The key to successful AI agents within an enterprise? Shared memory and context.

This, according to Asana CPO Arnab Bose, provides detailed history and direct access from the start – with guardrail checkpoints and human supervision, of course.

This way, “when you assign a task, you don’t have to go back and re-give all the context of how your business works,” Bose said at a recent VB event in San Francisco.

AI as an active teammate, rather than a passive add-on

Asana launched Asana AI Teammates last year with the philosophy that, like humans, AI agents should be plugged directly into a team or project to create a collaborative system. To pursue this mission, the project management company has fully integrated with Anthropic’s Claude.

Users can choose from 12 pre-built agents – for common use cases such as IT ticket deflection – or build their own, then assign them to project teams and immediately provide a historical record of which tasks have been completed and which have not yet been resolved. Agents also have access to third-party resources such as Microsoft 365 or Google Drive.

“When that agent is created, it doesn’t work for one person, it presents itself as a teammate and it gets all the same sharing permissions, that’s inherited,” Bose explained. Everything anyone does — including humans and AI — is documented to allow for “ease of explanation” and a “clear and reliable system.”

But like human workers, AI agents are monitored: Critically, workflows include checkpoints, where people can give feedback and ask the agent to tweak certain elements of a project or adjust research plans. It’s documented in what Bose calls a “human-readable way.”

Also important, the UI provides instructions and knowledge about the agent’s behavior, and approved admins can stop, edit and redirect API models if they take actions based on conflicting directions or start acting “in an unusual way.”

“The person with the right to edit can remove the conflicting items and bring them back to the correct behavior,” said Bose. “We lean toward a common human-understandable pattern of interaction.”

Overcoming the challenges of consent, integration

But since AI agents are still new, there are many challenges around security, accessibility and compatibility.

Asana users, for example, need to go through an OAuth flow and grant Claude access to Asana through their MCP and other public APIs. But letting all workers know that that integration exists — and more importantly, which OAuth grants are OK and should be avoided — can be a tall order.

Some of the challenges around direct OAuth grants between applications can be centralized identity providers, said Bose, or a centralized list of approved business AI agents with their skills, “almost like an active directory or universal directory of agents.”

Right now, however, beyond what Asana does, there is no standard protocol around shared knowledge and memory, Bose said. His team has gotten “a lot of interesting inbound inquiries” from partners who want their agents to operate Asana’s work graph and benefit from shared work.

“But since the protocol or standard doesn’t exist, now it has to be a routine conversation,” Bose said.

Finally, there are three questions that the CPO calls “very interesting” in AI orchestration today:

• How do you create, manage and secure an authoritative list of known approved AI agents?

• How can you do app-to-app integrations as an IT team without potentially configuring dangerous or malicious agents?

• Agent-to-agent interactions are now single-player only. Clouds can be independently connected to Asana or Figma or Slack. How can we finally reach a unified, multi-player outcome?

Further adoption of the modern context protocol (MCP) – the open standard introduced by Anthropic that connects AI agents to external systems in one action, instead of custom integrations for each individual pairing – is promising, he said, and its widespread adoption could open up new and exciting use cases.

However, “I think there probably isn’t a fundamental silver bullet out there right now,” Bose said.

Source link