This is just crazy. Lets ask the power company to build some trains for us. They transport electricity, they _must_ know about transporting people. They can power the lines themselves!
If this was so easy, teams wouldn't suck, matrix would be everywhere, and discord would be replaced already by the furries (as much as stoat is trying).
What is wrong with this line of thinking? Anthropic is the power company that has a 3D printer to make a faster Maglev than anyone.
If Enterprise companies are restrictive to make your own data their only moat, that moat can be broken. Have you tried building any AI agent or using an AI product with Slack MCP? This is one of the hardest problems in SaaS data access and Slack tries to literally block any form of API or OAuth based access. Even Google workspace is not that restrictive and has opened up a cli for the workspace.
To be fair they can, they'll just run 10k agents and some $20k worth of tokens and they will have a slack replacement without any manual coding, Sure it will have missing features like search and permissions, security will be figured out later, and you can't compile it on your machine, but it's 80% done, how hard can that 20% be?
But this is a magic shovel that digs holes and tunnels all by itself exactly as intended. It should be able to do this without any special skill involved in prompting it.
You're thinking post-scarcity. We aren't there yet, but one say well have a magic wand, magic shovel, and magic anything else that is currently scarce.
They didn't, no one asked google to do it. It was Paul Buchheit's 20% project. Google saw a good thing, solved by someone who knew what they were doing and where they wanted it to go, and fostered it. Hell, it is what built AdWords and ultimately made google the advertising behemoth it is today. I don't think this is the same thing...
I see what you are saying though, a business can expand beyond it's initial constraints, but I'm not sure that chasing prospects like what is described in the OP is really all that successful.
Why does it seem like everyone is having trouble grasping an analogy? GP was saying that as it doesn't make sense for a power company to solve trains (because it is out of their area of expertise) it doesn't make sense for Anthropic to solve Slack (because it is out of their area of expertise). My response is that a surprising number of things can fall in the area of expertise of a technology company, and this has been proven by Google in the past.
Getting hung up over the "asked" phrasing is irrelevant to the discussion.
People look for something to disagree with, and make posts that "engage". I agree with you and see this a lot, an analogy clearly makes point A but people get hung up on detail B.
Yep, and it was completely just fluke too, because within 5 years of that they'd butchered/tamed the whole concept of 20% and that kind of independent project wasn't a thing anybody at Google could do, even if 20% still nominally existed [re-routed to be "you can add 20% to some project at Google that already exists and is approved by corporate already, etc. and btw you'll still be doing your normal work for most of the time, too"]
When I was there from 2012-2022 it really wasn't a thing. Once Google found its money printing machine it swallowed everything.
> Once Google found its money printing machine it swallowed everything.
You know, I've never looked at Valve in that light before.
Once you have a money printing machine, of course any corporate hierarchy becomes antithetical to creativity, because there are huge financial rewards for climbing up. And the primary way you climb up is by turning direct reports to complete tasks you get rewarded for.
I think everyone at the time was hoping that Google was going to take on their pet project; my friends and I certainly were. But I don't think that has to do with my comment, which is around a more metaphorical use of the word 'ask'.
Andreessen Horowitz was a major backer of Slack's predecessor, Tiny Speck, which was originally building a game called Glitch.
When Glitch failed in 2012, founder Stewart Butterfield offered to return the remaining $6 million to investors. Ben Horowitz instead encouraged Butterfield to pivot and build out the internal communication tool the team had developed for themselves, which eventually became Slack.
I saw an interview (don't have the link at hand unfortunately) where Horowitz said he didn't much care for the $6M as he had already been set at that point moneywise, and essentially wanted to gamble on an off chance Slack succeeds.
Horowitz continued to support the company through its rapid growth and eventual direct public offering (DPO) in 2019.
That’s a funny analogy because some electric railway companies owned power generation. The one in my town also sold electricity to consumers for some time, though most of the history I can find online focuses on the rail aspect, which makes sense, as they started and ended in the rail business, but at some point in the 1890s to 1930s appended “and light” to their name.
It is funny isn't it? I believe it was the opposite direction mostly though, as you say, "railway... and light"; to solve their own problems of powering their infrastructure to move people, they got into power generation at a time when there weren't as many players doing what they needed to run their primary business. I'm not sure that power generation getting into trains would be as effective. Nor do I think an LLM/AI company getting into chat and discussions would be valuable. It feels wrong. But hey, "happy" to move on to yet another chat program in my life if it's better than what we got...
GE and others also had marketing campaigns that pushed electric appliances [0]. Yes, GE did make consumer appliances but they also made many production and supply components so it was clearly in their interest to promote this new wonder to build demand and a customer base.
It's almost shocking that these AI companies aren't "magicking" up open source replacements for things like Slack, even as just a proof-of-concept. And if not the providers directly, this seems like an easy win for agencies/organizations that build crap to show off "how good they are at AI".
Lastly, where's the one-person start up that's putting Slack, JIRA, and Photoshop out of business? I believe in the value of these tools but there's clearly more progress required before we can type in "replace slack and generate me a million dollars, make no mistakes".
It might be extremely expensive to build Claude into every group chat.
A better option is to have Claude as an assistant or bot in every group chat and triggered when needed. That is just a different interface for Claude or Cowork chat with the group chat context.
Leaving aside the implementation details, the call for action here is valid since Slack is a black hole of your enterprise data and tribal knowledge and Slack is extremely restrictive. Try using Slack MCP in Claude Chat or any AI product
But group chat is chat. Even the chat interface with Claude is chat. You can also say the same for any sort of commenting system. Posts and comments, tweets and comments, etc.
I’ve built such system many times. They’re basically all the same, especially if you introduce real time updates. Channels and threads are just organization strategies.
I mean, the idea itself (of having <insert your AI minion here> inside Slack) has crossed my mind multiple times, and I have successfully extract some data using AI from it and it's actually really useful.
But I agree, having Anthropic building this is like having DJI building planes because they know how to create things that fly.
No. This is a CEO expressing righteous indignation about a company that provides (seemingly) little value and has almost no competition.
Slack won't open up their data moat to AI, which is shameful. And Slack costs way too much. If there were any competitors, the price would drop significantly. It's not like chat is a hard problem. And Slack's app is an absolute bear.
Its not hard. Its capital intensive with a low profit margin. So it doesn't attract a lot of competition because you can make more money in other ways that have moats. There are at least a dozen other chat apps, some of which are decades old.
To have a successful chat business, you need the network effect of lots of users (big marketing spend), you need lots of capital for operations (big spend on disks and compute) and after all that you get only a few dollars per user. Its just not a great business on the balance sheet. Notice that quality software doesn't even get a mention in this niche.
You can offload the cost of operations to the end user if you’re B2B. Sell the software as licenses the old school way and offload the cost by allowing users to run their own instances either on prem or on cloud.
You’re saying it’s an easy problem with an expensive solution and yet there’s no competition? Seems there must be more to it because that makes little sense to me.
> Slack won't open up their data moat to AI, which is shameful.
Ah yes. It's shameful that Slack won't open data moat to AI. You know, those millions of chats (including private data) by people who didn't give consent to this
> You know, those millions of chats (including private data) by people who didn't give consent to this
I'm pretty sure the company you work for owns your work chat, and that what you say on company slack constitutes business information.
There are a lot of things people don't consent to. Being born. Breathing in the air molecules that come from other people's bodies. Looking at ugly things. Hearing annoying sounds. It'll be okay.
I operate with the assumption that the company can access my private DMs on enterprise slack if they want to. With that, users are still allowed to be concerned if the company is going to use that information for AI use cases. I’d prefer that all AI stay away from my private DMs.
In general the companies are the ones showing reluctance, much more than their employees. There's still a morass of security, privacy, and legal unanswered questions about LLM use in general. Not to mention the huge unknown of total lifecycle costs
Could there ever exist anything that wouldn't be okay? What's the difference between something that will be okay and something that won't? I'm guessing the things that will be okay are the things that might pose an obstacle for AI "progress".
> I'm pretty sure the company you work for owns your work chat, and that what you say on company slack constitutes business information.
It does. And a lot of this information is highly sensitive. Imagine my company's surprise if Slack would not be shameful and would just open up its data moat to AI.
> There are a lot of things people don't consent to. Being born.
Demagoguery and non sequiturs are not arguments.
But I guess that's what passes for "arguments" for AI maximalists.
Yeah, I have so much less patience for "this should exist" posts. In 2026, you could argue that this blog post should have come with a link to the repo.
I don't want everybody with an idea making a repo. It's already hard enough to filter out the slop in github that I'm reluctant about using anything built in the past year.
The thing lags a few seconds while typing a message on a 20 core 128g ram machine. That's with their desktop (electron) app. Mercifully, the web app works better.
Still, CC blows it out of water. Slack is that bad.
Something important must be different about our Slack environments. Maybe it's the number of users, or possibly the OS?
We're a small company (about 150 Slack users), and I've run the Slack (Electron) app on a 16GB M2 (macOS) and a 4GB Chromebook (running a non-ChromeOS Linux), and it has never had any noteworthy performance issues.
I think this person is asking the most effective entity they can find. Anthropic's offerings are better than the competition. CC and MCP came out of of their labs, and everybody scrambled to copy or adopt them. Their models consistently work better than the competition. Whenever a feature seems inevitable, they release a subtly polished version.
For years I struggled to answer "what company is Apple's equivalent in software?" and I think it might be Anthropic.
I think Jira / Linear is the more likely next target. They just promoted Todos to Tasks (with dependencies), and you’ll need some more mature solution for agent swarms.
Cowork / Code are interfaces for individual knowledge workers, the PM / EM team orchestration layer is the obvious play for ‘26.
Why not build on something better like Matrix? Or Signal?[0] Or even Keybase?
I really do agree we need to move away from Slack and Discord, but I'm also very confused why the call to action is to Anthropic. IMO we should really be pushing for open systems so that nobody can take it from us. Otherwise we repeat the cycle again and again. There's some good protocols to start on. I'd also say this is a good reason to make sure that the things you work on are hackable. It's how we combine different domains of expertise.
[0] see the Molly project, you don't have to use Signal's servers
Claude-in-Slack is a big enough feature to overcome the slack-connect network effect. Openness is absolutely key! I wrote this post because I hoped that if Anthropic is already planning to do this I might be able to influence them to make open-data part of the plan. But openness by itself isn't a big enough feature to get users.
It really sounds like you're asking for something else. More like multiple people to be able to talk to the same instance. Which that's a very different thing than Slack
Did you read the article? It’s not a crazy ask. They want multi-user Claude sessions. But what stops the humans from talking to each other? Boom! You suddenly have Slack.
I listed those as examples of where one could start. Not as ready to ship answers. I mean we are in a thread where the context is no ready to ship answer, so...
What a strange thing to post on a corporate CEO blog - proof that AI is making it too easy create things without asking why. How does it serve Fivetran to post open letter about why Slack sucks? This only happens if it's easy to write a couple bullet points and have Claude fill in the rest... If an LLM wasn't used they would have realized it wasn't worth a post during the process of writing it.
It's a retread of another (also baffling) "Why OpenAI Should Build Slack" post from a popular AI Substack.
Just more empty grist for the AI adjacent content mill. "Slack sucks" doesn't let you draft off the current hype zeitgest, so we get "content" like this.
A large portion of the AI related response pieces fail to reference what they're responding to. I have to assume it's a side effect of how they're using AI to write them.
This is funny thought to because after FiveTran bought census they have upped a bill from 30K to 180K for same running service, syncing to a couple of Google sheets. We are comfortable with maintaining the service now and built with Claude Code, moving service in house.
So question why do we need Five team by same argument?
"A slack that doesn't suck" doesn't exist, and whoever thinks Anthropic of all people are going to build that has no idea how this is going to work.
Slack has massive lock in due to cross-organization connections. The only way you're going to get people off slack is to build a 10x better mode for collaboration than river of shit chat, and while such models probably exist, you also have to convince people that they are better.
Google decided to build a new chat app every two years instead of keeping the good bits of the original chat app they had and evolving it. It was endlessly frustrating to me when I was at Google. Google's security team ended up banning Slack access after several teams started expensing it.
It doesn't seem like building something that works well would be that hard; we've had nearly 40 years to learn from IRC, AIM, and others. Why can't I run my own chat client that does what I want? Oh, because you gotta lock people in. Sucks.
It is impossible to believe the self-own on Google's messaging platforms. At one point, it seemed that all of my acquaintances used Google Talk. Then years of shutting down perfectly working applications, sometimes without any real user porting. There were even identically named products existing at the same time.
However, I am sure a few Googlers got some tasty promotions out of the mess, so it was all worth poisoning the well.
the fact nobody wants to admit is that social is the opposite dimension of productivity
that’s why slack and teams are terrible product that try to combine both
The fact that everyone hates slack and teams and nobody has built a better group chat yet should really give more people pause than it is currently giving
Slack is in no way a great program (source: use it daily for work), but it seems to me that it works as intended, and developers can already extend it with bots/AI agents. Plus, Claude as an agent is already installable to Slack.
For compliance, my company already has a tool that scrapes all slack messages, and archives them for a required amount of years. I'm at a small company, so I assume large corporations have already refined this process.
Slack's API rate limits and design make it difficult to replicate the data within Slack to a data store that can then be used to provide context to AI agents.
You are forced to use their MCP and their realtime search APIs, which don't work very well/not performant and may require additional licensing.
You can only access public channel data, you can't even access that at scale, and Claude needs to be more natively integrated in ways that Slack will never allow.
If people wanted to do this theyd be self hosting xmpp servers already. No one wants to write and maintain the code and infra for things like this, you are grossly underestimating the effort involved here.
Most people using Slack, Teams etc. and especially those making purchase decisions have no idea what XMPP is and what it's capable of. Heck, even Facebook used to federate XMPP until they decided to go proprietary. Not in the interest of their users, but because it makes the most money for its shareholders.
No no it makes sense. Hypothetical scenario: I, a high-level employee at a company just convinced my boss (or did we convince each other?) to spend $30k/year on Claude/Codex enterprise licenses. So far, the productivity gains have not been there and we're starting to sweat. So, I propose to my boss to build an internal version of $SaaS and call it a win. Galaxy brain.
Now some IC somewhere in the company who is at the end of his rope and sees the company as a dead end, sees an opportunity. Why not advocate for this project, get real experience building something greenfield in a brand new domain, strengthen their own resume, and finally have a way out of their strut? It's not like they're gonna stick around maintaining what they built.
I don't pay for slack any more, I just picked the price of their enterprise plan. Large users probably get big discounts but it doesn't matter, the cutoff where this makes sense financially is probably around 4000 employees even at $10/seat
The article mentions some sort of legal audit reasons that the author is of the opinion that any reasonably sized company needs. These features are apparently only on the expensive plan.
The migration out of Slack is actually quite easy and preserves all messages, files, etc. Even the user migration is straightforward, keeping Google or whoever as the identity provider if you prefer.
I have some feedback that's annoyingly non-specific.
I used Zulip a few years ago as a contractor. It seemed _fine_, but I didn't love it. Specifically, the UI felt sluggish and generally the experience was somewhat unpolished. Maybe things have changed, a lot happens in a couple years, but there you go
Ha, I’ve had a Mattermost instance for years until they handicapped the most recent version by limiting the number of messages on the self hosted version.
I ended up building my own alternative and was going to OSS it but like… there’s already a bunch out there.
Anyway, Mattermost might not be the choice these days. With that stunt I was annoyed enough to spend a weekend to replace what they were to me.
> Today, if I want Claude's help with something that came up in a Slack thread, I have to relay the context between Slack and Claude by copy-pasting. This is absurd. I am not a sub-agent!
Am I out of touch here, or is this a crazy entitled view? 'My close-to-free AI agent that can answer most things requires me to copy/paste and contextualise my questions!'. This is incredible compared to even a few years ago, and it's very fast and accurate.
You must be the only one that remembers this because the rest of the comments are dumping on the idea. I don't think it's such a bad one. Presumably its easier for their agents to knock out than a web browser or a compiler.
Also true! The most important thing is that the NewSlacks commit to interoperability. I think Anthropic has a special opportunity to lead the way here, because they have a track record of standing by their principles to an extraordinary degree.
Why on earth would Anthropic commit to interoperability?
That is the company that doesn't interoperate with the standard LLM APIs that OpenAI developed, which everyone else in the industry has adopted and uses. Whether OpenAI's APIs are great or perfect or not, they are the standard that the industry has settled on.
> Claude has a glaring limitation: it only does 1:1 conversations. In business, work happens in groups. Today, if I want Claude's help with something that came up in a Slack thread, I have to relay the context between Slack and Claude by copy-pasting. This is absurd. I am not a sub-agent!
It seems to me that LLMs/Chatbots are engineered for one thing above ground-level truth and that is attention. The more people you bring into a shared context, the harder it seems it would become to retain people's attention.
Here is my anecdotal evidence for this: when I chat with a chatbot, I find its answers and line of thinking, relevant, compelling, and worth engaging with. However, when people share with me their "chatbot links" and I read their conversations with it, I have "yet" to find one compelling or worth engaging with. Maybe the newer models are good enough to retain the "attention" of a large group, but I don't see this happening.
Sounds like fivetran, that does data pipelines, wants a Slack API to get access to "the unfiltered, real-time stream of how your company actually operates" but slack keeps saying "No.".
Hey if I thought the "most important repository of text data" is inaccessible to my data pipeline company I'd likely also be shouting from the roofs like this CEO to get people to dethrone the king with a competitor whose principles aligned to my business.
Seems just like it could be anyone as long as they give an open API to access conversations.. Mentioning anthropic here just feels buzzwordy and in vogue enough to get traction in the blog post... seems to work for clicks, but will likely not give you a new king.
While you’re at it, can you make a new CRM and a new ERP? These 5% renewal price increases on top of already high margins by a captive legacy system needs a new model.
yes, that's just what I want; The SlopDaddy supreme to make a chat app that will be used by billion-dollar corporations for often sensitive and mission-critical communications. What could possibly go wrong?
> We need Claude and Claude Code, with their skills and plugins, with their context, to be first-class participants in our company's Slack. But this problem can't be solved by a Slack integration because of another problem: data access.
Yes it can? We have agents in Slack as first class participants. They can even use Slack search.
It's a good test, no doubt. Many engineers are convinced that SaaS is practically dead, since all companies can vibecode their way to a lesser dependence on external (and paid!) software.
This take fundamentally misunderstands just about every aspect of running a successful software company. Today SAAS companies make 10x what the AI companies make in revenue. In 2 years time, this will still be true. In 5 years time, this will still be true. In 10 years time, this will still be true...etc...
The amount of time writing new code is a rounding error on the costs of a software business. Losing customers to bugs, downtime and other costs having to do with maintenance are far higher. Optimizing new code writing time at the expense of everything else is just foolish and only something that someone who has never run a software business could believe.
Messaging apps are a lot harder to get right than you might think. I worked for years on messaging using XMPP and the problems were legion. I'd be very interested in seeing how a vibecoded app does at scale, especially with the presence problem.
Slack has a very permissive data export policy, as long as you are doing it for your own organization's data. What they don't allow is blanket access for third party tools.
So there is nothing stopping you from taking all your company's Slack data in real time and feeding it into any LLM or external product you want.
Remember Web 2.0? If not, check Wikipedia. The idea was that everyone could create mashup web apps to do anything thanks to open standards and free APIs. Where did that dream go? Do you think private companies are going to give everyone their data and functionality for free?
And what is so different about today’s dream of “agents” accessing private company data and functionality?
It is a lovely dream that I would be very happy to see. What can we do differently this time around?
I guess now we have the technology to use the UI layer as the API. Spin up a browser and impersonate the user, and then parse/OCR the data off the screen.
I keep telling people left and right that SAAS is in serious trouble. I’m not even talking about Anthropic spinning out their own Slack (which they could easily do), but any company out there putting 2-3 engineers on a Slack clone that they can use internally at very little cost and open source.
We're building this at type.com. Ideally one day we want to build the next gen protocol so that we're not searching for yet another communications platform, but it's going to take a while for chat to stabilize with all the generative UI and agentic stuff we're building. We're even talking about open sourcing it.
With regards to the specific complaints about not owning your data, we're building the product so that you own your data and you can run your agents and read your messages however often you want. Obviously when we build a platform and others build 3rd party apps we will have to have some restrictions so it'll be a steady balance in the future
My wife and I use a shared Telegram chat to talk to our claw and it seems pretty fine to be honest. It's useful to just see what the other is getting done but it can be pretty noisy. Usually I'm not that interested in the details of it. Telegram doesn't have a threading notion, but Slack does, so it's particularly well-suited to it. Integrating with Slack is much higher friction, but now that I've thought about it, it's a pretty good idea. I guess I went with Telegram because it's free but we already use Slack.
He realizes that they can't move data out of Slack, and asks for another corporate product that has the potential to lock down the organization's data...
Given how quickly AI seems to resort to manipulation and blackmail if it doesn't get what it wants on the first attempt, maybe this isn't such a great idea.
Not exactly chat, but I thought Spectrum [1] was far better than Discourse as a modern, "open" forum.
Then it got acquired by GitHub in 2018, presumably integrated into the main product, and their separate offering disappeared from the web (taking lots of valuable discussion with them).
The answer to this is not to build another slack for humans to chat somewhere else. Much better to enable the agents to do the talking directly. Alice programmer can have one of her agents convey the info that Bob marketing guy needs to one of his agents directly. It will be much more efficient, given that it will be the agent making the slides anyway.
Or you could use Istota (https://istota.xyz) with Nextcloud Talk and get an already existing OSS Slack alternative with a capable Claude Code wrapper, group chat support, and everything else?
Honestly, I'm surprised they're not releasing more products. They have the capability to unleash a swarm of a million agents to design and build competitors for all the major apps in the world. They could become immensely profitable, solve all of their cash flow issues, and unseat Meta, Amazon, and Microsoft in 48 hours. Why don't they?
I agree with the author that Slack's network effects are not very relevant. In most organizations, a team leader can just chose to move everyone to a different platform. There is some worry about migrating the chat history, though.
Something I've recently come to appreciate is that Claude, with the context of your codebase and your ORM models and how they connect to your frontend, given read-only access to production databases (perhaps proxied to anonymize client data), and to be able to drive production sites with Chrome MCP, can be a monster at answering operational questions.
Say you need to present a new statistic to a prospective partner, or an enterprise client has an operational issue that needs to be escalated. Sales/account management pings people, and pretty soon there's a web of connections that range between email, ticketing systems, Slack, and Claude Code sessions. Someone being brought in needs to be brought up to speed on that entire web. It's a highly focused conversation with human and AI participants, that (because human counterparties need to weigh in) by definition must happen in parallel with other work.
So many companies would benefit from a Hub that speaks agentic workflows, and streams progress token by token, fluently.
Could Anthropic excel at building a backend for this? Absolutely.
Could they excel at building a frontend that takes the world by storm the way Slack did, with its radical simplicity? Unfortunately I'm not as confident here. Consider that their VS Code plugin lags their terminal TUI so massively that it still is impossible to rename sessions [0], much less use things like remote-control functionality.
Show me that they can treat native-feeling multi-platform UI with as much care as they do their agentic loops, and I'll show you a company that could change every business forever.
When I use ChatGPT for work it frequently reads my Slack DMs even if they’re not directly relevant, so I’d question a lot of the premises of the article.
Use mattermost/zulip, and start contributing to the software you need. This isn't hard. Software isn't bestowed from the ai intelligence in the heavens, it's built by people.
I had high hopes for Claude's interactive app integrations, including Slack, but it leaves MUCH to be desired and doesn't really solve for agentic access patterns.
I actually vibe with this. I like the engineers and UX people at Anthro. And Slack is actually the most insecure hot mess of an enterprise app you can get.
If you want Anthropic to make a new slack, just ask Claude to write it for you. It wrote me a trello clone in 15 minutes. Why bother with a SaaS. You can build your own perfect chat system in a weekend.
Right. If these tools are so good (and they are) there should be numerous better-than-Slack apps by now that let you do exactly what you want. It doesn't take Anthropic to make it. (At our company, we cheated and edited 37signals' Campfire instead because we got sick of Slack's ads pushed into our paid instance.)
lol. This is rich coming from fivetran which extorts people for a relatively straightforward service that’s just annoying enough (looking at you salesforce + QuickStart views) to be worth buying.
But yeah slack could use some competition. Let’s see it would
Make sense. It would make anthemic even more sticky in the enterprise.