Day 14: Run a Project Health Check
By 21 Days of AI · Last updated: July 4, 2026
The Concept
A project can look busy and still be unhealthy. Tasks may be moving while the outcome becomes less relevant, risks remain unowned, or stakeholders lose confidence. A health check creates a deliberate pause to examine the whole system rather than relying on the loudest current problem.
Use evidence, not mood
Health is not the project manager's feeling about the project. It is a view built from evidence: milestone movement, unresolved decisions, quality signals, capacity, risk exposure, and stakeholder confidence. Sentiment matters too, but it should be treated as an input to investigate, not a substitute for facts.
Look for causes
AI can help compare symptoms and possible causes. Ask it to test several explanations rather than accepting the first one. A late task may be caused by poor estimation, an unavailable dependency, unclear acceptance criteria, or competing priority. The response should change depending on the cause.
Example: read the whole picture
Imagine a project where the schedule is amber because the last two milestones moved. The team is working long hours, so a quick diagnosis might blame delivery speed. A wider health check could show that scope is still changing, the quality review is happening late, and the sponsor has not confirmed which outcome matters most. The late milestones are symptoms of several connected conditions, not evidence that the team simply needs to work harder.
Write the evidence beside each health area. For schedule, record the original and current dates. For scope, list decisions that changed the work. For quality, note defects, rework, or skipped checks. For capacity, compare committed work with available time. For confidence, capture what stakeholders say they understand and what they still question.
This makes the conversation less personal. Instead of asking why the team is behind, ask which condition is creating the most avoidable delay and who can change it.
Compare leading and lagging signals
Lagging signals tell you what has already happened: a missed milestone, an increase in defects, or an overspent budget. Leading signals suggest what may happen next: unanswered decisions, growing work in progress, repeated clarification requests, or a dependency with no confirmed date. A strong health check includes both. AI can help generate leading indicators, but select only signals the team can actually observe.
Make the check a conversation
Share observations before ratings. Ask the team and stakeholders what they see differently and where evidence is incomplete. If people disagree, record the disagreement as a question to resolve. A health check should create better visibility, not produce a score that ends the discussion.
A final check before sharing
For every concern, identify the evidence, likely cause, owner, and next action. If the action is only “monitor,” define what will be monitored and when the team will decide whether to intervene. Health improves when the check changes behaviour.
Choose the right response
Not every amber signal needs an immediate recovery plan. Some signals need better evidence first. Others need a decision, a change to scope, a capacity conversation, or an escalation. Name the response type so the project does not spend energy treating every concern as the same kind of problem.
For example, weak stakeholder confidence may call for a clearer demonstration rather than a new plan. A quality warning may need an acceptance decision. A capacity gap may require work to be removed, not a request for the team to work faster. The health check is useful when it points toward the right conversation.
Avoid false precision
Traffic-light ratings are a communication aid, not a measurement system. Two people may use amber differently. Define the evidence behind the rating and include a short explanation. If the team cannot agree on a colour, that disagreement is worth recording because it may reveal a difference in expectations or risk appetite.
Repeat the health check after a meaningful intervention. Compare what changed, what did not, and whether the original diagnosis was correct. This creates a learning loop and helps the team improve its ability to see problems earlier.
Use the result to choose the next management action, then return to the project team with the same evidence. A health check should not become a private report that surprises the people responsible for recovery.
Invite the relevant owners to validate the findings before the check becomes a formal status statement. Their corrections improve the evidence and create earlier ownership of the response.
Use this today
That shared review also makes the recovery plan easier to support.
It turns a rating into a shared next step.
Revisit the rating after the action so the team can see whether visibility and control improved.
Create a one-page health check. For every amber or red area, write the evidence, the owner, and the next conversation. Keep the check useful by limiting actions to what the team can actually do.
Remember this
- A health check is a learning moment, not a performance ritual.
- Symptoms deserve investigation before they receive solutions.
- Weak evidence is itself a project risk.
Prompt of the day
Copy this into your AI tool and replace any bracketed placeholders.
Prompt
Help me run a practical project health check. Project outcome: [OUTCOME]. Current milestones: [MILESTONES]. Progress evidence: [EVIDENCE]. Budget or capacity position: [POSITION]. Top risks and issues: [RISKS]. Stakeholder feedback: [FEEDBACK]. Team sentiment: [SENTIMENT]. Create a health check across outcome, scope, schedule, dependencies, risk, quality, capacity, stakeholder confidence, and decision-making. For each area, give: evidence to review, signs of health, warning signs, one question to ask, and one action if attention is needed. Do not assign a health rating without evidence.
Your 15-minute task
Run the health check on a real project. Choose the two areas where evidence is weakest and schedule one action to improve visibility. Share the findings as questions and observations, not as a surprise verdict.
Expected win
An evidence-based view of project health that reveals where attention is needed before the project becomes difficult to recover.
Power user tip
Ask AI to distinguish a symptom from a cause. A missed milestone may be a symptom of unclear scope, missing capacity, or a decision that has not been made.
Finished today?
Mark this lesson done on this device. No account is required, and you can continue straight to the next day.
Want useful AI learning updates in your inbox?
Get practical AI notes, course updates, and new resources by email. You can keep reading for free now, with no account required.
Get updates