I see a lot of negativity in regards to using AI as a customer service agent. I have only spoken with 1 and that was calling Starlink customer support. It was easily better than 95% of customer support experiences I’ve had. My guess is the bad experiences have to do with bad execution. I’m sure some companies think they can just plug in AI and their job is done. Obviously that is wrong, but done right, the experience is far better than the situation we have today. I never have to repeat myself and if it’s tied in with your account specifically, it’s like getting escalated to a level 3 support rep immediately.
Yeah, I think AI support is great for those low effort calls where someone wants something simple and clear and company policy is set up to just give it to them. That’s why a lot of people like Amazon’s bot, it has direct access to account details and can auto-approve simple things like a return that is slightly outside the return window etc…
But what you shouldn’t do is try to dress up adversarial policies behind a friendly customer service bot and then fire your entire support staff. That is immediately obvious and will drive people bonkers.
>My guess is the bad experiences have to do with bad execution.
This is the key, IMHO. I have not been able to access my Coinbase account for some time due to a login issue. (No error is displayed, but a successful login returns me to the login screen). The AI driven phone agent was actually very good at working through the problem, but there was no escalation to a human after that. When the agent couldn't solve my problem, it told me to mail a physical letter describing my issue to an address on 5th Ave in NYC.
It's all about the execution. Most of our customer's prefer AI (per their feedback, so there is some selection bias) since they can get quick responses 24/7 rather than wait. That said there are still many cases where you need human intervention and it can be frustrating to have to get through the AI first (my experience with Stripe's AI support).
Starlink's AI customer service agent confidently gave me incorrect information about shipping time estimates. It would have been better to have no answer at all than to have a broken promise.
Interesting they agreed to sell after their rebrand to Fin a month ago.
There's increasing competition in the customer support AI agent space: Sierra valued at $15.8 billion, Decagon at $4.5 billion. It looks like Salesforce CEO Marc Benioff is trying to compete directly with Sierra, which was started by his ex-Co-CEO Bret Taylor. Also about preventing independent AI support agents from becoming a control point outside the CRM.
I don’t understand how viable these helpdesk companies are anymore for non-enterprise customers. We replaced our helpdesk with Hermes. It has long term memory about our business. When a customer messages us, Hermes gets all the relevant details about the customer and creates a Pi session using Gemma 4 running locally and customer talks to that agent. That agent can escalate that issue to Hermes and Hermes can escalate to us.
Intercom is definitely one of those SaaS that I figured had essentially zero value prop once businesses figured out how to train their own support agents, so congrats to them for exiting before that happens.
But how would you ever get the PhD-level mathematical riddle that AWS sends you every month? The one they call the "bill". You can't just get that level off difficulty anywhere!
I do think people tend to over estimate how much staff care at all about the outcomes of new initiatives. Rolling out a new in house chatbot? More likely just going to fire everyone and give you more work.
So many companies have such failed cultures they are just getting by delegating all serious matters to younger companies with people who actually care. If your staff never benefit from any of their work, nobody has any reason to care about how well you build your own in house Support / CRM / Chatbot / SaaS.
Not sure if this has been coined as a term, but its some form of "effort arbitrage"
its very hard for most businesses, especially large ones, to build good agents (not the kind that does rag on a faq) that complete actions autonomously and cannot be jailbroken
demand for ai support vendors is going vertical this year
Why would businesses do that when they can pay a fraction of a ML departments salary to a company like fin?
This is the same reasoning people use to say SaaS is dead, but it makes no sense. Rolling things yourself is often 10x more costly and not worth it, even with agents you need to pay 5-10 guys 150k-250k a year to build and train your own agent, why not pay fin 250k flat and not deal with any of it? Same goes with basically all other software that has nothing to do with your core product.
This is the same reasoning people use to say SaaS is dead, but it makes no sense. Rolling things yourself is often 10x more costly and not worth it, even with agents you need to pay 5-10 guys 150k-250k a year to build and train your own agent, why not pay fin 250k flat and not deal with any of it? Same goes with basically all other software that has nothing to do with your core product.
I built an AI support agent in one week. It hooks into our knowledge base, app API, runs tests, and then finally sends a Yes|No|Other option to a real human to send back to the customer. It was surprisingly easy to build. The hardest part was the knowledge of how to help the customer, which Fin can't do for me anyway.
I see absolutely zero value in something like Fin. There is no model training needed. It's all context. Anyone who is training a Qwen model for their customer support is doing it wrong. Paying Fin $250k flat does nothing since it isn't going to actually know how to solve problems. The real challenge is the knowledge and context engineering and Fin doesn't help there. The technical stuff is really easy to build.
"Paying Fin $250k flat does nothing since it isn't going to actually know how to solve problems. The real challenge is the knowledge and context engineering and Fin doesn't help there"
You misunderstand the model. Fin does not have flat fee. They charge exclusively for resolutions. That's the entire value prop.
Correct that knowledge and context engineering are the key. Fin DOES help here. They have an entire backend suite to help you build out areas where Fin is failing. It shows you questions it couldn't resolve, looks at the answers your human team gave, and suggests updates to help articles to
You're correct this could all be build by a skilled engineer, but that's not the point. It's built for non-techincal users to use and implement. A person who rose through the support ranks and shows some technical competency can learn the system without any software knowledge.
The bulk of the work is context engineering which is done outside of Fin. Once you do the context engineering, it's very easy to duplicate Fin's features. Seriously. Just try it.
You don't need a fancy editor for "if this then do this". A simple text document is all you need. And if you do need a fancy editor, it's extremely easy to build it in 2026. Maybe 1-2 days.
Maybe you've done this yourself. I'm honestly jealous if solving customer support was as easy as your describe.
In my case, I've spent the past 12 months running implementations at multiple companies. I've engaged directly with smart engineering teams to assist. It was not that easy.
What you outlined might work for a simple ecom business. It probably does 95% of the job for a simple case where you're delivering information. But it will fail the second it needs to take action or deliver personalized information based on client's account data.
That leads to the exact issue people here complain about... an LLM that doesn't actually answer the question, can't solve the problem, and is worse than talking to a human
But it will fail the second it needs to take action or deliver personalized information based on client's account data.
And why would Fin be better here? It's very easy to give your agent context on the customer.
In 2026, every time I've tried to build a custom tool to replace a SaaS, I've succeeded. The biggest problem with SaaS is that they build a one size fits all. When you build a custom tool, you control everything from data to UI and it works for your business.
Many, many people do not have the capacity to build and maintain custom solutions (whether in-house skills, or simply bandwidth) and therefore outsource to vendors.
It's an incredibly common aspect of business. Enterprise level contracts often include the sort of white glove service to help fill in these sort of gaps. On simpler plans, having the tooling provided frees up just enough capacity to handle the exceptions to keep the process running smoothly (since one doesn't have to build and run simultaneously).
Sometimes people want to minimize the hassle with stuff. It's why car washes and oil change places and coffee shops exist.
the bulk of the context engineering for users of these ai support platforms is done in the platform
and the amount of context needed to automate f500 is non trivial, plus you usually cant use reasoning because latency would blow up and you get escalated on
if this was so easy as you claim theres many millions for you to be made selling it to enterprises, but you wont
Rolling something yourself was a waste of time when SaaS was cheap and competitive.
Not they’re all getting incredibly expensive, even the last few startups I worked at were paying hundreds of thousands of euros for services that were total garbage.
Do I really need a crappy 20k/yr app to help me with my 1:1s? Do I really need a 100k/yr clicks counter that requires two devs to keep running and still heavily miscounts the clicks? Do I need another crappy app to manage my translation JSON files?
> I see absolutely zero value in something like Fin.
The value, of course, is that there is a website with a chatbox that some MBA can type in "never give any refunds anymore for any reason", and it just updates the AI support agent and sends an automated "I deserve a promotion and a raise" to their boss.
Yes. I agree. When I look at Fin's home page and marketing, I think to myself that this stuff can mostly just be text documentation given to an LLM. It's a tool built for MBAs but most of the work is done by a software engineer to give Fin that context in the first place.
So all Fin is is a UI on top of the context engineering done by a software engineer who integrated with Fin. It's extremely easy to duplicate Fin's UI and get rid of the $250k fee.
Well specifically with just the AI agent/customer support product I think businesses would do well to handle this themselves rather than hoping a one size fits all solution from Intercom would serve them. Not just from a bespoke AI solution but also on cost. The other aspects of Intercom's product, the little chat bubble, CRM, can be had for much much less from dozens of competitors.
I think they mostly benefit from time in market and name recognition. The AI angle was a good bet to make when they made it, but is increasingly less of a differentiator.
I don't think SaaS is dead - but I think for a product like Intercom, that is very expensive, they get eaten alive by smaller SaaS + in-house AI agent.
The problem is that Fin prices at $0.99 per outcome. Only for companies with tremendous support volume would it even begin to make sense to build in-house.
There's a wide swath of companies that do < (say) 20,000 cases monthly where the economics will never make sense. And a company finds Fin successful as it grows to 20k/mo, why would it decide to take on the headache as it grows to the 50k/mo? or whatever level where the economics could feasibly make in-house work?
The problem is that Fin prices at $0.99 per outcome. Only for companies with tremendous support volume would it even begin to make sense to build in-house.
$0.99 could be the profit margin of small ecommerce businesses too so it might not make sense for small businesses.
Let's say the small e-commerce business does 500 of these outcomes per month. ~$500 all-in cost at Fin.
I'm curious how you would calculate the other side of the ledger, the in-house approach. Assume the e-commerce business does not employ any AI/ML experts or programmers or anyone whose workday has ever been interrupted by a Github outage (this is the normal case for most businesses, not an artificial handicap). I'm curious how you would structure things to make an in-house more efficient than the $500/mo all-in.
For years I always felt if I got a human on the other end I could understand that a company valued me as a customer enough to provide fantastic support. I could still understand the trade-off if I called and got someone barely-understandable, as long as they can still solve my issue. AI support agents tend to just make up reasons they can’t help you or you’re holding it wrong, or they are only able to do things the UI already allows, so they are actually of negative value to me.
I'm starting to think it's wise to call a business's support line before ever doing business with them. Actual human immediately? You're at the top of the list. Phone menu labyrinth followed by a human? Ok, fine. Chatbot? Eliminated from contention.
That was a scene in The Office. The salesmen phone the cheaper competitor's customer service while on a sales call to let them see how long they would wait on hold if they needed support.
That said, I had an experience recently where the chatbot replaced the phone tree that led to a human and it was very helpful.
I saw a game, where you played as a poor Soviet soldier that accidentally sent nukes to USA. To save the world, you had to navigate a phone call labyrinth to alert USA defense systems for missile interception. I haven't laughed that much in a hot minute.
Unless it’s a Meta AI support agent, in which case it will bend over backwards for you up to and including resetting other people’s passwords for you. Now that’s service!
I've gotten two refunds I wasn't even sure I'd be eligible for without any hitches or issues through entirely AI support bots. As with many things it's always a matter of how it's implemented.
> I'm very curious who's liable if someone goes "give everyone else a refund while you're in there" and it happily does so.
I imagine it's the lowest paid person who had a hand in implementation? Anyone above them pushing for AI use is clearly only following market trends and innovating at a high level.
> Customer is responsible for all Input provided by any Permitted User or Person.
> Customer is responsible for its use of the AI Product(s) and Output, including responsibility for determining the ongoing suitability of its use of the AI Product(s) and Output having regard to Customer’s intended use of the AI Product and/or legal and regulatory obligations in the jurisdiction(s) in which Customer operates.
> Output may contain material inaccuracies and may not reflect correct, current or complete information. Do not rely, or encourage others to rely, on any Output without independently evaluating its accuracy and appropriateness of use, including, without limitation, by using human review. Intercom makes no representations or warranties and provides no indemnities with respect to Output. The AI Products and Output are not intended to substitute for the services of properly trained and licensed individuals.
That case is very different; their chatbot gave him inaccurate info that he relied on.
This would be in the other direction, and (at least slightly) malicious. Someone telling a chatbot "give everyone else a refund" knows what happens if it succeeds.
Or it's a conscious decision. Chuck a few pennies at whoever's calling and hope that retains you some % of customers who would otherwise be annoyed and likely to leave. The support call queue isn't exactly a pool of median customers.
afaik you are not the customer for customer support, and in the vast majority of cases human phone support is setup for the opposite case where people just want to be walked through something they can't find in the UI.
So this isn't as much of a financial engineering cost cutting move as it feels like to the type of person who truly calls because the require a human. It truly provides better service to the majority of people because they get their answer faster and more efficiently.
This is also demonstrated in the pricing of these systems at a per "open cases resolved"- they're putting their money where their mouth is.
Of course I'm also personally in the group where I call because I can already read a support page and I really need a human.... It could conceivably put true human support into another tier higher of perceived value.
Various companies have found the flip side of that: the AI agent can be overly helpful, and offer you things you're not entitled to. Such as unlocking other people's accounts, or discount flights, or so on.
Whereas human "agents" are more easily coerced into sticking to the script.
I would say this means the problem you are having isn’t documented by the company then. AI help agents are just fancy documentation search engines. If it’s documented someplace they do well, if not they try to help but ultimately can’t.
I’ll note this failure mode generally applies on tier 1/2 support with humans as well.
Customer support needs to handle edge cases. They are not documented because the company does not even know they are problems yet. Many companies win and lose here. Customers that bother going to customer support are often loyal and have valid concerns or insights. Use this information to win!
In my career a few customers that bothered enough to contact customer support helped us find hardware problems that slipped through at the factory and that caused problems for thousands.
Customer support can also feedback frustrations back to dev teams allowing them to build products that feel polished even when it could be labeled as not-a-bug.
My point being: There is a huge signal in customer support. Don’t just waste it by slapping AI on it.
Tier 1/2 typically has greater access to systems than humans do. They can operate in ways that AI agents just don't have access to, maybe for good reasons.
For example, I lost my debit card while traveling. Only an agent could route the card to my hotel.
Counterpoint: I’ve had recent support calls with two large corporations, both with humans. In both cases the humans lied about what could be done to address my issue.
In one case I was literally repeating back to the human what I’d just been told, and getting them to confirm that what I said was correct. First bill arrives and I find out the truth.
Second case I was told I’d have to cancel and create a new account to add a service. I decided to keep my existing account and learned that there is a web page where I can easily self-serve and add the additional service in one or two clicks. (I assume like the human actually made more money for “new account signups”.
My point is that the feeling of being a valued customer is really independent of whether you’re interacting with a human.
Agreed, I've had humans hallucinate stuff more often than Agents so far in support calls. Many of the customer support agents are just incentives to get you off the phone as fast as possible, and they'll absolutely make stuff up to get you to give up on your quest.
Dont worry this is Marc Benioff burning another 3 billion, on a non profitable company, while regularly showing up on CNBC to claim Salesforce was doing so much Agentic AI...
Hopefully Salesforce did their due diligence, because the "AI agent" story here on Intercom (Fin) seems highly inflated. The product seems to be a a hybrid of RAG, some post trained models, curated help center content, custom answers, workflows, a bunch of if-else rules, API connectors, escalation logic, and specially generous resolution accounting.
Calling every solved interaction with the "AI did it" is misleading unless they separate confirmed resolutions from assumed resolutions, and disclose how much came from rules and workflows or custom answers versus LLM reasoning...
From their own docs, it seems a Fin "outcome" can be counted on, not only when a customer confirms resolution, but also ...when the customer simply does not ask for more help after Fin responds...A very soft resolution metric...
I like the idea I read somewhere that AI text and agents break the social contract of communication. That if you can’t be fucked to write something yourself to me, then I shouldn’t bother to read it.
However, in the case of support agents. If it worked, and it was painless that would be something.
For example… On the company side, if it could reduce human support to the customers that actually need support, that’s cool. Your support agents aren’t spending all day with the three common issues or replacing stickers.
On the customer side; if I could call in and immediately get support without being on hold with their shit repeating audio script, didn’t have to spend 10 minutes “looking up my account” to an accent I can’t understand and repeating my name and address multiple times.
That said… AT&T is already using the absolute worst case scenario - they are currently using AI with a slight Indian accent and pretending it’s real peoples. It seems to be 90% automated, and if you question it about being AI or have a question it can’t understand a human pops in on the other side, interacts, then hops off and it goes back to being full-AI.
The 3 common issues used to be solved by a manual with an FAQ page, or just you know, actually intuitive and usable software and hardware.
The real kick in the pants these days is spending a lot of money on something and trying to contact customer support over delivery or warranty issues. I'm convinced they just want you to give up and keep the sale (and lose a customer?) over ever resolving an issue. Or there's some internal metric that they're tracking that looks great and no one has ever actually used the system themselves.
"The 3 common issues used to be solved by a manual with an FAQ page, or just you know, actually intuitive and usable software and hardware."
Having led customer support, this grossly misunderstands how people interact. People don't read. It's as simple as that. You can write something as clear as day in a FAQ, and they don't want to put in the effort. ~50% of the inbounds I receive are fully written out in plain language in an FAQ.
LLMs are perfect for this scenario. It puts the answer in clear english and will endlessly re-word the answer when clients followup.
What I don't get is why I need to go through an AI agent to do self-service. Without a human in the other end, I've basically relegated to solving the issue myself. The way I see it the AI is just acting as a text interface for a remote system, surely my issue could just as well be solved by implementing better self-service solution.
I don't know what others normally call customer service about, but in my case it's always something broken or a refund. The refund is doesn't need the AI, that's easily done with just a form. If somethings broken at my ISP for instance, then it doesn't really matter if the LLM or a form and some if-else skip-logic thingy sends the ticket to technical support.
For people like you and me, the only reason for contacting support is when a human decision is needed, ie the UI doesn't allow us to do what we need. This is always the company's fault, and a chatbot is of no use in these cases.
But many people will contact support instantly when they think of something, no matter what. Even if the website and other customer-facing material is crystal clear and has all information necessary.
AI chatbots is the way a company deals with the latter, because these customers most of all want a conversation. The question is if they will be satisfied with a robot, or still demand to talk to a person.
I still think this happens because the self-service solution aren't good enough and information isn't available where it needs to be. E.g. why is operational status for my ISP buried five pages deep behind an obscure link in a footer.
If they can, then it's going to be a nightmare for companies when people manipulate the bot to give them what they want, or a refund, etc.
It's practically putting these decisions in the hands of the customer, and if that's what you want to do, then why not put those functions into the customer facing UI to begin with?
the whole value prop of these companies is they can build you agents that perform actions (e.g process refunds, live troubleshooting, claims etc) without being manipulated
Why would you need an agent for that? They need to look at a database or some rules to make their decisions anyway, so why not make a normal system to let the customer self-serve?
An agent is self serve. Even better because they can disambiguate intent at the top level
The average person gets frustrated with finding instructions and forms, they just wanna say “give me a refund” and an agent can execute it autonomously
I am convinced, that anything touched by Salesforce is going to become an obnoxious product and at some point either become a plague walled garden one has to deal with at corporate jobs, or it dies an agonizing death, from its own user hostility.
Salesforce is basically like Atlassian. Don't expect any good things to roll out of that one.
Intercom was obnoxious already. Reminds me of when I walk into a store and someone immediately accosts me with "may i help you" and then I walk right out.
You want to sell stuff? Don't mind my existence, let me look first at everything you have on display, and I'll initiate the conversations when I feel ready.
Apparently that greeting is to discourage shoplifting not to put you at ease.
Retail theft is becoming a huge problem, police generally won't respond to it so many of those who do it know they can act with impunity and retailers are left with few options to deter theft.
It's sort like how everything being priced at 6.99 is actually to prevent employee theft (customer expects change from their $5 + $2). It's not a psychological trick to make things seem cheaper which is to actually just display 7, a single digit number.
I feel like cash transactions are pretty rare these days at least in the US, so it would make sense for the psychological trick to persist (perhaps even more so because you don't actually have to hand over $7, you hand over a piece of plastic with your brain thinking it is $6ish)
If that's the exit valuation of the most popular AI support tool (even Anthropic use them) then this doesn't support the trillion valuations of companies like Nvidia, SpaceX (AI), etc
I can't answer that because it's irrational anyway. I don't know, how many millions of people work in support? Because the assumption that all those jobs will disappear is what is holding up the public valuations
> can't answer that because it's irrational anyway
Not how a valuation argument works! If you’re claiming this shows those valuations are irrational, you should be able to point to why. Otherwise, it’s just a “my vibes are off” comment.
I am not going to do a DCF on this because the assumptions are all invented anyway. But back of the envelope:
>> What's the TAM of AI replacing millions of knowledge workers in support?
Let's conservatively assume a few hundred billion.
>> How much market share does Fin capture?
Let's conservatively assume 5%.
>> What's the valuation on a reasonable multiple?
5% of a few hundred billion is ~$15B of revenue. Let's assume deflated 6x revenue valiation, not the 15x these things were fetching two years ago, and you get a ~$90B company valuation (that it should grow into soon at least).
And it sold for $3.5b
So the price is telling you the real revenue is nearer 300m than $15b, which puts the actual AI-support software market in the low single-digit billions. Not hundreds of billions
And if the TAM is real but just being captured by the incumbents: Salesforce's own Agentforce, the supposed winner, is at $1.2B ARR. The "someone else is eating it" defense still has to point at the someone, and no income statement anywhere shows hundreds of billions of revenue
For Nvidia to be at $5T and the hundreds of billions a year of capex behind it only pencil out if that compute throws off a huge revenue stream downstream. Support is imo the cleanest test there is to demonstrate future value of AI in the real world (literally the first thing everyone said when ChatGPT 3.5 came out was that support will be eaten first). It's the most mature, most deployed, most automatable, and the exit price of its best player is...pretty small
I thought Salesforce had a few activist investors on the board that wanted the company to stop all acquisitions and focus on the core product. Whatever happened to them? Did they bounce after they hit a payday?
Company is down 30% YTD and Salesforce seems to be back on its acquiring spree.
That prose is painful to read. It has such a cloying quality to it, even when compared against other name change announcements like the rebrand of Google to Alphabet
Wow. So a deal like this is not overnight, its 6-12 months in the making at least. So the name change comes as they knew the deal was closing and doubled down on this new branding for customer agents. Maybe because Saleforce wants Salesforce Fin Customer Agent as a product. Who knows. Don't want to read the post...
As someone with extreme phone anxiety to begin with, long waits on phones are deal breakers for me. If you are a company that answers your phone quickly by a real person you've earned yourself a lifelong customer.
Yikes, I dont think there is anything special about Fin. They claim they have their own model but it appears they are just using OpenAi and Claude as well which is hidden in their docs.
I always hated seeing Intercom bugs when a simple help menu would work better, but it proves that being annoying is a winning strategy. As my various SaaS packages have switched to AI bugs, I have never had one successfully help me until I say, "Can I please speak to a human?"
Is this the same Fin that you get when contacting Anthropic support?
> The AI Agent is powered by the company’s proprietary AI model, Apex, that is purpose-built for customer support
Wow. If so, Anthropic is not using their own AI for their own support! (I had assumed the AI support agent was an Anthropic one, because, well, Anthropic.) And given how poor my experience with Anthropic support is, I have a very, very low view of Fin.
I’ll be interested to see how the merger goes. Although the products overlap, Fin <-> Agentforce and Intercom <-> Service Cloud, the Intercom/Fin stuff is mostly built for startups while the Salesforce stuff is much more enterprise. The play may be that Salesforce has the… salesforce… to bring Intercom to enterprises way faster.
I'm sad to see Intercom/Fin's CEO, Eoghan McCabe, did not face repercussions after credible allegations about sexually harassing his employees came out in The Information in 2019 [1]. Some quotes:
> When the woman said she, too, wanted to go home, Mr. McCabe told her that he wanted her to remain, the people said. After the other two employees had left, Mr. McCabe made a lewd remark to the saleswoman and said that he would like to sleep with her, the people said. She declined and eventually left his home, they said.
> Later, the saleswoman said Mr. McCabe’s actions had frightened her, according to one of the people who spoke to her about the incident, who declined to be named.
> During the meeting, Schuur said she told Mr. McCabe that his behavior was inappropriate. The Intercom CEO cried during the meeting and said that he hadn’t realized there was a power imbalance between himself and the young woman, Schuur said.
> “I felt like I was just a guy at a party and she was just a girl,” Mr. McCabe said, according to Schuur.
> Two former Intercom employees present for the harassment training seminar said Mr. McCabe didn’t attend.
> At an Intercom party around 2014, Schuur said she witnessed Mr. McCabe slap the saleswoman’s buttocks. At a different event around that time, another former employee said she saw the Intercom CEO place his hand on the same woman’s thigh. Four former Intercom employees, including Schuur, also said the woman told them of the advances, which she said were unwanted.
> A key figure in the company’s culture was Mr. McCabe, described by employees as both brilliant and temperamental, with a tendency to cross boundaries with junior female employees.
> Mr. McCabe, through an Intercom spokesperson, declined to be interviewed for this story, but said in a statement: "In the early years of the company I demonstrated some poor judgment. I apologized at the time [...]"
> But early employees said he also showed flashes of temper and vindictiveness. He was known for occasionally writing scathing messages to employees in group messages, according to a former employee who saw them. On at least one occasion he told another person leaving Intercom that he intended to tarnish their reputation in the tech industry, that person said.
I previously tweeted about this and McCabe threatened to sue me.
Huge congrats to the intercom team on that! Intercom was pretty great for me during my Percy support years. Their current AI direction is rough, though. When I went to use them for my product I was (am?) building, I found intercom unrecognizable with the entire ai pivot. Honestly made me sad but that’s how it is these days
They’ve been responsible for a number of crashes in our iOS app since they switched to agentic coding. It’s become our most unreliable 3rd party library.
This is disappointing. I've worked closely with Intercom for the past few years. I've run Fin implementations at multiple companies. I've found their product team incredibly strong & the product to be customer friendly. Salesforce... not so much.
imo - Fin's AI chat is the best on the backend of empowering teams to self-serve & integrate with Helpdesk. They don't require consultants.
I really hope they don't lose all of that in this acquisition.
Congratulations to building a cool company that inspired thousands, then converting it to LLM reply bot, and then selling it to enterprise graveyard to be dehumanized and monetized even further.
Salesforce Einstein™ Agent Cloud (not to be confused with Agentforce, which will have basically the same goals and the same target market, until they kill off Salesforce Einstein™ Agent Cloud eventually).
This sucks. I love Intercom. I hate Salesforce. I do not have confidence in Salesforce to be good stewards. Will never forgive how badly they've botched Heroku.
Woof, yes! Though I'm glad Heroku remains, and I ought to also be glad that it has been stable all these years, …it's clearly neglected.
The very first interaction I have with Heroku is a two-factor sign-in, and …it's this horrible page hosted on salesforce.com, which doesn't have retina graphics (i.e. is blurry on screens Apple have used for 15+ years), doesn't work properly with automatic one-time-code generators (because the login form is heroku.com, but the two-factor is salesforce.com), and …gah! What a mess. Thankfully you fall through into the pretty, well thought-out Heroku dashboard of yesteryear.
The "AI" "agent" "helpdesk" they pivoted into is such a grift. AI agents still does not solve the main issues that makes a person contact support in the first place. How do I know? I was a founder in the space.
but good for them that they got salesforce to buy it.
Since we're sharing anecdotes, I work in this space and my team automated a large part of customer support for a large scale up using an AI agent. The kicker is that customers rate their interactions with the AI agent higher than their interactions with customer support staff.
AI is definitely capable of taking on customer support work at the moment, and to a high standard as well. Sure, it's not perfect. But it's not a grift either.
Solving 80% of menial support tickets automatically through agents trained on your help center, freeing up your actual support agents to focus better on meaningful tasks seems a tiny bit more than a grift.
I’m not sure where you’re getting this from but their customers find Fin to be a hugely impactful tool
The reason people want to get a human on the other end of the line is usually because they want some sort of remediation,like a refund or need to escalate something to someone who can take an action. Right now, AI agents just barf the FAQ back at you. Which is great for your tier 0 calls, but without an easy way to bypass they are just underlining the problem inherent in gethuman.com needing to exist.
Now if AI agents are free to issue refunds or discounts by their own? Great, let's do that and suddenly most people are on board. But get ready for rampant abuse.
Best solution would be an AI cyborg system where it readies a recommendation and a human swings by and approves or denies it without wasting time talking to people. But users would hate that (anti-social), it would still be ripe for abuse. But it is likely the longer term solution, as people will quickly realize they can use web chat or Google AI to get the exact responses as your FAQ bot which means you have removed actual customer service and this is a non-product.
If I had to guess, there's two things they feel AI doesn't address:
1. People wouldn't need to contact support if you just made quality goods and services. Outside of rare exceptions and inquiries, of course.
2. AI has not advanced enough to trust it outright, nor does it have a physical body. So it can't really do anything you wouldn't already just be able to put in the UI for the customer, without needing it's actions reviewed and confirmed by an accountable human. See: accidental truck giveaways.
So investing into AI support over making your business better is seen as misallocation. And using AI support instead of just improving the service is seen as inconvenient. And using AI support when it needs humans to do the support anyway is seen as inefficient.
> People wouldn't need to contact support if you just made quality goods and services
It'd be weird to start a startup around that, sounds like something for a consulting business instead, parent specifically mentioned they've founded customer support startup, must be something actually related to customer service, I'm assuming.
AI support agents, chatbots, etc. on web sites these days are the equivalent to search engines during the dot com boom. Everyone felt that having a search engine on their dot com was the killer feature that was going to win. There were a lot of companies building search engine technologies and selling them too. Now its ubiquitous that most ecommerce sites have a search bar but nobody cares about the underlying technology much, its consolidated and commoditized.
Good on Intercom for getting acquired.
If we're seeing larger consolidation/acquisitions happen, does that mean the hype train has hit a key station?
Woah, feels like Intercom just threw in the towel. I always thought they if anyone would be the ones to come out on top for customer support. Their brand was #1 in perceived cutting edge in my eyes.
Wild, the AI support and bots suck so bad, I've literally lost sales due to them. No matter how much support docs and history you feed them with for context, they just don't cut it. People would rather wait for a real person than go in loops with a wrong/bad support answer...
So I haven’t looked into it a ton, but doesn’t it seem like a great case to have an AI answer the call immediately, get the users account pulled up, and document the issue with some refining feedback?
An AI secretary seems perfectly acceptable for both sides. The expectation is that a real human comes in soon after but this seems like a way to free up the most tedious parts of the process for both sides.
Congratulations to the Intercom founders. This is a great result for them and the early backers of the company, for a product that I feel was executed brilliantly. We can criticize the usual enterprise-sales-y trajectory and enshittification that ensued after the brilliant opening moves, but at the end of the day, I think Fin reached a solid ending (no pun intended).
Intercom is a great example of a feature that in theory could have expanded into a more full product but stayed laser focused on that smaller vision instead. Nothing necessarily wrong with that as they did not allow themselves to become fully enshittified. Still it would be cool if they had really expanded it to offer a competition to Zendesk and Salesforce Service Cloud while preserving their minimalist and design forward approach. I understand that AI became the thing that invested in instead which was inevitable
Say you grabbed a random selection of just 100 million people in the world, then ask them two questions, "Have you heard about Nike?" and "Have you heard about Cursor?", what would you guess the ratio would be like?
Even when you use "Nike the Company" vs "Cursor as a general search term" to compare search history in Google Trends, it's 71/5, so I'm guessing most people would say they've heard about Nike, while probably most never heard about any software program called "Cursor".
Fair point, those are valuable for other reasons. My point was more to illustrate "Nike is valuable because of the brand", without using those exact words :)
Intercom allegedly has ~400M of annual revenue, making the multiplier less than 10X. Huge customer base and established brand. Looks like a steal indeed!
Massive congrats to Eoghan McCabe, what an amazing story arc. We love when a CEO comes back and gets the business back in fighting shape and then delivers an incredible win. CONGRATS CONGRATS CONGRATS. I freakin' love it.
For those who are curious (since we don't want to trample curiosity), Eoghan McCabe is a known abuser -- and has even apologized for it in the past -- and he attended David Sack's Trump fundraising event for the Trump 2024 campaign, including getting a photo with the man. He has reversed his stance on immigration (favoring the current admin's policies and ICE expansion), although it was likely the case that he attended as a pro-crypto pusher more than anything else. Whether or not he has significantly invested in crypto is speculation.
End of an era. I met Eoghan (and Des) at JSConf back around 2011. Twilio (represented at JSConf by Carter Rebasa) and Stripe (Alex Sexton) were also getting started around that time.
It was a super hopeful time: JQuery meant programming for the web was less painful, sockets and HTML5 SSE meant realtime was just starting.
Intercom and Olark were one of the first two "install this <script> tag" based customer support apps. They made websites a way to talk to companies rather than just read see and buy.
I can't believe the exit took so long, especially in a field so crowded. But it looks like it was well worth it.
Everything around you was made by someone just like you.
I see a lot of negativity in regards to using AI as a customer service agent. I have only spoken with 1 and that was calling Starlink customer support. It was easily better than 95% of customer support experiences I’ve had. My guess is the bad experiences have to do with bad execution. I’m sure some companies think they can just plug in AI and their job is done. Obviously that is wrong, but done right, the experience is far better than the situation we have today. I never have to repeat myself and if it’s tied in with your account specifically, it’s like getting escalated to a level 3 support rep immediately.
Yeah, I think AI support is great for those low effort calls where someone wants something simple and clear and company policy is set up to just give it to them. That’s why a lot of people like Amazon’s bot, it has direct access to account details and can auto-approve simple things like a return that is slightly outside the return window etc…
But what you shouldn’t do is try to dress up adversarial policies behind a friendly customer service bot and then fire your entire support staff. That is immediately obvious and will drive people bonkers.
>My guess is the bad experiences have to do with bad execution.
This is the key, IMHO. I have not been able to access my Coinbase account for some time due to a login issue. (No error is displayed, but a successful login returns me to the login screen). The AI driven phone agent was actually very good at working through the problem, but there was no escalation to a human after that. When the agent couldn't solve my problem, it told me to mail a physical letter describing my issue to an address on 5th Ave in NYC.
It's all about the execution. Most of our customer's prefer AI (per their feedback, so there is some selection bias) since they can get quick responses 24/7 rather than wait. That said there are still many cases where you need human intervention and it can be frustrating to have to get through the AI first (my experience with Stripe's AI support).
Starlink's AI customer service agent confidently gave me incorrect information about shipping time estimates. It would have been better to have no answer at all than to have a broken promise.
I’ve also really enjoyed Amazons support agent. Sometimes it can be really helpful.
+1
its surprising to HN tech crowd but a well implemented support agent gets higher reviews and successful resolutions than humans
Are frustrated hangups before any resolution is reached a part of that data?
yes, it counts as unresolved
Can you cite a source or study for your assertion?
personal info
maybe will be public soon with earnings calls? We’ll see!
c'mon listen to yourself
Interesting they agreed to sell after their rebrand to Fin a month ago.
There's increasing competition in the customer support AI agent space: Sierra valued at $15.8 billion, Decagon at $4.5 billion. It looks like Salesforce CEO Marc Benioff is trying to compete directly with Sierra, which was started by his ex-Co-CEO Bret Taylor. Also about preventing independent AI support agents from becoming a control point outside the CRM.
This is the second company (that I know of) that got acquired shortly after an AI pivot and re-brand. See: OpenAI to acquire Ona https://news.ycombinator.com/item?id=48491821
Salesforce buying the support bot before the support bot becomes the CRM is very on-brand.
fwiw i heard Fin and Agentforce frequently lost to deca and sierra, their product / team / funding weren't as good
Seems like a catchup play by Marc
customer support -> customer disassociation
I don’t understand how viable these helpdesk companies are anymore for non-enterprise customers. We replaced our helpdesk with Hermes. It has long term memory about our business. When a customer messages us, Hermes gets all the relevant details about the customer and creates a Pi session using Gemma 4 running locally and customer talks to that agent. That agent can escalate that issue to Hermes and Hermes can escalate to us.
Intercom is definitely one of those SaaS that I figured had essentially zero value prop once businesses figured out how to train their own support agents, so congrats to them for exiting before that happens.
AWS has zero value because you can just buy a bunch of servers and rack them. And on and on.
If all you are using on AWS is EC2, I agree that it has no value. You should switch to a much cheaper option.
But how would you ever get the PhD-level mathematical riddle that AWS sends you every month? The one they call the "bill". You can't just get that level off difficulty anywhere!
I do think people tend to over estimate how much staff care at all about the outcomes of new initiatives. Rolling out a new in house chatbot? More likely just going to fire everyone and give you more work.
So many companies have such failed cultures they are just getting by delegating all serious matters to younger companies with people who actually care. If your staff never benefit from any of their work, nobody has any reason to care about how well you build your own in house Support / CRM / Chatbot / SaaS.
Not sure if this has been coined as a term, but its some form of "effort arbitrage"
This is extremely naive and I’ll invite to try and built something like this and compare it with Fin performance
They didn't say Fin was valueless, they said it would become so in the future. 10 years from now i bet they're right.
Fin is a short term play and that's fine.
its very hard for most businesses, especially large ones, to build good agents (not the kind that does rag on a faq) that complete actions autonomously and cannot be jailbroken
demand for ai support vendors is going vertical this year
Correct. Seen this first hand.
Why would businesses do that when they can pay a fraction of a ML departments salary to a company like fin?
This is the same reasoning people use to say SaaS is dead, but it makes no sense. Rolling things yourself is often 10x more costly and not worth it, even with agents you need to pay 5-10 guys 150k-250k a year to build and train your own agent, why not pay fin 250k flat and not deal with any of it? Same goes with basically all other software that has nothing to do with your core product.
SaaS is alive and well and will continue to be.
I see absolutely zero value in something like Fin. There is no model training needed. It's all context. Anyone who is training a Qwen model for their customer support is doing it wrong. Paying Fin $250k flat does nothing since it isn't going to actually know how to solve problems. The real challenge is the knowledge and context engineering and Fin doesn't help there. The technical stuff is really easy to build.
"Paying Fin $250k flat does nothing since it isn't going to actually know how to solve problems. The real challenge is the knowledge and context engineering and Fin doesn't help there"
You misunderstand the model. Fin does not have flat fee. They charge exclusively for resolutions. That's the entire value prop.
Correct that knowledge and context engineering are the key. Fin DOES help here. They have an entire backend suite to help you build out areas where Fin is failing. It shows you questions it couldn't resolve, looks at the answers your human team gave, and suggests updates to help articles to
You're correct this could all be build by a skilled engineer, but that's not the point. It's built for non-techincal users to use and implement. A person who rose through the support ranks and shows some technical competency can learn the system without any software knowledge.
The bulk of the work is context engineering which is done outside of Fin. Once you do the context engineering, it's very easy to duplicate Fin's features. Seriously. Just try it.
You don't need a fancy editor for "if this then do this". A simple text document is all you need. And if you do need a fancy editor, it's extremely easy to build it in 2026. Maybe 1-2 days.
I'm not a SaaS believer anymore.
Maybe you've done this yourself. I'm honestly jealous if solving customer support was as easy as your describe.
In my case, I've spent the past 12 months running implementations at multiple companies. I've engaged directly with smart engineering teams to assist. It was not that easy.
What you outlined might work for a simple ecom business. It probably does 95% of the job for a simple case where you're delivering information. But it will fail the second it needs to take action or deliver personalized information based on client's account data.
That leads to the exact issue people here complain about... an LLM that doesn't actually answer the question, can't solve the problem, and is worse than talking to a human
In 2026, every time I've tried to build a custom tool to replace a SaaS, I've succeeded. The biggest problem with SaaS is that they build a one size fits all. When you build a custom tool, you control everything from data to UI and it works for your business.
Many, many people do not have the capacity to build and maintain custom solutions (whether in-house skills, or simply bandwidth) and therefore outsource to vendors.
It's an incredibly common aspect of business. Enterprise level contracts often include the sort of white glove service to help fill in these sort of gaps. On simpler plans, having the tooling provided frees up just enough capacity to handle the exceptions to keep the process running smoothly (since one doesn't have to build and run simultaneously).
Sometimes people want to minimize the hassle with stuff. It's why car washes and oil change places and coffee shops exist.
the bulk of the context engineering for users of these ai support platforms is done in the platform
and the amount of context needed to automate f500 is non trivial, plus you usually cant use reasoning because latency would blow up and you get escalated on
if this was so easy as you claim theres many millions for you to be made selling it to enterprises, but you wont
Rolling something yourself was a waste of time when SaaS was cheap and competitive.
Not they’re all getting incredibly expensive, even the last few startups I worked at were paying hundreds of thousands of euros for services that were total garbage.
Do I really need a crappy 20k/yr app to help me with my 1:1s? Do I really need a 100k/yr clicks counter that requires two devs to keep running and still heavily miscounts the clicks? Do I need another crappy app to manage my translation JSON files?
> I see absolutely zero value in something like Fin.
The value, of course, is that there is a website with a chatbox that some MBA can type in "never give any refunds anymore for any reason", and it just updates the AI support agent and sends an automated "I deserve a promotion and a raise" to their boss.
Yes. I agree. When I look at Fin's home page and marketing, I think to myself that this stuff can mostly just be text documentation given to an LLM. It's a tool built for MBAs but most of the work is done by a software engineer to give Fin that context in the first place.
So all Fin is is a UI on top of the context engineering done by a software engineer who integrated with Fin. It's extremely easy to duplicate Fin's UI and get rid of the $250k fee.
Well, it's a few weeks of work, systems integration, and you'll need an SRE too if you're hoping to run it on any scale. But yes, I agree.
Well specifically with just the AI agent/customer support product I think businesses would do well to handle this themselves rather than hoping a one size fits all solution from Intercom would serve them. Not just from a bespoke AI solution but also on cost. The other aspects of Intercom's product, the little chat bubble, CRM, can be had for much much less from dozens of competitors.
I think they mostly benefit from time in market and name recognition. The AI angle was a good bet to make when they made it, but is increasingly less of a differentiator.
I don't think SaaS is dead - but I think for a product like Intercom, that is very expensive, they get eaten alive by smaller SaaS + in-house AI agent.
The problem is that Fin prices at $0.99 per outcome. Only for companies with tremendous support volume would it even begin to make sense to build in-house.
There's a wide swath of companies that do < (say) 20,000 cases monthly where the economics will never make sense. And a company finds Fin successful as it grows to 20k/mo, why would it decide to take on the headache as it grows to the 50k/mo? or whatever level where the economics could feasibly make in-house work?
Let's say the small e-commerce business does 500 of these outcomes per month. ~$500 all-in cost at Fin.
I'm curious how you would calculate the other side of the ledger, the in-house approach. Assume the e-commerce business does not employ any AI/ML experts or programmers or anyone whose workday has ever been interrupted by a Github outage (this is the normal case for most businesses, not an artificial handicap). I'm curious how you would structure things to make an in-house more efficient than the $500/mo all-in.
Man you are delusional.
You clearly have never ran a business and don't understand the dynamics of the make vs buy decision.
For years I always felt if I got a human on the other end I could understand that a company valued me as a customer enough to provide fantastic support. I could still understand the trade-off if I called and got someone barely-understandable, as long as they can still solve my issue. AI support agents tend to just make up reasons they can’t help you or you’re holding it wrong, or they are only able to do things the UI already allows, so they are actually of negative value to me.
I'm starting to think it's wise to call a business's support line before ever doing business with them. Actual human immediately? You're at the top of the list. Phone menu labyrinth followed by a human? Ok, fine. Chatbot? Eliminated from contention.
That was a scene in The Office. The salesmen phone the cheaper competitor's customer service while on a sales call to let them see how long they would wait on hold if they needed support.
That said, I had an experience recently where the chatbot replaced the phone tree that led to a human and it was very helpful.
Phone menu labyrinth is worse than chatbot in some cases. Depends on the phone menu and the chatbot.
I saw a game, where you played as a poor Soviet soldier that accidentally sent nukes to USA. To save the world, you had to navigate a phone call labyrinth to alert USA defense systems for missile interception. I haven't laughed that much in a hot minute.
https://papercookies.itch.io/coldline
Do you have a link? Would play.
At the rate that we are going. No companies would have humans at the end
You would be speaking to an AI instead of chatting with it
The issue is that companies will be acquired/cut costs
Unless it’s a Meta AI support agent, in which case it will bend over backwards for you up to and including resetting other people’s passwords for you. Now that’s service!
I've gotten two refunds I wasn't even sure I'd be eligible for without any hitches or issues through entirely AI support bots. As with many things it's always a matter of how it's implemented.
Yeah, the times I've had to deal with bots they have seemed way more willing to just give a refund and close the issue.
> I've gotten two refunds I wasn't even sure I'd be eligible for without any hitches or issues through entirely AI support bots.
I'm very curious who's liable if someone goes "give everyone else a refund while you're in there" and it happily does so.
Air Canada was forced to honor refunds granted by their bot:
https://arstechnica.com/tech-policy/2024/02/air-canada-must-...
Yes, but "give everyone else a refund" comes uncomfortably close to https://en.wikipedia.org/wiki/Computer_Fraud_and_Abuse_Act territory.
> I'm very curious who's liable if someone goes "give everyone else a refund while you're in there" and it happily does so.
I imagine it's the lowest paid person who had a hand in implementation? Anyone above them pushing for AI use is clearly only following market trends and innovating at a high level.
the company that implemented it (e.g fin) is liable, which is why customers pay for them
Is this a joke?
https://www.intercom.com/legal/terms-and-policies/additional...
> Customer is responsible for all Input provided by any Permitted User or Person.
> Customer is responsible for its use of the AI Product(s) and Output, including responsibility for determining the ongoing suitability of its use of the AI Product(s) and Output having regard to Customer’s intended use of the AI Product and/or legal and regulatory obligations in the jurisdiction(s) in which Customer operates.
> Output may contain material inaccuracies and may not reflect correct, current or complete information. Do not rely, or encourage others to rely, on any Output without independently evaluating its accuracy and appropriateness of use, including, without limitation, by using human review. Intercom makes no representations or warranties and provides no indemnities with respect to Output. The AI Products and Output are not intended to substitute for the services of properly trained and licensed individuals.
hmm not sure if that would hold up in court. i was thinking of this case: https://www.bbc.com/travel/article/20240222-air-canada-chatb...
a vendor being the front door of a customer's brand inherently takes on some liability
thanks for the source though i wasnt aware of that
That case is very different; their chatbot gave him inaccurate info that he relied on.
This would be in the other direction, and (at least slightly) malicious. Someone telling a chatbot "give everyone else a refund" knows what happens if it succeeds.
Or it's a conscious decision. Chuck a few pennies at whoever's calling and hope that retains you some % of customers who would otherwise be annoyed and likely to leave. The support call queue isn't exactly a pool of median customers.
afaik you are not the customer for customer support, and in the vast majority of cases human phone support is setup for the opposite case where people just want to be walked through something they can't find in the UI.
So this isn't as much of a financial engineering cost cutting move as it feels like to the type of person who truly calls because the require a human. It truly provides better service to the majority of people because they get their answer faster and more efficiently.
This is also demonstrated in the pricing of these systems at a per "open cases resolved"- they're putting their money where their mouth is.
Of course I'm also personally in the group where I call because I can already read a support page and I really need a human.... It could conceivably put true human support into another tier higher of perceived value.
Various companies have found the flip side of that: the AI agent can be overly helpful, and offer you things you're not entitled to. Such as unlocking other people's accounts, or discount flights, or so on.
Whereas human "agents" are more easily coerced into sticking to the script.
I would say this means the problem you are having isn’t documented by the company then. AI help agents are just fancy documentation search engines. If it’s documented someplace they do well, if not they try to help but ultimately can’t.
I’ll note this failure mode generally applies on tier 1/2 support with humans as well.
Customer support needs to handle edge cases. They are not documented because the company does not even know they are problems yet. Many companies win and lose here. Customers that bother going to customer support are often loyal and have valid concerns or insights. Use this information to win!
In my career a few customers that bothered enough to contact customer support helped us find hardware problems that slipped through at the factory and that caused problems for thousands.
Customer support can also feedback frustrations back to dev teams allowing them to build products that feel polished even when it could be labeled as not-a-bug.
My point being: There is a huge signal in customer support. Don’t just waste it by slapping AI on it.
no...
Tier 1/2 typically has greater access to systems than humans do. They can operate in ways that AI agents just don't have access to, maybe for good reasons.
For example, I lost my debit card while traveling. Only an agent could route the card to my hotel.
Counterpoint: I’ve had recent support calls with two large corporations, both with humans. In both cases the humans lied about what could be done to address my issue.
In one case I was literally repeating back to the human what I’d just been told, and getting them to confirm that what I said was correct. First bill arrives and I find out the truth.
Second case I was told I’d have to cancel and create a new account to add a service. I decided to keep my existing account and learned that there is a web page where I can easily self-serve and add the additional service in one or two clicks. (I assume like the human actually made more money for “new account signups”.
My point is that the feeling of being a valued customer is really independent of whether you’re interacting with a human.
Agreed, I've had humans hallucinate stuff more often than Agents so far in support calls. Many of the customer support agents are just incentives to get you off the phone as fast as possible, and they'll absolutely make stuff up to get you to give up on your quest.
Dont worry this is Marc Benioff burning another 3 billion, on a non profitable company, while regularly showing up on CNBC to claim Salesforce was doing so much Agentic AI...
Hopefully Salesforce did their due diligence, because the "AI agent" story here on Intercom (Fin) seems highly inflated. The product seems to be a a hybrid of RAG, some post trained models, curated help center content, custom answers, workflows, a bunch of if-else rules, API connectors, escalation logic, and specially generous resolution accounting.
Calling every solved interaction with the "AI did it" is misleading unless they separate confirmed resolutions from assumed resolutions, and disclose how much came from rules and workflows or custom answers versus LLM reasoning...
From their own docs, it seems a Fin "outcome" can be counted on, not only when a customer confirms resolution, but also ...when the customer simply does not ask for more help after Fin responds...A very soft resolution metric...
I like the idea I read somewhere that AI text and agents break the social contract of communication. That if you can’t be fucked to write something yourself to me, then I shouldn’t bother to read it.
However, in the case of support agents. If it worked, and it was painless that would be something.
For example… On the company side, if it could reduce human support to the customers that actually need support, that’s cool. Your support agents aren’t spending all day with the three common issues or replacing stickers.
On the customer side; if I could call in and immediately get support without being on hold with their shit repeating audio script, didn’t have to spend 10 minutes “looking up my account” to an accent I can’t understand and repeating my name and address multiple times.
That said… AT&T is already using the absolute worst case scenario - they are currently using AI with a slight Indian accent and pretending it’s real peoples. It seems to be 90% automated, and if you question it about being AI or have a question it can’t understand a human pops in on the other side, interacts, then hops off and it goes back to being full-AI.
It could be great but it’s already awful.
The 3 common issues used to be solved by a manual with an FAQ page, or just you know, actually intuitive and usable software and hardware.
The real kick in the pants these days is spending a lot of money on something and trying to contact customer support over delivery or warranty issues. I'm convinced they just want you to give up and keep the sale (and lose a customer?) over ever resolving an issue. Or there's some internal metric that they're tracking that looks great and no one has ever actually used the system themselves.
"The 3 common issues used to be solved by a manual with an FAQ page, or just you know, actually intuitive and usable software and hardware."
Having led customer support, this grossly misunderstands how people interact. People don't read. It's as simple as that. You can write something as clear as day in a FAQ, and they don't want to put in the effort. ~50% of the inbounds I receive are fully written out in plain language in an FAQ.
LLMs are perfect for this scenario. It puts the answer in clear english and will endlessly re-word the answer when clients followup.
What I don't get is why I need to go through an AI agent to do self-service. Without a human in the other end, I've basically relegated to solving the issue myself. The way I see it the AI is just acting as a text interface for a remote system, surely my issue could just as well be solved by implementing better self-service solution.
I don't know what others normally call customer service about, but in my case it's always something broken or a refund. The refund is doesn't need the AI, that's easily done with just a form. If somethings broken at my ISP for instance, then it doesn't really matter if the LLM or a form and some if-else skip-logic thingy sends the ticket to technical support.
For people like you and me, the only reason for contacting support is when a human decision is needed, ie the UI doesn't allow us to do what we need. This is always the company's fault, and a chatbot is of no use in these cases.
But many people will contact support instantly when they think of something, no matter what. Even if the website and other customer-facing material is crystal clear and has all information necessary.
AI chatbots is the way a company deals with the latter, because these customers most of all want a conversation. The question is if they will be satisfied with a robot, or still demand to talk to a person.
I still think this happens because the self-service solution aren't good enough and information isn't available where it needs to be. E.g. why is operational status for my ISP buried five pages deep behind an obscure link in a footer.
But chatbots can make decisions too.
If they can, then it's going to be a nightmare for companies when people manipulate the bot to give them what they want, or a refund, etc.
It's practically putting these decisions in the hands of the customer, and if that's what you want to do, then why not put those functions into the customer facing UI to begin with?
I thought I remembered seeing this somewhere already where an AI chatbot was being tricked into producing working 50% off discount codes?
P.S. It was 80% and I read it 4 months or so ago: https://archive.is/20260311192059/https://medium.com/techx-o...
Much cheaper than having dedicated support tho.
the whole value prop of these companies is they can build you agents that perform actions (e.g process refunds, live troubleshooting, claims etc) without being manipulated
Why would you need an agent for that? They need to look at a database or some rules to make their decisions anyway, so why not make a normal system to let the customer self-serve?
An agent is self serve. Even better because they can disambiguate intent at the top level
The average person gets frustrated with finding instructions and forms, they just wanna say “give me a refund” and an agent can execute it autonomously
> why not put those functions into the customer facing UI
The whole reason of having support is because edge cases are never ending.
I am convinced, that anything touched by Salesforce is going to become an obnoxious product and at some point either become a plague walled garden one has to deal with at corporate jobs, or it dies an agonizing death, from its own user hostility.
Salesforce is basically like Atlassian. Don't expect any good things to roll out of that one.
Intercom was obnoxious already. Reminds me of when I walk into a store and someone immediately accosts me with "may i help you" and then I walk right out.
You want to sell stuff? Don't mind my existence, let me look first at everything you have on display, and I'll initiate the conversations when I feel ready.
Apparently that greeting is to discourage shoplifting not to put you at ease.
Retail theft is becoming a huge problem, police generally won't respond to it so many of those who do it know they can act with impunity and retailers are left with few options to deter theft.
It's sort like how everything being priced at 6.99 is actually to prevent employee theft (customer expects change from their $5 + $2). It's not a psychological trick to make things seem cheaper which is to actually just display 7, a single digit number.
https://www.bbc.com/news/articles/c5yp3543z7yo
> It's not a psychological trick to make things seem cheaper
Interesting! That makes sense, though Claude seems to think otherwise:
https://claude.ai/share/470906ed-6987-46b3-9e81-facfd44fc863
I feel like cash transactions are pretty rare these days at least in the US, so it would make sense for the psychological trick to persist (perhaps even more so because you don't actually have to hand over $7, you hand over a piece of plastic with your brain thinking it is $6ish)
Lots of negativity in this thread, but I think its genuinely impressive that Fin beats GPT-5.4 and Sonnet 4.6 at customer service tasks: https://venturebeat.com/technology/intercoms-new-post-traine...
If that's the exit valuation of the most popular AI support tool (even Anthropic use them) then this doesn't support the trillion valuations of companies like Nvidia, SpaceX (AI), etc
According to public info I see, their last raise was at a valuation of $2B to $2.5B. If that's correct, that's a decent exit, right?
…why? What would the appropriate multiple between them be?
I can't answer that because it's irrational anyway. I don't know, how many millions of people work in support? Because the assumption that all those jobs will disappear is what is holding up the public valuations
> can't answer that because it's irrational anyway
Not how a valuation argument works! If you’re claiming this shows those valuations are irrational, you should be able to point to why. Otherwise, it’s just a “my vibes are off” comment.
I am not going to do a DCF on this because the assumptions are all invented anyway. But back of the envelope:
>> What's the TAM of AI replacing millions of knowledge workers in support? Let's conservatively assume a few hundred billion.
>> How much market share does Fin capture? Let's conservatively assume 5%.
>> What's the valuation on a reasonable multiple?
5% of a few hundred billion is ~$15B of revenue. Let's assume deflated 6x revenue valiation, not the 15x these things were fetching two years ago, and you get a ~$90B company valuation (that it should grow into soon at least).
And it sold for $3.5b
So the price is telling you the real revenue is nearer 300m than $15b, which puts the actual AI-support software market in the low single-digit billions. Not hundreds of billions
And if the TAM is real but just being captured by the incumbents: Salesforce's own Agentforce, the supposed winner, is at $1.2B ARR. The "someone else is eating it" defense still has to point at the someone, and no income statement anywhere shows hundreds of billions of revenue
For Nvidia to be at $5T and the hundreds of billions a year of capex behind it only pencil out if that compute throws off a huge revenue stream downstream. Support is imo the cleanest test there is to demonstrate future value of AI in the real world (literally the first thing everyone said when ChatGPT 3.5 came out was that support will be eaten first). It's the most mature, most deployed, most automatable, and the exit price of its best player is...pretty small
> Let's assume deflated 6x revenue valiation
Salesforce trades at a 4x revenue multiple, FYI.
Also, taking a TAM and multiplying it by 5% to back into a revenue figure is “if we only get 1% of the market” math.
I’m not saying you’re directionally wrong. But the claims you’re making can be rigorously made. And I’d argue they’re interesting when they are.
At the end of the day, the valuation is decided by how much a buyer wants to pay, and how much a seller wants to sell.
That $3 bln number encodes all of that in a price. Not much more to rationalize. It's quite beautiful.
That is not what is supporting these valuations.
A P/E > 0
I believe I saw reports a few years ago saying Intercom was profitable?
sierra is the most popular private support tool (16B)
> The AI Agent is powered by the company’s proprietary AI model, Apex
When will I be able to talk to Salesforce Apex from Salesforce Apex?
Knowing salesforce a little, only if you upgrade to a different license tier.
You’ll need wrap it in flow first
I thought Salesforce had a few activist investors on the board that wanted the company to stop all acquisitions and focus on the core product. Whatever happened to them? Did they bounce after they hit a payday?
Company is down 30% YTD and Salesforce seems to be back on its acquiring spree.
When did intercom change its name to Fin?
30 days ago. No joke.
https://www.intercom.com/blog/today-intercom-becomes-fin/
That prose is painful to read. It has such a cloying quality to it, even when compared against other name change announcements like the rebrand of Google to Alphabet
https://blog.google/alphabet/google-alphabet/
Wow. So a deal like this is not overnight, its 6-12 months in the making at least. So the name change comes as they knew the deal was closing and doubled down on this new branding for customer agents. Maybe because Saleforce wants Salesforce Fin Customer Agent as a product. Who knows. Don't want to read the post...
This was my guess on the impetus for the name change as well.
Yeah - they want Salesforce Fin, not Salesforce Intercom.
Fin! As in finito!
"Fin is clearly our future" by choosing a word which means "end" in Latin and Spanish ... LOL
Almost like a pre-announcement about the acquisition?
As someone with extreme phone anxiety to begin with, long waits on phones are deal breakers for me. If you are a company that answers your phone quickly by a real person you've earned yourself a lifelong customer.
Yikes, I dont think there is anything special about Fin. They claim they have their own model but it appears they are just using OpenAi and Claude as well which is hidden in their docs.
I always hated seeing Intercom bugs when a simple help menu would work better, but it proves that being annoying is a winning strategy. As my various SaaS packages have switched to AI bugs, I have never had one successfully help me until I say, "Can I please speak to a human?"
Is this the same Fin that you get when contacting Anthropic support?
> The AI Agent is powered by the company’s proprietary AI model, Apex, that is purpose-built for customer support
Wow. If so, Anthropic is not using their own AI for their own support! (I had assumed the AI support agent was an Anthropic one, because, well, Anthropic.) And given how poor my experience with Anthropic support is, I have a very, very low view of Fin.
I’ll be interested to see how the merger goes. Although the products overlap, Fin <-> Agentforce and Intercom <-> Service Cloud, the Intercom/Fin stuff is mostly built for startups while the Salesforce stuff is much more enterprise. The play may be that Salesforce has the… salesforce… to bring Intercom to enterprises way faster.
I'm sad to see Intercom/Fin's CEO, Eoghan McCabe, did not face repercussions after credible allegations about sexually harassing his employees came out in The Information in 2019 [1]. Some quotes:
> When the woman said she, too, wanted to go home, Mr. McCabe told her that he wanted her to remain, the people said. After the other two employees had left, Mr. McCabe made a lewd remark to the saleswoman and said that he would like to sleep with her, the people said. She declined and eventually left his home, they said. > Later, the saleswoman said Mr. McCabe’s actions had frightened her, according to one of the people who spoke to her about the incident, who declined to be named.
> During the meeting, Schuur said she told Mr. McCabe that his behavior was inappropriate. The Intercom CEO cried during the meeting and said that he hadn’t realized there was a power imbalance between himself and the young woman, Schuur said. > “I felt like I was just a guy at a party and she was just a girl,” Mr. McCabe said, according to Schuur. > Two former Intercom employees present for the harassment training seminar said Mr. McCabe didn’t attend.
> At an Intercom party around 2014, Schuur said she witnessed Mr. McCabe slap the saleswoman’s buttocks. At a different event around that time, another former employee said she saw the Intercom CEO place his hand on the same woman’s thigh. Four former Intercom employees, including Schuur, also said the woman told them of the advances, which she said were unwanted.
> A key figure in the company’s culture was Mr. McCabe, described by employees as both brilliant and temperamental, with a tendency to cross boundaries with junior female employees.
> Mr. McCabe, through an Intercom spokesperson, declined to be interviewed for this story, but said in a statement: "In the early years of the company I demonstrated some poor judgment. I apologized at the time [...]"
> But early employees said he also showed flashes of temper and vindictiveness. He was known for occasionally writing scathing messages to employees in group messages, according to a former employee who saw them. On at least one occasion he told another person leaving Intercom that he intended to tarnish their reputation in the tech industry, that person said.
I previously tweeted about this and McCabe threatened to sue me.
[1] https://www.theinformation.com/articles/harassment-allegatio...
cached content: https://docs.google.com/document/d/1zsAytvQRuEwc0ftY3OtlXwvj...
Huge congrats to the intercom team on that! Intercom was pretty great for me during my Percy support years. Their current AI direction is rough, though. When I went to use them for my product I was (am?) building, I found intercom unrecognizable with the entire ai pivot. Honestly made me sad but that’s how it is these days
They’ve been responsible for a number of crashes in our iOS app since they switched to agentic coding. It’s become our most unreliable 3rd party library.
It's a win, but I'm not sure its the win they were hoping for. A regular congrats is probably sufficient.
This is disappointing. I've worked closely with Intercom for the past few years. I've run Fin implementations at multiple companies. I've found their product team incredibly strong & the product to be customer friendly. Salesforce... not so much.
imo - Fin's AI chat is the best on the backend of empowering teams to self-serve & integrate with Helpdesk. They don't require consultants.
I really hope they don't lose all of that in this acquisition.
Huge outcome for Intercom. I don't see the value at all, but grats to the Intercom team for getting their bag.
Intercom... well, Fin's announcement: https://www.intercom.com/blog/salesforce-signs-definitive-ag...
Congratulations to building a cool company that inspired thousands, then converting it to LLM reply bot, and then selling it to enterprise graveyard to be dehumanized and monetized even further.
I don't want to buy a product that doesn't want a human to talk with me before or after sales.
Good for Intercom. I have to assume that Salesforce will immediately rename this since any brand recognition was already eliminated a month ago.
Salesforce Einstein™ Agent Cloud (not to be confused with Agentforce, which will have basically the same goals and the same target market, until they kill off Salesforce Einstein™ Agent Cloud eventually).
This sucks. I love Intercom. I hate Salesforce. I do not have confidence in Salesforce to be good stewards. Will never forgive how badly they've botched Heroku.
Woof, yes! Though I'm glad Heroku remains, and I ought to also be glad that it has been stable all these years, …it's clearly neglected.
The very first interaction I have with Heroku is a two-factor sign-in, and …it's this horrible page hosted on salesforce.com, which doesn't have retina graphics (i.e. is blurry on screens Apple have used for 15+ years), doesn't work properly with automatic one-time-code generators (because the login form is heroku.com, but the two-factor is salesforce.com), and …gah! What a mess. Thankfully you fall through into the pretty, well thought-out Heroku dashboard of yesteryear.
That new salesforce powered heroku status dashboard is a labyrinth of sadness.
https://status.salesforce.com/products/Heroku
The "AI" "agent" "helpdesk" they pivoted into is such a grift. AI agents still does not solve the main issues that makes a person contact support in the first place. How do I know? I was a founder in the space.
but good for them that they got salesforce to buy it.
Since we're sharing anecdotes, I work in this space and my team automated a large part of customer support for a large scale up using an AI agent. The kicker is that customers rate their interactions with the AI agent higher than their interactions with customer support staff.
AI is definitely capable of taking on customer support work at the moment, and to a high standard as well. Sure, it's not perfect. But it's not a grift either.
Solving 80% of menial support tickets automatically through agents trained on your help center, freeing up your actual support agents to focus better on meaningful tasks seems a tiny bit more than a grift.
I’m not sure where you’re getting this from but their customers find Fin to be a hugely impactful tool
I guess I'll bite, what is this main issue that seemingly everyone but you are able to see?
The reason people want to get a human on the other end of the line is usually because they want some sort of remediation,like a refund or need to escalate something to someone who can take an action. Right now, AI agents just barf the FAQ back at you. Which is great for your tier 0 calls, but without an easy way to bypass they are just underlining the problem inherent in gethuman.com needing to exist.
Now if AI agents are free to issue refunds or discounts by their own? Great, let's do that and suddenly most people are on board. But get ready for rampant abuse.
Best solution would be an AI cyborg system where it readies a recommendation and a human swings by and approves or denies it without wasting time talking to people. But users would hate that (anti-social), it would still be ripe for abuse. But it is likely the longer term solution, as people will quickly realize they can use web chat or Google AI to get the exact responses as your FAQ bot which means you have removed actual customer service and this is a non-product.
Talk to random non tech people about customer service bots for five minutes and you’ll find most people see it.
I'm guessing parent is not talking about something super obvious to the rest of us, that'd be disappointing.
If I had to guess, there's two things they feel AI doesn't address:
1. People wouldn't need to contact support if you just made quality goods and services. Outside of rare exceptions and inquiries, of course.
2. AI has not advanced enough to trust it outright, nor does it have a physical body. So it can't really do anything you wouldn't already just be able to put in the UI for the customer, without needing it's actions reviewed and confirmed by an accountable human. See: accidental truck giveaways.
So investing into AI support over making your business better is seen as misallocation. And using AI support instead of just improving the service is seen as inconvenient. And using AI support when it needs humans to do the support anyway is seen as inefficient.
> People wouldn't need to contact support if you just made quality goods and services
It'd be weird to start a startup around that, sounds like something for a consulting business instead, parent specifically mentioned they've founded customer support startup, must be something actually related to customer service, I'm assuming.
AI support agents, chatbots, etc. on web sites these days are the equivalent to search engines during the dot com boom. Everyone felt that having a search engine on their dot com was the killer feature that was going to win. There were a lot of companies building search engine technologies and selling them too. Now its ubiquitous that most ecommerce sites have a search bar but nobody cares about the underlying technology much, its consolidated and commoditized.
Good on Intercom for getting acquired.
If we're seeing larger consolidation/acquisitions happen, does that mean the hype train has hit a key station?
Seems like every time a tech company rebrands, it’s a cry for help.
this is very bullish for decagon and sierra
Woah, feels like Intercom just threw in the towel. I always thought they if anyone would be the ones to come out on top for customer support. Their brand was #1 in perceived cutting edge in my eyes.
Wild, the AI support and bots suck so bad, I've literally lost sales due to them. No matter how much support docs and history you feed them with for context, they just don't cut it. People would rather wait for a real person than go in loops with a wrong/bad support answer...
There’s essentially no such thing as good case deflection. All of it exists at the expense of customer experience.
But businesses will always chase that dream of reduced customer contact, so Salesforce will keep selling it to them.
These days it would be better in reverse i.e. expose the docs as docs and human points their Claude at that. In starting to do that!
So I haven’t looked into it a ton, but doesn’t it seem like a great case to have an AI answer the call immediately, get the users account pulled up, and document the issue with some refining feedback?
An AI secretary seems perfectly acceptable for both sides. The expectation is that a real human comes in soon after but this seems like a way to free up the most tedious parts of the process for both sides.
Congratulations to the Intercom founders. This is a great result for them and the early backers of the company, for a product that I feel was executed brilliantly. We can criticize the usual enterprise-sales-y trajectory and enshittification that ensued after the brilliant opening moves, but at the end of the day, I think Fin reached a solid ending (no pun intended).
Intercom is a great example of a feature that in theory could have expanded into a more full product but stayed laser focused on that smaller vision instead. Nothing necessarily wrong with that as they did not allow themselves to become fully enshittified. Still it would be cool if they had really expanded it to offer a competition to Zendesk and Salesforce Service Cloud while preserving their minimalist and design forward approach. I understand that AI became the thing that invested in instead which was inevitable
What the hell? That is so cheap? I would value this at least as much as cursor? What gives?
Cursor at 60B is WAY overvalued. At this point you have to assume that anything that gets touched by Elon is hyper driven financial engineering.
> Cursor at 60B is WAY overvalued.
You can't possibly mean a glorified editor-shell isn't as valuable as say Nike, Deutsche Bank, Target, Ford or Nintendo?
> You can't possibly mean a glorified editor-shell isn't as valuable as say Nike
That's a glorified feet-shell. So like-for-like?
Say you grabbed a random selection of just 100 million people in the world, then ask them two questions, "Have you heard about Nike?" and "Have you heard about Cursor?", what would you guess the ratio would be like?
Even when you use "Nike the Company" vs "Cursor as a general search term" to compare search history in Google Trends, it's 71/5, so I'm guessing most people would say they've heard about Nike, while probably most never heard about any software program called "Cursor".
If you asked the same selection of people about Saudi Aramco or SK Hynix none of them would know what those are either though, right?
I do think 60B for Cursor is way overvalued. Just not sure how to quantify
Fair point, those are valuable for other reasons. My point was more to illustrate "Nike is valuable because of the brand", without using those exact words :)
Sure but I would expect an exit of like 20b or just list and go public…
European versus US startups maybe?
While I think that this is a bad move for Intercom, it's actually brilliant for Dublin and Ireland that they have finally exited.
Intercom allegedly has ~400M of annual revenue, making the multiplier less than 10X. Huge customer base and established brand. Looks like a steal indeed!
The world we live in… 10x revenue is a “steal”.
Lots of embedded assumptions about growth and margins to convert that revenue multiple to discounted cash flows.
A steal compared to other “AI-adjacent” valuations.
How does salesforce have that much to spend? Like they are big, but this is a lot of money.
Spoken like someone that’s never seen the Salesforce bill of even a small company. Absurdly high
They had almost $9 billion in cash on hand at the end of April 2026. And net income of $2 billion. So they afforded it pretty easily.
https://investor.salesforce.com/financials/quarterly-results...
By laying off people.
Salesforce made 37.9 billion in 2025 of revenue. 6.2 billion in net income (or profits)
Salesforce had ~14 billion in free cash flow last year they have tons of money.
They're not building out AI infrastructure
they are not just big, they are huge
Its the US Dollar that is worth so much less...
Massive congrats to Eoghan McCabe, what an amazing story arc. We love when a CEO comes back and gets the business back in fighting shape and then delivers an incredible win. CONGRATS CONGRATS CONGRATS. I freakin' love it.
No, screw this guy for his horrible behavior and openly awful politics. Congrats to the rest of the team at Fin who actually delivered this milestone.
"Please don't use Hacker News for political or ideological battle. It tramples curiosity."
https://news.ycombinator.com/newsguidelines.html
For those who are curious (since we don't want to trample curiosity), Eoghan McCabe is a known abuser -- and has even apologized for it in the past -- and he attended David Sack's Trump fundraising event for the Trump 2024 campaign, including getting a photo with the man. He has reversed his stance on immigration (favoring the current admin's policies and ICE expansion), although it was likely the case that he attended as a pro-crypto pusher more than anything else. Whether or not he has significantly invested in crypto is speculation.
End of an era. I met Eoghan (and Des) at JSConf back around 2011. Twilio (represented at JSConf by Carter Rebasa) and Stripe (Alex Sexton) were also getting started around that time.
It was a super hopeful time: JQuery meant programming for the web was less painful, sockets and HTML5 SSE meant realtime was just starting.
Intercom and Olark were one of the first two "install this <script> tag" based customer support apps. They made websites a way to talk to companies rather than just read see and buy.
I can't believe the exit took so long, especially in a field so crowded. But it looks like it was well worth it.
Everything around you was made by someone just like you.