Overview of Pack your agentic stack in Slack
This Stack Overflow Podcast episode features Ryan Donovan interviewing Jamie DeLange, Chief Product Officer at Slack, about how Slack is positioning itself as a natural home for AI agents. The conversation covers Slack’s evolution from chatbots to a full agent platform, how Slack context can improve agent usefulness, and why keeping agents inside collaborative channels may reduce chaos, improve trust, and create better auditability for teams.
Why Slack is a strong fit for agents
- Slack already acts as the place where a lot of organizational context lives:
- conversations
- files
- channel history
- team knowledge
- Agents work well in Slack because they can tap into this real-time, unstructured context instead of relying on static docs or disconnected tools.
- Slack’s existing app and bot ecosystem gave it a head start as agents became more capable.
Slack’s agent strategy
Existing platform foundations
Slack has spent years building infrastructure that now supports agents, including:
- Block Kit
- App Home
- message actions
- bots and apps
- the new agents SDK
New capabilities mentioned
- MCP server: lets agents use Slack data and take Slack actions outside of Slack.
- Real Search API: gives agents real-time access to conversations, files, and channel context.
- Slackbot: Slack’s native AI assistant, designed to operate with current Slack context plus organizational context.
Context is the key unlock
Jamie emphasized that the real value is not just “searching everything,” but surfacing the right context.
Main ideas
- Slack has a strong recency bias, which can be a feature: what people are actively discussing is often the most relevant.
- Developers often want “all the context,” but too much context can create:
- context rot
- token bloat
- unnecessary complexity
- Slack can abstract away some of that burden by handling storage, ranking, and relevance for the developer.
Possible future contextual signals
Jamie hinted at broader context systems beyond chat history, such as:
- relationships between people
- organizational role and permissions
- channel-specific context
- richer smart APIs that help apps understand what matters without overloading the model
Slackbot as a personal and organizational agent
Slackbot is being positioned as more than a Q&A bot.
What it can help with
- answering policy questions like PTO
- searching Slack and connected systems
- understanding user role and organizational context
- acting as a bridge to other agents or apps
Example use case
A user could ask Slackbot how to file PTO, and Slackbot could:
- find the relevant HR policy
- determine the user’s context
- hand off needed data to a Workday-like system or custom HR agent
- reduce the need for a separate form-filling workflow
Collaboration, not solo-agent silos
A major theme of the conversation was avoiding isolated “agent holes” where a single person and their AI work in a silo.
Risks discussed
- too much trust in self-affirming agents
- work moving too fast without team review
- oversized PRs or projects that are hard for others to understand
- teams losing shared context when work happens privately
Slack’s answer
Slack wants agents to work in channels, where:
- the team can see what the agent is doing
- the agent can see team feedback
- work leaves an audit trail
- others can jump in and continue the work
This is especially useful for:
- code review
- triage channels
- bug fixing
- shared operational workflows
Noise, trust, and agent management
Jamie noted that agents are already increasing message volume and complexity inside Slack.
Observations
- AI interactions are growing rapidly inside Slack.
- Agentic apps have seen major growth.
- Agents are often verbose, creating more notification and review overhead.
Product direction
Slack is focusing on:
- better notification management
- agent discovery
- summarization
- helping users filter bot-only or human-only activity
- reducing the need for constant human supervision as trust in agents grows
Security, compliance, and enterprise readiness
Slack’s advantage for enterprises is that it already fits existing governance needs.
Benefits mentioned
- legal hold support
- data loss prevention
- security approvals already in place
- auditable message history and workflows
This makes Slack a more practical place to deploy agents than a completely separate system.
Key takeaways
- Slack is evolving from a chat platform into a context layer for agents.
- The company believes the best agent experiences are embedded in the same place teams already collaborate.
- Real-time search, MCP, and Slackbot are central to making agents more useful and less disruptive.
- The future likely involves a mix of:
- agents
- tools
- skills
- apps
- human collaboration inside Slack
Developer resources and next steps
If you want to build on Slack:
- visit slack.dev
- check out the newly redesigned developer docs
- explore agent support, notifications, Block Kit, unfurls, and the rest of the Slack app ecosystem
Notable stats
- Over 50% of conversational interactions in March were AI vs. non-AI
- 144% growth in agentic apps compared with the second half of the previous year
These figures suggest that agent usage inside Slack is already scaling quickly.