Perplexity’s new Computer in Slack feature takes the chat app people already live in all day and quietly turns it into a shared AI workspace that can actually get work done, not just draft a few emails. Instead of bouncing between tabs and tools, you can now pull in an AI “teammate” that understands your company’s context, works across your stack, and leaves its work right where the rest of the team is already talking.

At its core, Computer is an AI orchestrator: it doesn’t just answer questions, it coordinates a whole crew of specialist models and tools on a secure server, writing code, doing deep research, querying connected apps, and generating finished artifacts like reports, docs, decks, and even simple apps. Under the hood, it can tap things like Claude Code and GPT-style coding engines, route tasks across 400‑plus integrations, and then hand you back something tangible: a marked‑up contract, a filled‑in spreadsheet, a slide deck ready to present. The wild part is how much “work” that adds up to when you let it loose on real company tasks: Perplexity says that in its very first two weeks of limited availability, Computer saved Max subscribers over $91 million worth of labor based on benchmarks from firms like McKinsey, Harvard, MIT, BCG, and Nielsen, and that it has now crossed $776 million in labor‑equivalent work for Enterprise, Pro, and Max users combined.

What makes this story interesting is that Computer wasn’t dreamed up as some abstract enterprise platform; it was born inside Perplexity’s own Slack workspace. The earliest version lived in a single shared channel and went by the tongue‑in‑cheek name “ASI” – short for “artificial superintelligence” and “a Slack integration” at the same time. People would just tag it in a thread and ask it to do the kind of jobs that normally eat up afternoons: pull data from Snowflake, research a market segment, prep a board memo, write or refactor code. Within its first four weeks, just as an internal Slack bot, it had already performed $1.6 million worth of work for about 300 employees, again using external productivity benchmarks to estimate value.

The numbers were impressive, but something else was happening in that public channel that’s arguably more important: everyone could see how everyone else was using Computer, and that made the tool itself better. A single well‑crafted prompt from a finance lead or an engineer would turn into a pattern the rest of the company could copy, remix, or improve on. New workflows emerged out in the open: someone tags Computer to clean up a messy dataset, someone else realizes they can extend that into a weekly automated report, and suddenly there’s an informal library of “plays” the whole team can run. That’s the key idea Perplexity is betting on with Computer in Slack: the AI doesn’t just assist individuals, it sits inside the shared context where collaboration already happens.

In practice, that shared context is what separates “an AI chatbot in Slack” from “a shared AI workspace”. Every request to Computer in Slack lives in a thread along with all the back‑and‑forth: the original ask, clarifying questions, links to docs or dashboards, the output, and everyone’s feedback in between. Teammates can jump into the same thread and add details the AI needs, correct assumptions, or attach reference files, while project leads can transparently hand off work in public instead of quietly forwarding emails or tickets around. When the result comes back, it’s right there in the same channel where the conversation started, so you’re not chasing links to some separate tool just to see what the AI did.

That design fits how people actually use Slack day to day. If a task touches multiple stakeholders or needs visibility – a product spec, a legal review, a launch checklist – you just tag @Computer in a shared channel and describe what you need in plain language. For more private or half‑baked work, like a first draft of a note or some early‑stage research, you can DM Computer instead and treat it like a personal coworker who doesn’t mind rough ideas. In both cases, Computer might ask follow‑up questions before it starts, which is usually a sign that it’s trying to nail down the exact output format or decision criteria before it disappears to go do the job. Once that back‑and‑forth is done, it works in the background so you can keep moving on to something else while it grinds through your request.

The range of work you can offload is pretty broad because Computer isn’t limited to text generation; it’s more like a general‑purpose operator that can read, write, and act across your tools. In Slack, you can ask it to search the web, analyze business data, generate content, or stitch those together into multi‑step workflows like “research, then draft, then format, then deliver”. Perplexity’s own examples are the kind of jobs that usually involve multiple people and tools:

• “Review this partnership proposal, fact‑check the statistics, flag risky terms, and return a marked‑up Word document with suggested edits.” Computer can read the document, hit the web to verify claims, apply basic legal and business heuristics, and then give you a redlined version to review.
• “Draft this quarter’s OKR kickoff deck using our strategy docs, prior results, and current backlog.” Here, it has to synthesize internal documents, recent performance, and future plans into slides that feel tailored to your team, not generic templates.
• “Find the sales call appointments that showed up in GoHighLevel, review the recordings, and send a summary report to the sales manager in Slack.” This is straight‑up operations work: query a CRM, process call recordings, extract key points, and then post the summary to the right person in the right channel.

Crucially, Computer isn’t locked into Slack. Everything you or your teammates do with it in Slack is mirrored in the main Perplexity web app, so you can start a task in a channel and then continue it later in the browser without losing the context. If you kick off a complex research project from Slack but want to refine the output, add more sources, or spin off multiple variations, the full history is sitting in Perplexity’s interface ready for deeper editing. That two‑way bridge matters for real teams: Slack is where you launch and coordinate work, while the web app is where you might polish it, plug into more data sources, or rerun workflows over time.

Under the hood, the new MCP connector gives Computer much deeper access to what’s actually happening in your Slack workspace, which makes its help feel less like a generic chatbot and more like someone who works there. With that connector, Computer can search workspace knowledge, read through relevant threads, and send messages, all while staying grounded in the ongoing conversations and documents around it. That context is what lets it, for example, draft a launch email that matches how your team already talks about a product, or pull the right internal doc instead of the first thing it finds on the public web. It’s a step toward AI that really lives inside your organization, not just next to it.

Installation is intentionally straightforward: you can add Perplexity Computer directly as a Slack integration or grab it from the Slack Marketplace like any other app. Once it’s installed and connected to your stack, you don’t have to teach people a new tool; they just learn a new teammate’s handle. For teams that are already paying for Perplexity Enterprise, Pro, or Max, Computer in Slack becomes a kind of amplifier: it takes the raw modeling power they already had and drops it directly into the channels where the work is being discussed, scoped, and reviewed.

Taken together, Computer in Slack is less about yet another bot and more about a different way of working: one where big chunks of knowledge work can be delegated to an AI operator that everyone can see, shape, and reuse in real time. When the requests, the context, and the output all sit together, you don’t just move faster – you build a shared playbook of how your team works with AI, visible in every channel, ready for the next person to pick up and push a little further.