Slack code lets teams build software with AI together

Slack code makes AI coding a team activity

Slack Code changes how teams work with coding agents. Today, most developers interact with these agents in separate, individual sessions. Slack Code moves that work into shared, project-based channels where several people can participate at once.

The change expands who can take part in software creation. Engineers and product managers can work alongside designers, marketers, and other business users. Each participant can give the coding agent instructions in natural language, suggest changes, review results, and approve work. The technical execution remains visible to everyone involved.

Sateja Parulekar, vice president of product marketing at Slack, describes this model as a “multi-player AI experience.” She said marketers and product designers can now interact with coding agents alongside engineers and product managers in a “safe, governed environment.”

For executives, the important change is the cost of coordination. Software work often requires business requirements to pass through product teams before engineers translate them into code. Shared agent sessions can shorten this cycle. A product designer can comment directly on a prototype. A marketer can clarify a web-page requirement. An engineer can review the resulting code in the same workspace.

This broader participation requires clear ownership. Natural-language access makes it easier for employees to request software changes, but ease of access does not remove the need for engineering standards. Teams still need named owners for architecture, testing, security, deployment, and final approval. Slack’s governance controls become important as more employees gain direct access to code-producing agents.

The strategic value is therefore wider participation with controlled execution. Companies can bring business knowledge closer to software development while retaining technical review. If managed well, Slack Code can reduce handoffs and make product decisions visible to everyone responsible for the outcome.

Code channels automate the workflow around AI coding sessions

Slack Code also targets a more operational problem: setting up and managing the work around an AI coding task. A user can tag a coding agent in a Slack channel or direct message with a request. Slack then creates a dedicated code channel, adds links to relevant documents, and invites the people needed for the task.

These channels are designed for defined pieces of work. Examples include fixing a bug, updating a web page, or building an application feature. The team can exchange normal Slack messages and files while giving instructions to the coding agent from the same channel.

The agent’s work remains visible through “session artifacts.” These include code diffs, which show exactly what code the agent changed. Slack Canvas documents can capture supporting information. Live HTML can show working prototypes and previews. Participants can review these outputs, request revisions, and sign off on the result using natural language.

Slack also connects the temporary development session to the wider project. Updates from the code channel appear in the original Slack channel. When the task is finished, Slack automatically archives the dedicated channel. This creates a defined lifecycle for short AI-assisted development projects without requiring employees to maintain another collaboration space manually.

For executives, traceability is the key benefit. Giving an AI agent a task is easy. Understanding what it changed, who reviewed the change, and how the result moved toward approval is the harder operational problem. Code channels put those activities in a shared record where the relevant employees can see the work as it develops.

This model can also reduce context switching. Requirements, discussions, files, generated code changes, and previews can remain connected within Slack. That can shorten review cycles and make accountability clearer, especially when software work spans engineering and business teams.

The value will depend on workflow discipline. Automated channel creation does not guarantee good requirements or reliable code. Organizations still need testing, engineering review, security controls, and deployment processes. Slack Code can streamline the coordination layer around those controls, giving teams a more structured way to move from an AI request to a reviewed output.

Permissions and governance are the immediate challenge

Slack Code gives more employees direct access to coding agents. That increases the importance of access control. An agent may be able to read conversations, retrieve workspace data, modify code, and produce outputs that affect production systems. Each action needs clear authorization.

Slack says Code channels follow its existing security and permissions model. Agent access to conversations and data depends on controls already configured for the Slack workspace. Administrators can also add guardrails. For example, they can prevent employees outside engineering from shipping code until an engineer reviews it.

This distinction matters for executives. Natural-language interfaces reduce the technical skill required to request a code change. They also increase the number of people who can initiate development work. Governance must therefore define who can invoke an agent, which systems and data the agent can access, what actions it can perform, and who has final approval.

Will McKeon-White, senior analyst at Forrester, calls permissions “a persistent challenge for anything multiplayer.” He identified several approaches used by collaborative systems. Permissions can follow the user for a specific task, remain persistent while invocation rights differ between users, or derive from the shared room where the agent operates. McKeon-White said each approach has flaws.

The core issue is delegated authority. When an employee asks an agent to perform work, the system needs to determine whose permissions govern every subsequent action. That becomes more complex when several employees with different access rights participate in the same channel. Companies need policies that remain predictable when users, agents, repositories, and sensitive information interact in one workflow.

For CIOs and CISOs, deployment should therefore start with permissions and approval paths. High-impact actions deserve tighter controls than low-risk tasks such as generating a prototype. Engineering review, repository permissions, testing, and deployment authorization should remain explicit. Slack Code can make AI development more accessible while administrators retain control over how generated work reaches business systems.

Slack code launches with four agent integrations and no additional subscription charge

Slack Code starts with four integrations: Anthropic’s Claude Code, Devin, GitHub Copilot, and Vercel. Slack plans to expand the available options over time. The feature is available across all Slack subscriptions at no extra cost.

Supporting several coding tools is strategically important. Enterprises already have different development environments, approved vendors, security requirements, and developer preferences. A shared Slack interface gives teams a common place to coordinate work while allowing different agents to handle the underlying coding tasks.

This approach can also reduce workflow fragmentation. Employees can initiate agent work from Slack, bring relevant participants into the session, review generated artifacts, and keep project discussions connected to the coding activity. Teams can choose an agent appropriate to their existing technology environment while maintaining a consistent collaboration process.

For CIOs, the four integrations create a choice architecture that requires governance. Each agent has its own capabilities, security characteristics, data-handling policies, permissions, and integration model. Enterprises should decide which agents are approved for specific repositories, data classifications, and development tasks. Vendor access should align with existing software-development and AI governance policies.

The absence of an additional Slack charge lowers the initial barrier to enabling Code channels for existing customers. Total operating cost will still depend on the coding agents selected, existing commercial agreements, implementation requirements, and internal governance. Executives should therefore evaluate the workflow as a complete system rather than treating Slack availability as the full cost calculation.

The larger strategic point is interoperability. Slack is positioning its workspace as the place where employees coordinate with a range of AI agents. Supporting Claude Code, Devin, GitHub Copilot, and Vercel at launch gives enterprises several paths into that model. Expanding the integration catalog could make Slack increasingly important as the management and collaboration layer through which teams use specialized AI tools.

Slack code brings software creation closer to the conversations that drive it

Software development already involves extensive coordination in Slack. Requirements, bug reports, design decisions, approvals, and status updates often move through channels before developers act on them in separate tools. Slack Code brings AI-assisted code creation into that existing workflow.

Wayne Kurtzman, research vice president at IDC, said Slack Code “closes the gap between where work gets done and where code gets written.” Dedicated code channels allow employees to coordinate a project and interact with coding agents in the same environment. Kurtzman sees this as supporting Salesforce’s wider goal of making work more seamless and collaborative within Slack across the applications a business uses.

Will McKeon-White, senior analyst at Forrester, makes a related point. Software development and its coordination have “happened in Slack forever,” he said. Slack Code makes it easier for employees to remain in Slack and broadens participation in code creation.

For executives, the main benefit is lower coordination overhead. A feature request can involve product managers defining requirements, designers reviewing interfaces, engineers assessing implementation, and business stakeholders approving the outcome. Giving those participants access to a shared agent session can shorten handoffs and preserve the context behind decisions.

The model also improves visibility. Participants can see requests, agent responses, code changes, previews, and approvals in one collaborative space. That can make it easier to understand why a change was requested and who participated in its review. This context becomes valuable when projects move quickly or involve several functions.

However, reducing context switching should remain a means to an operational outcome. The useful measures are shorter development cycles, faster reviews, fewer coordination delays, and reliable software delivery. Keeping employees inside Slack has limited value by itself. Executives should assess whether the workflow improves those business and engineering outcomes.

Adoption will also vary between organizations. Kurtzman said Slack Code “may seem early to market for some companies,” while others are using a first-mover advantage. Enterprises with mature AI coding practices and strong development controls are likely to have a clearer path to collaborative agent workflows. Organizations still establishing AI permissions, review processes, and engineering standards will need to resolve those foundations first.

Dynamic multi-agent collaboration creates a harder governance problem

Slack Code currently focuses on people collaborating with coding agents in shared channels. The longer-term technical challenge is deeper collaboration between multiple AI agents. These systems could divide work, coordinate tasks, and respond to each other with less direct human intervention.

Will McKeon-White, senior analyst at Forrester, identifies this as the longer-term challenge. “Multi-agent collab today in successful environments isn’t very dynamic,” he said. He added that “true dynamic agents cooperating can create extremely interesting emergent behavior.”

Emergent behavior means outcomes that arise from interactions between agents rather than from a single predefined sequence of instructions. An agent may adjust its actions based on another agent’s output, which then changes subsequent decisions across the system. As this process becomes more dynamic, predicting the complete workflow in advance becomes harder.

McKeon-White pointed to recent Anthropic research on multi-agent systems. Anthropic has explored architectures in which multiple agents work together on complex tasks, providing evidence that coordinated agent systems can support forms of work that benefit from parallel investigation and delegation.

For business leaders, the constraint is control. A single agent already raises questions about permissions, data access, validation, and accountability. Multiple cooperating agents multiply the number of interactions that governance systems must track. An action initiated by one agent may influence another agent with different tools, context, or access rights.

This makes observability essential. Enterprises adopting multi-agent workflows need records of which agent performed an action, what information it used, which permissions applied, and how its output influenced subsequent actions. Human approval points should reflect the risk and business impact of the task.

The opportunity remains significant. Dynamic agent collaboration could allow complex projects to be divided across specialized systems and executed with greater automation. The organizations best positioned to use that capability will be those that can preserve permissions, traceability, testing, and accountability as autonomy increases. For executives, multi-agent adoption should therefore advance alongside the controls required to understand and govern agent behavior.

Slack plans to extend collaborative AI workflows beyond software development

Slack Code starts with software development, but Slack sees the underlying model applying across more business functions. The core design is a shared workspace where employees and AI agents can work on a defined task, review changes, and approve an output together.

Marketing is one potential use case. Teams could use an agent to develop campaign materials while marketers, designers, product teams, and approvers review the work in a shared channel. Changes and decisions would remain connected to the project rather than being spread across separate conversations and AI sessions.

Legal document review is another example. Sateja Parulekar, vice president of product marketing at Slack, said these workflows can involve a “very detailed review of a document and tracking changes and a preview of what’s going to be sent out to the customer.” The same collaborative structure used for code could therefore support documents and other business assets that require several rounds of review.

For executives, the larger opportunity is workflow design. Many knowledge-work processes follow a similar sequence: gather context, generate or revise an asset, obtain specialist input, track changes, secure approval, and deliver the result. Agents can participate directly in these stages while Slack provides the shared workspace for human oversight.

Governance becomes more important as this approach reaches additional departments. A legal agent may handle confidential documents. A marketing agent may work with customer data or brand assets. Each workflow requires access rules, review requirements, retention policies, and clear accountability appropriate to the business risk involved.

This creates a practical test for expansion. Organizations should identify workflows with frequent handoffs, repeatable review stages, and clear ownership. Those processes provide a stronger basis for collaborative agents because management can define what the agent may do, which people need to participate, and what must be approved before completion.

Parulekar expects the model to reach a wider set of teams. “In the future, we see this extending to a lot of different teams and use cases for technical and non-technical users,” she said. That direction makes Slack Code an early implementation of a broader agent-based collaboration strategy.

Slack is positioning itself as a shared workspace for people and AI agents

Slack Code sits within a wider effort to make AI agents a regular part of workplace collaboration. Slack is adding different ways for employees to engage with agents while keeping those interactions connected to their daily work.

Two additional features support this strategy. “Agent DMs” provide a space for individual conversations between a user and an agent. A new “agents tab” gives users a central place to view existing agent conversations. Slack has also announced Claude Tag, which makes Anthropic’s agent available to teams.

Together, these capabilities support both individual and group interaction. An employee can work privately with an agent through a direct message and use shared environments when a task requires input, review, or approval from colleagues. Code channels add a structured workspace for development projects, including planning, coding, testing, and reviewing outputs.

Sateja Parulekar, vice president of product marketing at Slack, describes the strategy as a “multi-player experience” for people and AI agents. “Whether it’s the user experience that we’re evolving, like the agent tab, or coding channels where you can plan, code, test, and ship inside of Slack, it’s all part of that bigger vision of making Slack a great home for agents and furthering that vision of multiplayer AI,” she said.

For C-suite leaders, this strategy has implications beyond developer productivity. AI adoption can create fragmented workflows when employees move between communication tools, standalone AI interfaces, business applications, and approval systems. Slack is seeking to concentrate more of the interaction and coordination within a workspace employees already use.

That positioning also changes the role Slack wants to play in the enterprise technology stack. Slack increasingly serves as an interaction layer through which employees can call agents, provide business context, involve colleagues, inspect results, and manage work across other systems. Its integrations with Claude Code, Devin, GitHub Copilot, and Vercel demonstrate how that model can support third-party AI services.

The strategic question for executives is whether Slack can provide sufficient governance and interoperability as agent use grows. Centralizing agent interactions can improve visibility and simplify the employee experience. It also makes permissions, auditability, data controls, and vendor management central requirements.

The direction is clear: Slack wants AI agents to become participants in the same workflows used by human teams. Slack Code provides a concrete implementation of that strategy. Its long-term business value will depend on whether organizations can connect agent capabilities to measurable work outcomes while maintaining effective control over access and execution.

Concluding thoughts

Slack Code addresses a real constraint in enterprise AI adoption: coordination. Coding agents can generate work quickly, but software delivery still depends on shared context, technical review, clear permissions, and accountable decisions. Moving those activities into a common Slack workflow could reduce handoffs and make AI-assisted development easier to govern.

For executives, the key question is whether this model improves business outcomes. Measure development cycle time, review speed, engineering workload, and the quality of delivered software. Wider employee access to coding agents has value when it produces faster, controlled execution.

Governance should come first. Define which agents employees can use, what data and systems they can access, who can approve generated code, and which actions require human review. These controls become more important as Slack expands the same collaborative model into marketing, legal, and other functions.

Slack’s broader direction is significant. It wants to become a shared workspace where employees and specialized AI agents plan and execute work together. Slack Code gives enterprises a practical way to test that model today. The companies that benefit most will combine faster agent-driven execution with strong permissions, traceability, and human accountability.