As AI agents become more embedded in workplace tools, Asana is positioning its approach around collaboration rather than individual productivity.
“We believe in AI being ‘multiplayer’ by design,” said chief product officer Arnab Bose. “The future of the agentic enterprise will only be realized if agents can work independently and with multiple people, versus just a copilot.”
Asana made its AI Teammates feature generally available this month; it’s a paid add-on that provides customers with AI agents that are capable of completing tasks autonomously within Asana’s work management platform. Users can build their own AI teammate agents or use one of 21 off-the-shelf agents focused on job roles such as marketing, IT, and operations.
Unlike standalone AI assistants, Asana’s AI teammates operate within shared workflows, with access to projects and portfolios across the platform, Bose said. Users can assign tasks to an agent, review its output, and provide feedback. An auditable record of prompts and actions carried out by the agent is then available to all co-workers.
The collaborative approach leads to greater transparency around information generated by agents, said Bose. “As you train the AI agent and get more work done within Asana, all of that reinforced learning is shared with the human beings with access to the same agent,” he said. “You’re getting institutional memory; you’re not just getting individual memory and individual productivity boost.”
Connecting to third-party apps
AI teammates are also able to connect to third-party applications via “a bi-directional sync” to retrieve data and take actions such as creating new documents, Bose said. Currently this means API connections to Google Drive and Microsoft 365 apps. Connectors to other business applications, such as HubSpot and Salesforce, are in development.
API connectors are well-suited to tasks such as updating docs, but for more non-deterministic tasks, Asana is working on MCP connectors for AI teammates. These are likely to arrive next quarter, Bose said, and are a better fit for unstructured agent-to-agent interactions.
An example would be AI teammates connecting to Slack’s revamped Slackbot. “Slackbot has an MCP agent,” said Bose, “so you can have the Asana AI teammate ask the Slackbot, ‘Hey, what’s up with this particular project? Are there any interesting comments in these channels which I should use to flag status risk?’ We would get back answers like that.”
Competing with third-party agents
Asana believes that AI teammates will also help protect the value of Asana as third-party AI agents such as Claude Cowork and ChatGPT become more capable of working autonomously across software-as-a-service applications — that’s part of the thinking behind the so-called SaaSpocalypse, where work applications such as Asana theoretically become the layer underneath a general-purpose AI agent.
Bose doesn’t see external agents as a risk to Asana’s value proposition, and there are two reasons for this, he argues.
One is that Asana’s embedded, collaborative AI agent approach is better suited to the complexities of enterprise work management than general purpose AI agents.
“If you take a look at how Claude Cowork or any of these coding agents slash productivity agents work today, they are highly optimized for individual or personal productivity,” said Bose. These agents are essentially disconnected from colleagues, he said, with research and reinforcement loops “all happening on an individual basis; they’re not happening in the project or task context.”
The other point, he contends, is that Asana still stands to benefit even if a user interacts with the work management app via a AI agent, as the agent still connect to Asana’s Work Graph — a data map of the relationships between all the work, people and information held in its platform — which allows Asana monetize these agent interactions.
“If it’s a single human who’s completing a task … [and] you don’t end up using Asana UI, you’re still getting massive benefits out of the Work Graph, which is ultimately good for us,” said Bose.
He added: “If you’re updating a document or you’re responding to an email, you could pull data out of Asana to get the most relevant organizational context … that just makes our Work Graph stickier.”
In the shorter term, Asana is likely to face more competition from other productivity software vendors that are also building AI agents into their apps.
Craig Le Clair, VP and principal analyst at Forrester, said many business applications already feature AI builder tools that can address similar use cases to AI teammates, “so this in itself not unique.” Yet Asana has a “data advantage,” according to Le Clair, as other AI agents may not be grounded in the enterprise data and human work patterns to the same degree.
Since it’s connected to Asana’s Work Graph, AI teammates agents provide better context than AI tools such as ChatGPT, said Le Clair, with the main competition for Asana coming instead from “horizontal” platforms that integrate AI agents into the product suites where workers already spend their day: Microsoft 365 Copilot and apps such as Teams, for instance, or Salesforce, with Agentforce and Slack.
Asana is “well-positioned against other pure-play collaborative work management vendors, but the real threat are these more general alternatives,” he said.
Asana’s AI Teammates feature is priced $15 per user per month for 100 agent “requests.” Customers that go over this limit will be charged at the same rate for an additional 100 requests, though Asana will waive these additional fees for a year for customers that sign up before July 31, 2026.