SoftTech AI Registry
Asana · Asana AI

AI Teammates

Configurable AI agents created inside an Asana domain that can be assigned tasks, @mentioned, triggered by rules, forms or recurring tasks, and that read and persistently change Asana work (tasks, subtasks, comments, custom fields, projects, sections, pages, tables, capacity plans) within the access explicitly granted to them, using configured skills, automatically accumulated task-scoped memories and user-authorised connected applications.

Recorded characteristics

Function
Asana documents AI Teammates as AI-powered agents built into Asana that multiple colleagues can assign work to, ask questions of, and review output from. An AI Teammate is configured with a name, a human-facing description, behaviour guidance (instructions on how it speaks, reasons and delivers work), an Access configuration (specific Asana projects and teams, selected attachments, and connected apps), Memory, and Sharing settings. Documented triggering paths: assigning a task to the AI Teammate; @mentioning it in a task comment; updates to objects it already collaborates on; Asana rules; form submissions where it is the default assignee; recurring tasks; and self-scheduled follow-ups it sets during an execution. Users can interrupt an execution in the task thread with guidance, and can inspect an execution through "View activity". Skills are packaged units of expertise (guidance, trigger conditions and optional uploaded attachments) added to a Teammate from the Asana Skills Library, authored from blank, or imported from a Markdown file. Skills load selectively: a skill's guidance and attachments enter the working context only when that skill activates. Library skills are copied onto a Teammate, so edits do not propagate between Teammates. AI Teammates are created from a gallery (+ Create > AI teammate, or Agents > AI Teammates), from templates, from an "Analyze project" recommendation flow, or as a custom Teammate. Distinctness from the Registry's existing Asana Smart Workflows record: Asana documents AI Studio as the surface for building rule-based automated workflows for high-volume repeatable tasks, while AI Teammates handle collaborative work requiring context, judgment and back-and-forth interaction. Asana describes them as complementary, with AI Studio routing structured work in and an AI Teammate then working inside those tasks. A rule may trigger a Teammate, but the Teammate is a separate configured agent, not a rule action.
Data access
Established by Asana documentation: - On creation, an AI Teammate can see public tasks and projects in the domain, like a new user. It must be explicitly added as a member or collaborator on private Work Graph objects (tasks, projects, teams) to see or search them. If added to a team, it inherits that team's access levels. - Within its access, a Teammate can search tasks, subtasks, projects and project updates, milestones, portfolios, messages, comments, attachments, custom fields, status updates, goals and goal updates, and Reporting objects; public workspace data is searchable by default. - A Teammate's response is limited to the intersection of its own access and the triggering user's access: it cannot reveal information the triggering user does not already have permission to view. - Authorised members may upload key resources (playbooks, policies, instructions) for ongoing reference; skills may carry their own uploaded attachments. - External content (Google Drive, OneDrive/SharePoint, and other connected apps) is reached using the triggering user's own access tokens and connected account. - Memories are additional task-scoped context (see memory controls below). - Asana states an AI Teammate cannot elevate its own permissions and cannot bypass permissions. Not established: any automatic organisation-wide access beyond domain-public data, and access to private work that the Teammate has not been explicitly added to.
Actions
Can take actions
External actions
Conditional
Human confirmation
Conditional
Permission basis
Mixed
Administrative control
Documented controls: - Licensing and enablement: AI Teammates is a paid add-on assigned per user by domain, super or division admins in the admin console (individually or by CSV bulk import). Separately, a user's ability to consume AI requests is toggled per user under Users > Licenses > AI spending; disabling it removes that user's ability to use AI Teammates. - Usage governance: admins view requests used versus available, number of Teammate creators, number of Teammates created, members using Teammates, monthly usage and total completed requests in the Billing tab. - Teammate-level configuration: the creator sets access (projects, teams, attachments, apps), behaviour guidance, and sharing. Sharing grants AI Teammate admin (change settings, modify, delete), Editor (modify and apply to work) or User (find and apply to work). Only invited members can use a Teammate. - Skills: editors and admins add, edit, author and attach files to skills. - Apps: a Teammate's admin or editor adds an app to the Teammate; each individual then connects their own account. Per user and per app, actions can be set to Always allowed, Needs approval, or Not allowed, separately for Dash and for all AI Teammates. Read-only actions default to Always allowed; write/delete actions default to Needs approval. - Domain app administration: by default all apps are allowed; admins may use Asana App management to block apps or allow only approved apps, review connected apps and export app activity. - Memory: authorised members view memories on the Teammate's profile page and can delete them; memories are not manually created. - Execution oversight: View activity, mid-run interruption via comment, and comment threads visible to everyone with task access. Administrator configuration is distinct from runtime authority: admins license and restrict, while each run resolves against the Teammate's own object access and the triggering user's access.
Default state
Not established
Availability
Asana's current Help Center articles present AI Teammates as generally available functionality on Starter, Advanced, Enterprise and Enterprise+ plans with the AI Teammates add-on. Asana's own capability tables state what AI Teammates "can and can't do today", indicating an actively changing feature surface; one approval behaviour (adding a task to a project) is documented as temporarily bypassed by an internal flag. No beta or preview designation is applied to AI Teammates as a whole in the articles reviewed. Accessed 21 September 2026.
Licensing
Documented as an add-on license that can be added to any Starter, Advanced, Enterprise or Enterprise+ user license. Purchase is through an Asana sales representative. After purchase, an admin must assign the add-on license to a user in the admin console; licenses may be assigned individually or in bulk by CSV. Use also depends on the per-user AI request (AI spending) permission. Asana Skills are documented at feature tier rather than add-on tier in the Skills article. No pricing amounts are recorded here.
External model or provider
Not established. The Asana AI Teammates documentation reviewed does not identify the underlying models or model providers used by AI Teammates. Asana publishes a separate selectable-model list for AI Studio; that list is not stated to apply to AI Teammates and has not been carried into this record.
Limitations and uncertainty
ESTABLISHED - AI Teammates persistently change Asana work: create, edit and complete tasks and subtasks; set assignee, due date, description and custom field values; post, edit and delete their own comments; create projects; create and reorder sections; bulk update tasks; create dependencies, milestones, Asana Pages, and Asana Tables records; create capacity plans, workloads and allocations; merge duplicate tasks; attach files. - Explicit negatives in Asana's table: cannot create goals, create custom fields, remove a task from a project, send messages or status updates, create dashboards or charts, generate images or PDFs, rename or delete sections, update start date/task type/approval status on existing tasks, generate Google Slides, or bypass permissions. - External write is real but connection-dependent and per-user authorised (Gmail, Google Calendar, Outlook Mail and Calendar, Slack, HubSpot, Canva, Microsoft Teams, Google Drive documents and Sheets, SharePoint/OneDrive files, custom MCP apps), which is why external action is recorded as conditional rather than yes. - Human confirmation is conditional, not required: many Asana-side operations run with no approval, while specific operations are gated. Removing task collaborators and removing project members always require approval; assigning a task, creating tasks with assignees, deleting a task the Teammate did not create, adding collaborators, and adding members to a private project require approval only when the operation would grant someone new access. For connected apps, write/delete actions default to Needs approval but each user may set any action to Always allowed. Comment-based review of a draft after the fact is not treated as confirmation before action. CONFIGURATION-DEPENDENT / NOT ESTABLISHED - Which skills a Teammate holds, and therefore which behaviours it exhibits, varies per Teammate; not every Teammate has every available skill. - Which external systems a Teammate can read or write depends on apps added to that Teammate and on each user's own connected account and per-action settings. - Capability-level default operational state is not established: licensing, add-on assignment and the existence of per-action approval defaults do not establish whether AI Teammates operate by default in an Asana domain. - Models and providers for AI Teammates are not disclosed. - Runtime identity: Asana describes an AI Teammate as an addressable member-like object with its own object access, but also states that requests resolve against the triggering user's access and that connected-app actions run on the triggering user's credentials and appear as though that user performed them. This record therefore does not classify a distinct non-human security principal. - The temporary flag bypassing approval for adding a task to a project is documented as temporary and may change.

Evidence