Project management software giant Asana is rolling out what it calls “AI teammates”—bots that can participate in handling and discussing work via Asana’s platform in similar ways to actual humans. 

Unlike some AI assistant and copilot products that take direction only from one human user, Asana’s AI teammates are designed to work with multiple humans, similar to how an actual new hire might receive assignments, feedback, and comments from a range of coworkers.  

The aim is to offer a set of AI tools that integrate not only with software companies already use, but with the Asana-based workflows they rely on to divvy up and discuss work. So while the bots won’t necessarily replace human employees or potential hires outright, interacting with them through Asana will feel fairly similar to working with a flesh-and-blood colleague.

“It is a shared experience, which means you can bring it into a project, and it’ll behave and look like a team member, and it can pick up tasks,” says Asana chief product officer Arnab Bose. “And when it picks up tasks, everybody on the team who is a human being, who’s on the project, can give that AI agent feedback.” 

The feature is launching with 21 prebuilt virtual teammates that can handle tasks like planning product launches, drafting marketing campaign briefs, managing IT service queues, and coding web content. Users can also create their own AI teammates with custom prompts. The bots can be added to conversations on Asana and draw on the company’s existing Asana Work Graph, a data structure that maps relationships among projects, people, and tasks.

That context helps them understand their assignments and suggest relevant collaborators or files to reference. AI teammates can also be scheduled to perform routine tasks, such as scanning an Asana board daily or weekly to flag potential issues that could affect deadlines.

Critically, Asana’s virtual teammates can also read and write to files in cloud systems like Google Drive and Microsoft SharePoint, which means they can directly access and contribute to projects in the environments where companies already work. For instance, a marketing campaign brief creator can pull in existing notes and high-level strategy documents to guide its drafts. Additional integrations with tools like customer-relationship management (CRM) software are likely on the way, as are features to let users craft custom integrations. 

Already, the bots can even be directed to feedback given through comments in word processing documents, meaning users don’t have to radically change their workflows to accommodate their new digital teammates. 

“You don’t have to go ahead and learn a new way of working,” says Bose. “It’s not like you’d end up with some special format inside Asana, and now people have to stop using Google Docs or Office 365 and start using this.” 

Humans can still control at a fine-grained level what the AI teammates have access to. Teams can also configure who has what level of control over the bots. For instance, advanced users may be authorized to edit AI teammates’ stored memory data, which can help clear out erroneous instructions they may have picked up from less experienced colleagues, Bose says. And the agents generally won’t respond to users who aren’t paying for AI teammate access, which currently costs $15 per month per user, though other Asana users will still be able to see the bot discussions and any other output they generate. 

That $15 monthly fee includes the cost of calls to underlying AI models, so customers don’t have to pay separately for AI use or budget AI access tokens. Asana routes calls to the appropriate AI models for each task, and Bose says he anticipates many users will see that and other aspects of AI management being handled behind the scenes by Asana as a benefit rather than a limitation.  

The release of teammates, which Bose says will be available immediately to customers with enterprise-style sales relationships with Asana and self-service customers over the summer, comes as the software industry rolls out ever more ways to work with AI agents. Those include integrations with office software suites and creativity platforms, specialized desktop and command-line apps letting AI manipulate local files, and the familiar web chat interfaces that made large language models mainstream.  

In essence, Asana is betting that integration with existing work graph data, along with its collaborative approach to working with AI, will make the project management platform the preferred hub for managing AI. 

“Asana is designed for this moment, frankly,” says Bose. “It’s designed for this moment where every individual contributor is becoming a manager of multiple AI agents, and you need to coordinate them, and you need to align with other human beings.”