Can AI agents run a sprint? Six project management tools compared
Compare the documented agent capabilities of Linear, ClickUp, Notion, Asana, monday, and Stellary across one reproducible sprint benchmark.
11 min read
Programme steering is about propagation
A project has a goal, a team and a date. A programme has several of each, linked. Each project is already steered on its own; the programme is steered between them. The hard part is not the volume of information but its transversality: a two-week slip on the authentication workstream is not a local problem when three other workstreams wait for its API. Task tools show each project. They rarely show the propagation.
What an AI agent does well here
- Consolidates signals from every project — remaining work versus capacity, blockers, pending decisions, open risks — onto one scale.
- Spots slipping dependencies by comparing the expected date with the probable date, and says so when the gap appears, not at the next committee.
- Drafts the trade-offs. Reassign, postpone, cut scope, accept the delay: each option costs the other projects something. The agent writes the options with their known effects.
- Executes what was approved, under permissions: anything that touches another project becomes a proposal, never a direct change.
Where the agent stops
An agent that settles a programme’s trade-offs is a bad agent, even when it is right. The decision commits teams and budget and needs a named human owner. The agent knows the programme’s data, not the client who promised a date verbally or the team coming out of three hard months. And teams accept an agent preparing the committee; they do not accept one cutting their scope. In Stellary that boundary is a setting: a steering agent runs in supervised or approval mode, and cross-project actions are proposals.
The weekly loop
- Goals and margin per project — what each can lose before it endangers the programme.
- Consolidated signals — the portfolio view the programme lead opens on Monday.
- Gaps and options — for every signal off trajectory, the agent drafts the options.
- Decisions on record — date, owner, rationale, expected effect, triggered actions.
- Actions and follow-up — the agent executes what was approved and reports the effect the following week.
FAQ
What is AI-assisted programme steering?
Using AI agents to consolidate the signals of several linked projects — progress, dependencies, margin, risks — and to prepare the programme lead’s trade-offs. The agent reads, computes and proposes; the decision stays human, and every cross-project action goes through an approval.
Can an AI agent arbitrate between two projects?
It can prepare the arbitration: options, known costs, effects on the other projects. It should not settle it: the decision commits teams and budget and depends on context the board does not hold.
