It’s a Planv1.12.0

Before You Trust an AI Project Plan, Check Its Dependencies and Assumptions

A step-by-step review for AI-drafted project plans: check that scope, schedule, and budget agree, validate dependencies and capacity, tie risks to actions, and log assumptions before you commit to a date.

Andrii PoluosmakUpdated 6 min read

A plan can look ready for approval before it reflects the project’s scope, dependencies, people, risks, and constraints. Before you promise a date or assign work, compare the draft with the conditions your team faces. Use AI project planning to draft the plan, then have your team test its details.

PM World Journal’s October 2026 article “AI for Project Planning: From Plausible Artifacts to a Coherent Plan” describes how scope, schedule, cost, responsibilities, and risk can each look complete while failing to fit together. Check the links between these parts before you commit.

Start by separating a draft from a decision

Use AI to organize information and draft planning materials. PM World Journal’s “AI for Project Planning” describes uses such as drafting a work breakdown structure, calculating a critical path, estimating costs, assigning responsibilities, and identifying risks. Check the project context and inputs behind each output.

Have reviewers trace each planning assumption to its supporting evidence. Find the basis for each task duration, then verify the work sequence behind each predecessor. Ask resource owners to confirm that named people have the time and skills for their assignments. Connect each risk response to the work, cost, or people it requires.

The Project Management Institute’s 2023 article “Shaping the Future of Project Management With AI” says complex tasks call for more human intervention. It lists schedules and risk analyses as tasks AI can assist with, and recommends refining and testing those outputs. Have experienced reviewers focus on decisions that depend on project context, including dependencies, constraints, capacity, and uncertainty.

Check that scope, schedule, and budget agree

Work from the approved scope. Match each deliverable or work package to the schedule and estimate. Confirm that every promised result has the work, duration, and cost it needs. Check the schedule for approvals, procurement, testing, commissioning, transition, and closeout.

PM World Journal’s “AI for Project Planning” recommends checking completeness, boundaries, deliverables, traceability to requirements and assumptions, and whether teams can schedule, estimate, assign, and assess each work package for risk. Run these checks:

  • List scope items with no scheduled activity.
  • Flag scheduled activities with no clear scope or deliverable behind them.
  • Identify activities with no cost or stated cost basis.
  • Check any estimate based on a benchmark or percentage the team has not confirmed.
  • Track how a date change affects budget timing.

Recalculate schedule dates and float with a separate calculation. Check dependencies, calendars, and constraints against project conditions. PM World Journal’s article recommends these checks. Ask a scheduler or delivery lead to confirm the logic.

Validate dependencies, sequence, and capacity

Trace each dependency through the work. For each major task, compare its predecessor and successor with technical requirements, approvals, procurement lead times, access windows, and team handoffs. Ask the responsible specialist to confirm whether the planned start fits the dependency and project conditions. Check that the predecessor delivers what the next activity needs.

Check parallel tasks against the availability of teams, equipment, decisions, and information. Ask the people doing critical or near-critical work to confirm the sequence and calendar. If a dependency remains uncertain, log the assumption and name the evidence that would confirm or invalidate it.

Check capacity against the actual resource plan. PM World Journal’s “AI for Project Planning” notes that a general-purpose AI lacks an organization’s resource inventory, existing commitments, skills, and calendars unless someone supplies that information. Ask functional managers to confirm each person’s availability, capability, and authority. Flag overallocations and assignments the people named have not accepted.

Review RACI assignments with the people named in them. Confirm that major activities have one accountable role and that the person responsible has time and skills to do the work. Keep unconfirmed assignments off the commitment list until the team resolves them.

Connect risks to contingencies and actions

For each significant risk, review its cause, uncertain event, and consequence. Check the basis for its probability and impact ratings. Test whether the response can work and identify the risk that remains afterward.

Trace each response into the rest of the plan. Name its owner and the trigger that tells them to act. List any scheduled task, budget, contingency, or specialist the response needs. Add that work to the schedule and estimate. Update other affected plan components when a response changes a dependency or date.

PM World Journal’s “AI for Project Planning” advises teams to connect each selected response to an owner, due date, trigger, resources, residual exposure, and relevant schedule and budget components. For each response, name the risk owner, action trigger, authority, and funding.

Make assumptions and approvals visible

Build an assumption log. For each uncertain input, record its source, the person who can verify it, the evidence or decision needed, the due date, and what changes if the assumption proves false. Include items such as approval duration, supplier lead time, a person’s availability, or an unconfirmed technical interface.

Keep facts separate from estimates and decisions. Use a source document to verify requirements and past project data to support estimates. Ask the sponsor to approve trade-offs. Record the basis for each planning choice instead of treating an AI-generated explanation as the source of a project constraint.

Before you approve the plan, review how its parts connect. Ask the AI to flag missing links, duplicates, contradictions, and changes the team has not carried across the affected artifacts. Have the relevant owners resolve the findings. PM World Journal’s article warns that an early omission can carry through later outputs when each draft becomes the next prompt’s input. Check underlying assumptions against project records or their owners, even when the documents agree.

A practical review checklist

Review these points with the people accountable for the plan before treating it as a delivery commitment:

  1. Scope: Does work cover every commitment? Does each work package tie to an approved requirement or deliverable?
  2. Schedule: Has someone checked the dates and float with a separate calculation? Do dependencies, calendars, approvals, and constraints match the project?
  3. Budget: Does each major estimate have a stated basis? Do costs and timing align with the schedule?
  4. Capacity: Have resource owners confirmed availability, skills, authority, and workload?
  5. Risk: Do material risks have feasible responses, owners, triggers, time, and funding?
  6. Assumptions: Has someone named and assigned each unverified input and linked it to the decisions it affects?
  7. Approval: Has an authorized person accepted the integrated plan and any changes to scope, schedule, or cost?

In a controlled sprint-planning experiment at a mid-sized digital agency, researchers reported risk capture of 36.4% for AI-only planning and 86.7% for hybrid planning. The April 2026 arXiv paper describes the hybrid process as including human review of AI outputs and a structured session for risk identification and assumption documentation. Ask your team to name the risks and dependencies the generated draft may have missed.

Before you commit to your next AI-assisted plan, run this checklist against one real project. Ask the owners to verify dependencies and assumptions, record unresolved items, and approve the plan after they review its connected commitments.

Key takeaway

An AI-generated plan is a draft until the people who own the work confirm its dependencies, capacity, and assumptions. Approve it only when its parts agree and someone accountable accepts the remaining uncertainty.

Frequently asked questions

Can AI create a project schedule?

AI can draft schedule information. PM World Journal’s “AI for Project Planning” recommends checking calculations, durations, dependencies, and calendars against actual conditions before using the schedule.

What should I verify first?

Start with assumptions that could shift the delivery date or leave work without the people it needs: key dependencies, approvals, long-lead items, resource availability, and constraints. Assign an owner to each item the team cannot confirm yet.

When is an AI-assisted plan ready to approve?

Approve it after the scope, schedule, budget, resources, and risk responses agree, the team assigns owners and due dates to major assumptions, and an authorized decision-maker accepts the remaining uncertainty.

Andrii Poluosmak

Founder

Andrii founded It’s a Plan and leads its product and roadmap, from the open-source edition to Pro for teams. He works with the project’s contributors and the teams that run It’s a Plan.

All posts by Andrii Poluosmak

Put your next project on the board.

The open-source issue tracker you host yourself.