Why Project Manager Performance Must Shift From AI Activity to OKRs

Performance review season often exposes a weak point in project management: teams measure what is easy to count, then hope it reflects value.
That gap is getting wider with Artificial Intelligence. AI can draft updates, summarise risks, compare delivery options, prepare first-pass business cases, and automate routine coordination work. A Project Manager can look more productive than ever while the project still fails to deliver anything meaningful.
That creates a hard question for leaders: are Project Managers being measured on the work that will matter in 2027?
Traditional scorecards were already imperfect. In the AI era, they can be actively misleading. Tool usage, output volume, and automation counts may show motion, but they do not prove better outcomes. For organizations that want performance reviews to drive real progress, the better route is clear: move from activity metrics to Objectives and Key Results.
At OKR Asia, we see OKRs as the most practical way to connect AI-enabled project management with measurable business and human value.

AI has made old PM scorecards less reliable
For years, Project Managers have often been assessed through familiar measures:
Whether milestones were met
Whether budgets stayed within range
Whether risks were logged
Whether stakeholders received updates
Whether team capacity looked well managed
These measures still matter. A project that ignores time, cost, scope, and risk is not well managed.
The problem is that AI makes many traditional measures easier to satisfy without improving the result. A PM can create cleaner reports in less time. They can generate more options, more meeting notes, more risk statements, and more polished plans. They can respond faster and look more organized.
Yet none of that proves the project is producing value.
This is where performance management can drift into metric theatre. People learn what the scorecard rewards, then shape their behavior around it. If the scorecard rewards AI tool usage, people use tools. If it rewards the number of automations built, people build automations. If it rewards faster reporting, people report faster.
The team may still be solving the wrong problem.
By 2027, the difference between average and excellent Project Managers will not be who can use AI the most. It will be who can use AI with judgment, context, ethics, and focus. That needs a different measurement system.
The first wave of AI metrics has been rocky
Many organizations are in their first or second review cycle with AI-related performance measures. It is understandable that early attempts feel uneven. Everyone is learning. The tools are changing fast. Policies are still catching up.
Still, three common phases are already becoming visible.
Phase | What organizations measured | Why it fell short |
Activity-based measurement | Logins, prompt volume, tokens spent, frequency of AI use | Easy to count and easy to game. It says little about quality or value. |
Output-based measurement | Faster delivery of documents, more reports, increased team capacity, time spent correcting AI work | Better than activity counts, but often rewards doing the same work faster. |
Builder-based measurement | AI agents deployed, use cases approved, AI business cases submitted | Shows initiative, but does not prove that the work changed outcomes. |
The lesson is not that these measures are useless. Some can help during early adoption.
For example, tool usage data can reveal whether people have access, confidence, and basic training. Correction time can show where AI output creates hidden rework. The number of approved use cases can show whether a team is experimenting.
These are useful management signals. They are poor performance goals.
Once a metric becomes a formal review target, people treat it differently. A PM who knows they are being assessed on AI usage may use AI even when a direct conversation, a simple checklist, or their own judgment would be better. A team measured on the number of AI agents deployed may create tools that add maintenance burden but little value.
That is why AI adoption metrics should sit below the main performance frame. They can support learning. They should not define success.
Value is the missing link in Project Manager Performance
Project management expert Stephen Devaux has long argued for judging projects through the value they create, not only through schedule and cost control. That principle becomes more important as AI enters the workplace.
AI does not make a project valuable on its own. It only has value when it helps the organization achieve a valued result.
These business results might be:
Higher revenue from a faster product launch
Better margin through reduced rework
Lower operational risk
Faster service recovery for customers
Safer delivery in high-risk environments
More time for skilled staff to focus on human judgment
Better access to essential services
In public or social projects, outcomes such as lives protected, time saved, or trust improved
This is the heart of the shift. Project Manager performance must be tied to the value of the project, not the visible volume of project management work.
A PM using AI to produce a 40-page risk report has not necessarily helped. A PM using AI to find three overlooked delivery risks, bring the right people into the decision, and prevent avoidable rework has created value.
The difference is not the tool. The difference is the outcome.
AI should make Project Managers more valuable to the organisation, not merely more active inside the system.
OKRs give AI-era project work a better frame
OKRs work because they separate ambition from evidence.
An Objective describes the meaningful outcome the project team is trying to achieve. It should be clear, human, and important.
Key Results define how the team will know progress is real. They should be specific, measurable, and focused on results rather than tasks.
That structure fits AI-enabled project management because it keeps the conversation on value.
A weak AI metric might say:
`Use AI tools in 80% of project planning activities.`
A stronger OKR might say:
Objective
Improve delivery confidence for priority customer projects while reducing preventable rework.
Key Results
Reduce preventable rework from approved requirement changes by 20%
Identify and resolve critical delivery risks at least two weeks earlier than the previous project baseline
Improve stakeholder decision turnaround from five working days to two working days
Maintain agreed data protection and review standards across all AI-assisted project artifacts
The second version allows AI to help, but AI is not the goal. The goal is better delivery confidence, less waste, faster decisions, and safe tool use.
This is the real benefit of OKRs. They reduce the temptation to reward visible busyness. They ask whether the project created evidence of progress.
What good PM OKRs look like in practice
Good PM OKRs in the AI era should connect three things:
Business value
Delivery quality
Human judgement
If an OKR only measures business value, it may encourage short cuts. If it only measures delivery quality, it may miss whether the project mattered. If it only measures human behavior, it may become soft and hard to assess.
The best set balances all three.
Here are a few illustrative examples.
PM situation | Weak AI metric | Better OKR approach |
A PM uses AI for planning | Generate all project plans using AI | Improve plan quality by reducing missed dependencies and late scope changes |
A PM uses AI for risk analysis | Produce weekly AI-generated risk reports | Reduce unplanned escalations by identifying high-impact risks earlier |
A PM uses AI for stakeholder updates | Send automated updates every Friday | Improve decision speed and reduce confusion among key stakeholders |
A PM builds an AI agent | Deploy three project support agents | Reduce manual coordination time while maintaining accuracy and trust |
A PM uses AI for lessons learned | Summarise all retrospectives with AI | Increase reuse of lessons across future projects and cut repeat issues |
Notice the pattern. AI may be part of the method, but the Key Result measures whether something got better.
A useful PM OKR might read:
Objective
Increase the success rate of complex delivery programs through earlier risk detection and better decisions.
Key Results
Cut late-stage dependency surprises by 30%
Reduce average time from issue identification to decision from six days to three days
Improve sponsor confidence score from the project health check
Document all AI-assisted recommendations with source context and human review
This kind of OKR rewards strong project management. It also rewards responsible AI use. The PM still needs to interpret signals, manage trade-offs, ask better questions, and guide people through uncertainty.
AI supports the work. The PM remains accountable for judgment.

How to shift performance reviews towards OKRs
A move to OKRs does not require tearing up every existing performance process. It does require clearer intent.
The goal is to make the review conversation less about activity and more about evidence of value. For Project Managers, that means changing what gets discussed before, during, and after the project.
Start with the value question
Every project should begin with a direct question:
What value will this project create, and how will we know?
This sounds simple, but many project measures skip it. They start with delivery dates, budgets, or workstreams. Those are needed, but they are not enough.
A value statement should name the desired change. For example:
Reduce customer onboarding delays
Improve compliance readiness
Lower support demand after launch
Increase adoption of a new internal process
Protect service continuity during a platform migration
Once the value is clear, the team can decide where AI might help. AI might help with pattern detection, scenario planning, content drafting, knowledge retrieval, or process automation. The tool choice comes after the outcome, not before it.
Keep AI metrics as health indicators
AI adoption still needs measurement. Leaders need to know whether teams have access, capability, and safe practices.
Use AI metrics as health indicators, such as:
Percentage of PMs trained in approved AI tools
Number of AI-assisted artefacts reviewed for accuracy
Time spent correcting poor AI outputs
Common use cases where AI saves effort
Common use cases where AI creates risk
These measures can guide coaching and governance. They should not become the main definition of PM performance.
A PM who uses AI sparingly but creates excellent project outcomes should not be penalised. A PM who uses AI constantly but produces poor decisions should not be rewarded.
Review Key Results through evidence
Performance reviews work better when they examine evidence, not impressions.
For PM OKRs, evidence might include:
Decision logs showing faster resolution
Before-and-after rework patterns
Risk registers showing earlier detection
Stakeholder feedback on clarity and confidence
Delivery health checks
Financial or operational impact where available
Lessons reused in later projects
This approach also reduces bias. Instead of asking whether a PM looked busy or sounded confident, the review asks whether the project changed in the intended way.
Make responsible AI use part of quality
AI introduces new risks. It can produce inaccurate summaries, hide assumptions, reuse sensitive information in unsafe ways, or create false confidence.
A good PM performance framework should include responsible AI practice as part of delivery quality. That does not mean adding a long compliance checklist to every review. It means expecting Project Managers to show care in how AI-assisted work was produced.
Good Key Results can include safeguards, such as:
Critical AI-assisted recommendations receive human review
Sensitive data is handled within approved tools only
Project assumptions are documented and checked
AI outputs are traceable enough for stakeholders to challenge
This keeps the standard practical. Responsible use is not separate from delivery. It is part of delivery.
The human work becomes more visible, not less
One fear around AI is that it will reduce the role of the Project Manager. In reality, it should expose which parts of the role were merely administrative and which parts create real value.
AI can help draft a plan. It cannot own trust.
AI can summarise stakeholder feedback. It cannot judge the politics, history, incentives, and hidden concerns behind that feedback.
AI can suggest risks. It cannot decide which risk deserves attention when the team is tired, the deadline is close, and trade-offs are painful.
The Project Managers who stand out in the next few years will be strong in areas AI cannot replace:
Framing the right problem
Building shared understanding
Challenging weak assumptions
Creating decision clarity
Managing tension without blame
Protecting focus when everything feels urgent
Connecting project work to organizational value
Knowing when an AI answer is plausible but wrong
OKRs make this human work easier to recognize. They show whether the PM improved the conditions for success, not just whether they maintained project paperwork.
This matters for performance reviews. If reviews focus only on task output, AI will flatten the differences between Project Managers. If reviews focus on outcomes, judgment and leadership become clearer.
A practical OKR model for PM performance in 2027
A strong PM performance model for the AI era should have three layers.
Layer one is project value.
This is the main OKR layer. It answers what the project is meant to improve.
Example:
Objective
Deliver a smoother customer onboarding journey for enterprise clients.
Key Results
Reduce average onboarding delay caused by internal hand-offs
Improve first-month customer satisfaction feedback
Cut repeat support requests linked to unclear setup steps
Maintain agreed project scope without late surprise changes
Layer two is delivery system improvement.
This layer measures how the PM improves the way work gets done.
Example:
Objective
Improve delivery predictability across the project team.
Key Results
Reduce unresolved cross-team dependencies
Increase early identification of schedule risks
Improve decision response time from project sponsors
Reuse lessons from previous comparable projects
Layer three is responsible AI contribution.
This layer tracks whether AI helped safely and meaningfully.
Example:
Objective
Use AI to improve project clarity without weakening trust or control.
Key Results
Reduce manual reporting time while keeping sponsor satisfaction steady or higher
Confirm all AI-generated project summaries against source records
Identify recurring risk themes across project data and act on the top issues
Keep sensitive project information within approved tools and processes
This model avoids two extremes. It does not ignore AI. It also does not worship AI activity. It treats AI as part of the delivery system, measured by contribution to value.
The real test is whether behavior changes
A performance framework only matters if it changes decisions.
If a PM knows they will be assessed on AI usage, they will use AI more. If they know they will be assessed on project value, they will ask better questions:
What outcome are we actually trying to create?
Which measure proves progress?
Where is AI useful, and where might it mislead us?
What decision needs human judgment?
What evidence will we bring to the review?
These are the questions that belong in performance management now.
The future of Project Manager performance is not more dashboards, more automation counts, or more polished reports. It is clearer alignment between work and value. OKRs offer a simple, disciplined way to make that shift.
As AI takes over more routine project activity, organizations need PMs who can turn speed into better outcomes. Measure that, and the review season becomes more than an annual exercise. It becomes a signal for the kind of project leadership the future will reward.
Contact us at transform@asiapmo.com or via the contact form link to learn how your Project Managers can focus on business outcomes, not just project deliverables.


Comments