By admin 
The core proposition of Slack Code is elegantly simple yet profoundly impactful: currently, the vast majority of AI coding agent work occurs in isolation, a one-on-one interaction between a single developer and an AI, hidden from the rest of the team. Slack Code transforms this into a "multiplayer" experience. When a user initiates a task by tagging a coding agent within any conversation, the agent automatically spins up a dedicated, project-specific code channel. All subsequent work, including code diffs, live previews, and the evolving project plan, is conducted openly within this channel, readily accessible through dedicated tabs. Once the task is completed, the channel is archived, leaving behind a meticulously documented, searchable audit trail of the entire process.
Rob Seaman, Slack’s interim CEO, articulated the vision during a press briefing, highlighting the shift in bottlenecks. "One of the things I love about this is that code is no longer the bottleneck," Seaman stated. "Ideas, taste, judgment, craft – those are the things that are the bottleneck, and you’ve effectively extended the population that can contribute ideas, taste, judgment, and craft to anybody that exists in your Slack." This statement underscores Slack’s belief that by democratizing access to AI coding capabilities and making the process transparent, the collective intelligence and diverse perspectives of the entire team can be leveraged, accelerating innovation and improving the quality of the final product.
This launch carries significant weight for Slack, positioning it as a pivotal player in the rapidly evolving enterprise AI landscape. The current AI coding boom has largely been characterized by individual productivity gains, with developers leveraging tools like Claude Code or OpenAI’s Codex in solitary terminal sessions. Slack is strategically betting that the next frontier in AI development lies in controlling the collaborative layer that surrounds these powerful agents. This bet is particularly crucial for Salesforce, Slack’s parent company, which is actively seeking to revitalize its narrative and demonstrate the tangible value of its AI investments.
Slack Code Channels: AI Coding Agents at Work in the Open
Slack executives, in collaboration with Cognition leaders, detailed a workflow for Slack Code channels that draws parallels to a newsroom rather than traditional pair programming. This model emphasizes transparency, rapid iteration, and broad participation. Jeff Wang, president of new enterprise at Cognition, the creators of the Devin coding agent, showcased a live demonstration that vividly illustrated this new paradigm. In the scenario, a developer reported a bug within an engineering channel. Devin acknowledged the issue with an emoji, responded directly within the thread, commenced its investigation, and subsequently opened a pull request. Wang highlighted the agent’s intelligence, noting, "It even knows the code owner, so you can see it tagged Theo into this as well." He further elaborated on Devin’s operational independence: "Every time Devin is doing something like this, it does have its own computer. So here, it’s actually using Chrome and the DevTools to test if the feature is working correctly."
Following Devin’s initial work, the task seamlessly transitioned into a dedicated code channel. This is where the "multiplayer" aspect truly shines. Any team member – be it an engineer, a product manager, or a designer – can readily engage with the ongoing development. In Wang’s demonstration, a designer, mid-task, uploaded a Figma file into the channel, and the AI agent incorporated the design element without disruption. The agent concluded its work by presenting the code changes, accompanied by illustrative screenshots and a recorded demonstration verifying the successful implementation of the feature. This robust verification loop is central to Slack’s pitch. Unlike agents operating on a developer’s local machine, cloud-based agents like Devin can generate a verifiable, auditable record of their work, ensuring accuracy and completeness. Wang emphasized the scalability and auditability benefits: "Scaling things, auditing things, giving it to everybody – that is much easier with these cloud agents form factor than it is with local agents."
The impact of Slack Code extends beyond just code channels. Slack is also introducing a broader overhaul of how AI agents are integrated into the platform. This includes agent direct messages (DMs) that mimic conversations with human colleagues, a new "Agents" tab providing a centralized hub for all agent sessions with live status updates and control buttons, and an streamlined "Add to Slack" flow. This flow simplifies the deployment of agents from various platforms, including Lovable, n8n, OpenAI, LangChain, and Airtable, requiring only a few clicks with automated OAuth and configuration.
The Strategic Reimagining of Software Development: Code as the Least Scarce Resource
The fundamental strategic argument underpinning Slack Code is that AI has fundamentally inverted the economics of software development. Historically, the act of writing code was the scarce and expensive component of the development lifecycle. Slack’s executives contend that AI has rendered this step significantly more accessible and less costly. Consequently, the primary constraint has shifted upstream, to human judgment, creativity, and strategic decision-making.
"One of the things I love about this is that code is no longer the bottleneck," reiterated Seaman during the press briefing. "Ideas, taste, judgment, craft – those are the things that are the bottleneck, and you’ve effectively extended the population that can contribute ideas, taste, judgment, and craft to anybody that exists in your Slack." This perspective suggests that by automating the more rote aspects of coding, AI empowers humans to focus on higher-level strategic and creative endeavors, accelerating the overall innovation process.
Cognition provided internal data to substantiate these claims of enhanced velocity. Wang reported, "We’ve seen our internal merged PR count go up 10x in the last few months, versus our headcount has only gone up like 40 percent." He described a workflow where engineers initiate a Devin task, then pivot to another project, and subsequently launch another agent. "Soon you have everybody working on like dozens of agents at a time," he remarked, illustrating a significant increase in development throughput.
Wang further elaborated that this pattern transcends traditional engineering roles. "Believe it or not, a lot of our bugs are reported by our sales team. They report it in Slack, and then someone who’s technical applies them to fix the bug." Seaman seized upon this example as a microcosm of Slack Code’s overarching thesis: "So much of that stuff never even made its way to a product manager into a backlog because the communication vehicles weren’t there, the motivation wasn’t there, the knowledge that it could actually be fixed so quick wasn’t there – and we’ve effectively knocked all of that down." This highlights how Slack Code can streamline the entire bug reporting and resolution pipeline, fostering greater cross-functional collaboration and efficiency.
Wang extended his vision to the future of development, predicting the automation of "toil work – fixing bugs, fixing CI/CD, or fixing vulnerabilities, all these things engineers probably don’t want to do." What will remain, he asserted, is work that "requires creativity, planning, business logic." He offered a prediction that might raise eyebrows among some engineering leaders: while human oversight is currently mandatory for every code merge, "I suspect maybe in the next year it’s just going to go through automatically." This suggests a future where AI agents are trusted to handle routine code integration, freeing up human developers for more complex and strategic tasks.
Addressing the "AI Slop" Concern: Transparency as the Antidote
A primary concern surrounding the democratization of software creation through AI is the potential for a surge in "AI slop" – the term used to describe plausible-looking but poorly conceived output generated at scale by inexperienced users. Slack’s executives, however, propose a counterintuitive solution: visibility.
"The multiplayer part is a guard against that, actually, because people can see your work, people can comment on your work," explained Katie Steigman, Slack’s VP of product. She contrasted this with the current opaque model: "If I’m doing God knows what in terminal with an agent, versus being able to do it in a place where people can see my intent and actually change and shape my work – or slap my hand and tell me that’s slop, because that’s real." This emphasis on collective review and feedback is intended to act as a natural quality control mechanism.
Steigman, a product manager herself, shared her personal workflow as an example. "When I put PRs up as a product person, I almost always tag in an engineer from my team. I don’t just send a PR and ask for an approval," she stated. "Almost every time, an engineer will say something like, ‘Come on, you can make that a little bit tighter,’ or they’ll actually give it some specific technical guidance, and the agent will take one more rev and produce code that has been touched by an engineer to a certain extent." This illustrates how even non-technical contributors can leverage AI agents, with technical colleagues providing oversight and refinement, ensuring higher quality output.
Seaman framed the argument more broadly, suggesting, "I think the moral arc of multiplayer AI bends towards higher quality and less duplication." He referenced Shopify, where CEO Tobi Lütke has reportedly advocated for restricting agentic coding to public channels. The rationale is that this approach "immediately disseminates every single thing that’s happening in the company" and levels the playing field. Despite these optimistic projections, the skepticism regarding AI’s reliability is not unfounded, with industry data providing supporting evidence.
Gartner’s projections from last year indicated that "more than 40 percent of agentic AI projects will be canceled by the end of 2027," citing escalating costs and unclear business value as primary reasons. Similarly, McKinsey’s most recent "State of AI" survey revealed that while 62 percent of organizations are experimenting with AI agents, only about a third have begun scaling AI adoption, and just 39 percent report any discernible bottom-line impact. The persistent gap between enthusiasm for AI agents and their tangible value creation remains a defining challenge in the market Slack is entering.
Inside Slack Code’s Security Model: Inheriting User Permissions
For enterprise clients, a critical design consideration within Slack Code is its robust permissions model. When directly questioned about the potential for agents in code channels to leak access across teams – for instance, granting finance access to a legal repository – Seaman offered a firm assurance: agents inherit the precise permissions of the human user who invokes them, and nothing more.
"Everything is done on behalf of the user, using the user’s ACLs, both in Slack and in the systems that they’re connecting to," he clarified. "There’s no god permissions or bot-level permissions… Within Slack, the agent has access to information that the user has access to, and access to the channels that it’s been added to." This model ensures that AI agents operate within the existing security framework of the organization, mitigating the risk of unauthorized data access.
Steigman further elaborated on the security protocols. She explained that when an agent initiates a code channel, "the only thing that agent gets from the code channel is the context of the conversation" that triggered the task. On the execution front, Wang confirmed that Devin operates within isolated sandboxes with "minimum viable access," and importantly, includes an optional mode that completely disables internet connectivity. "You’ve heard all these stories about the agent escaping and causing havoc," he acknowledged, "but we have different security configurations." This multi-layered security approach is designed to prevent agents from acting autonomously or maliciously.
This "agents as extensions of existing users" model represents a significant differentiator compared to standalone agent platforms. Such platforms often necessitate IT departments to provision new service identities and manage a complex array of one-off permissions. Slack Code elegantly sidesteps this challenge, addressing the pervasive issue of shadow IT. Because agents produce standard pull requests into platforms like GitHub, existing release gates and review processes remain fully intact. "It reduces that barrier upfront to get that initial PR up," Steigman noted. "Then the due diligence still happens in GitHub." This seamless integration ensures that the adoption of AI agents does not bypass established security and governance protocols.
Slack Code and Salesforce’s High-Stakes AI Turnaround
The introduction of Slack Code arrives at a critical juncture for its parent company, Salesforce. The company’s stock experienced a significant decline of approximately 18 percent in the year leading up to January, significantly underperforming the Nasdaq. This downturn was fueled by Wall Street’s skepticism regarding AI’s potential to erode demand for traditional enterprise software. Compounding these challenges, OpenAI poached Slack CEO Denise Dresser in December to serve as its chief revenue officer, leading to Rob Seaman’s elevation from product chief to interim CEO of Slack.
In response to these pressures, Salesforce has aggressively pivoted to position Slack as the central AI gateway for enterprise workflows. In January, the company launched a revamped Slackbot powered by Anthropic’s Claude, which was quickly hailed as the "fastest-adopted feature" in Salesforce’s 27-year history. Slack Code represents a strategic extension of this initiative, moving beyond mere question-answering to the active production of tangible artifacts within Slack itself, including functional code, prototypes, and documents.
A notable strategic shift is also evident in today’s announcement. In mid-2025, Reuters reported that Salesforce had initially moved to "block rival AI firms from accessing Slack data," a defensive maneuver aimed at protecting its proprietary information. In stark contrast, today’s announcement positions Slack as an open platform actively courting these very AI companies as partners. The company plans to open its code channel APIs to all developers, signaling a significant change in its approach. Software engineering is presented as just the initial use case, with future applications in marketing campaign generation and legal document review within dedicated agent channels already on the horizon. The strategic calculus appears to have evolved from safeguarding Slack’s data to making Slack indispensable as the primary venue where AI agents – regardless of their origin – perform their work.
Industry partners have responded positively to this strategic pivot. Vercel CTO Malte Ubl expressed enthusiasm, stating, "A whole team can gather in one code channel, watch the agent work, steer it together, and ship a preview." Mario Rodriguez, GitHub’s chief product officer, echoed this sentiment, calling Slack "a strategic part of a broader GitHub promise: humans set direction, agents close the loop." These endorsements highlight the perceived synergy between Slack’s collaborative platform and the evolving capabilities of AI agents.
The Future of AI Coding: Coexistence of Multiplayer Channels and Single-Player Terminals
While Slack is championing its multiplayer approach, its executives are not suggesting the obsolescence of the single-player terminal environment. When asked about the future of tools like Claude Code and Codex, Seaman predicted a clear division of labor. "The overwhelming majority of the work is actually going to happen in these multiplayer environments," he stated. "But there’s going to be deep, immersive, intensive, single-player thought work that’s going to happen in terminals." This suggests a scenario where complex architectural refactoring or in-depth problem-solving might still occur in the focused, isolated environment of a terminal, while routine tasks and collaborative development will migrate to Slack Code channels.
Seaman further posited that the adoption curve for these tools will mirror previous waves of automation. "People are going to open these things at the beginning" – meticulously reviewing every code diff and every step of the agent’s thought process – "and then build trust in the system and open it less and less over time." This gradual acclimatization process is expected to foster confidence in AI-generated code and facilitate broader adoption.
Ultimately, Slack Code represents more than just a product launch; it signifies a fundamental workflow redesign aimed at making teams, rather than individuals, the primary unit of AI adoption. McKinsey’s research consistently emphasizes that organizations achieving significant value from AI are those that fundamentally redesign their workflows around these new technologies, rather than simply appending them to existing processes. If Slack Code succeeds, the company that once redefined how colleagues communicate could very well redefine where software is created. Conversely, if it falters, the increased transparency might simply lead to everyone witnessing the collective accumulation of "AI slop."
Regardless of the ultimate outcome, the era of the lone developer silently interacting with an AI agent in a private tab appears to be drawing to a close. As Wang aptly put it, "The bottlenecks have shifted." The crucial question Slack Code aims to answer is whether the collective intelligence and collaborative oversight of the "crowd" will effectively shrink these bottlenecks, or merely amplify their visibility and the noise surrounding them.
