The future of work is about doing the work itself right where conversations happen, not just discussing what needs to be done. Slack is evolving from a communication platform into an agentic operating system where humans and AI agents collaborate seamlessly to accomplish tasks.
Kurtis Kemple, Senior Director of Developer Experience at Slack, is at the forefront of this transformation. He is building the infrastructure that enables both first-party and third-party agents to integrate directly into collaborative workflows. This shift did not happen overnight. It emerged from Salesforce's push to integrate AgentForce into Slack, which revealed critical insights about what it takes to support agentic experiences at scale.
Through partnerships with companies like Anthropic, Vercel, and Replit, Slack has been stress-testing what it means to build a platform that supports structured, grounded agentic workflows while maintaining the flexibility that makes Slack powerful.
Scaling orchestration with agentic workflows
The concept of agentic workflows represents a fundamental shift in how we think about productivity tools. Rather than simply automating individual tasks, Slack is enabling orchestrated chains of AI agents that can pass context and delegate responsibilities across multiple specialized systems.
When thinking about how humans and agents truly collaborate, Kemple notes that it goes beyond just writing code together; it is about actual handoffs of work.
Imagine an engineering manager who asks GitHub Copilot to triage their top pull requests and then hands those results to OpenAI Codex to open the top five in sandbox environments for review. All of this happens without ever leaving Slack.
This orchestration layer is where Slack's unique value proposition emerges. The platform acts as the integration point where user intent, multi-agent collaboration, and tool capabilities converge. Companies are already seeing the impact. Some organizations report saving millions annually by deploying agents that handle deal pipeline triage and categorization, effectively offloading toil without asking the AI to make critical business decisions.
From prompt engineering to context engineering
The evolution from prompt engineering to context engineering marks a critical maturity in how we build AI-powered systems. Kemple introduces the concept of "leaky prompts," which is the phenomenon where conversations inevitably drift into chaos unless actively managed. Because developers only control half of the experience (the system prompt, not the user's input), maintaining alignment requires treating the problem as one of information architecture rather than just prompt design.
Slack is a wonderful home for that context because its very structure provides hierarchical data. Threads, channels, messages, and canvases provide rich context that can be harnessed at different levels of granularity.
- A design team collaborating on a landing page might need thread-level context to feed Vercel's v0 agent.
- A knowledge management channel might require channel-level context to power a Q&A agent.
- Deep research projects might need workspace-level synthesis across multiple channels and time periods.
To support these varying needs, Slack is building purpose-built APIs like real-time search optimized for LLM consumption rather than end users. The platform provides SDKs that allow app developers to access and utilize Slack's context data while integrating secondary tools like Figma, Jira, and Vercel.
Building conversations, not applications
This strategic vision leads to a provocative piece of advice for developers. Kemple urges teams to stop building standalone apps and start building conversations instead. Since Slack already possesses a multi-turn, collaborative UI purpose-built for AI user experiences, developers can find product-market fit faster by building directly within Slack rather than creating custom web interfaces.
The conversational approach offers several advantages. It reduces the overhead of creating and maintaining custom interfaces. It enables rapid prototyping, allowing teams to iterate on conversation-driven automation without the friction of full application development. Finally, it naturally supports both individual and multi-agent collaboration scenarios in real time.
This flexibility makes AI feel like a natural participant in the workflow rather than a separate tool that requires context switching.
Accelerating integration with developer tooling
Slack's approach to AI-powered integrations prioritizes platform extensibility while maintaining a streamlined developer experience. The company has consolidated its tooling around the Bolt framework and a unified CLI that makes it possible to go from idea to deployed agent in weeks rather than months.
Kemple notes that they are working hard to make the "click to create" agent experience super simple. He envisions a future where developers can easily vibe code Slack apps and deploy them seamlessly into their workspaces.
The platform supports both ends of the technical spectrum. Workflow Builder provides a low-code solution for non-developers to create automation and orchestrate simple agent interactions. Meanwhile, developers get full SDK access to build complex integrations that span multiple Slack surfaces.
Eliminating toil to amplify productivity
The most actionable advice for engineering leaders comes from Kemple's personal approach to productivity. Rather than attempting to overhaul entire workflows at once, he recommends throwing AI at "toil" and time-dependent tasks.
Start with the most chaotic, annoying part of your day or week. Spend 30 minutes researching whether AI tools within your existing systems can address that specific pain point and invest another 15 minutes trying it out. When you start saving five minutes here and 15 minutes there, the productivity gains compound quickly.
Kemple uses personal agents for deep research and calendar optimization, tasks that represent "toil" because they must be done but do not require unique human judgment. By automating these tasks, leaders free up time for high-value work that only they can do.
Slack's evolution into an agentic operating system is a fundamental reimagining of how work gets done. By providing the infrastructure for humans and AI agents to collaborate naturally within existing conversational workflows, Slack is reducing the friction between intent and action. For engineering leaders, the message is clear. Stop building standalone AI applications and start building conversations.
For a closer look at how conversational interfaces are reshaping the future of work, listen to Kurtis Kemple discuss these ideas in depth on the Dev Interrupted podcast.
