Your non-technical colleagues are officially out of excuses for why they can’t fix that bug themselves.
With Slack Code, teams can now drop into temporary channels and build software directly alongside AI coding agents. Slack Code creates short-lived channels where anyone can prompt an AI agent to build software in public. Think of it as group-chatting a product into existence, while the AI does the actual heavy lifting. At launch, the tool plugs into Anthropic’s Claude, Cognition’s Devin, GitHub Copilot and Vercel, while OpenAI’s ChatGPT is set to join the list soon.
According to Slack, more than 70% of its internal code channels open and close within a single day, often taking a task from idea to merged pull request. Take internal metrics with a grain of salt, of course, but the speed advantage is noteworthy.
The Slack Code Workflow
To be clear, Slack Code isn’t an IDE, a new language or a code editor. It’s a multiplayer wrapper built on top of the AI agents businesses are already using.
The flow is simple: call out an agent in a message, it creates a single-purpose channel to execute the task. Everyone collaborates on live artifacts like code diffs, spec plans and HTML previews directly in the thread. When the pull request is done, the channel self-terminates, the thread is then archived for future context.
What’s different from how most AI-assisted development works today is the audience. Currently, the typical workflow involves an engineer opening an IDE or browser-based agent, working through a problem privately, and eventually bringing the output back to the team for review. Slack Code moves that interaction into a shared space from the start. The iteration happens in public, with context from across the business available to the agent and visible to everyone involved.
If engineering bandwidth is your main constraint, this dynamic changes things. A customer support lead who knows how to reproduce a bug can accelerate the minor fixes that are usually at the bottom of a sprint. Meanwhile, product managers get to stop filing tickets and start actively showing the AI what to build.
Slack’s own documentation gives examples beyond pure software development, including co-writing marketing plans and redlining legal contracts. That suggests the company sees this as a general collaborative workflow rather than an engineering-specific tool.
More from Artificial Intelligence
- AI Can Process Medical Data. But Does It Actually Understand How Our Bodies Work?
- The Death Of The Click: Google Gave Us The Click, Is AI Taking It Away?
- Quite Contrary: Kacper Staniul On Why Chat Is “The Worst Interface AI Has Ever Had”
- Laptops Are Getting More Expensive, And AI Might Be To Blame
- If AI Agents Skip The Dashboard, What Is The SaaS Interface Actually For?
- What Does Stripe’s $7 Billion Model Routing Deal Reveal About The Real AI Economy?
- Pax Silica And The AI Schoolyard: Is There Any Room Left For Neutrality?
- ChatGPT Can Now Remember What You Do On Your Computer – But Are We Giving It Too Much Context?
What Slack Code Doesn’t Solve
The breakdown happens at the validation stage. While natural language prompting allows non-technical employees to articulate their desired outcomes, it doesn’t enable them to assess whether the generated code is secure or architecturally sound. Democratising the input process in a shared channel doesn’t take the fundamental responsibility of code review away from qualified software engineers.
Slack confirms that code channels inherit workspace permissions and allow high-stakes changes to be routed for human approval. However, this only manages access control within Slack. It doesn’t address repository permissions, cloud infrastructure access, dependency auditing, secrets management or deployment gates – all of which still require separate, explicit governance.
The upside of collaboration is that more participants in a software workflow can mean more context and earlier feedback. The downside is that it can also mean conflicting instructions, unclear ownership and a false sense that the hard part was generating the code.
The hardest problems in software development rarely involve implementation speed. The true difficulty is in deciding what to build, navigating system constraints, managing dependencies across a growing codebase and determining if a change is safe for release. Slack Code may reduce the time it takes to go from an idea to a first draft. It doesn’t change how much expertise is required to decide whether that first draft should ever reach production.
This constraint isn’t a dealbreaker, but it does require realistic expectations. Teams struggling with requirement gathering and ticket scoping will find value in public AI channels. Conversely, organisations where code review, testing and deployment cause the primary delays will find that generating code faster simply shifts constraints further downstream.
