A project review pack arrives on a Monday morning. The commentary is clear, the risks have been summarised, the actions are neatly grouped and the recommendation at the end sounds entirely reasonable.
It may have taken someone several hours to produce. Or it may have taken an AI tool a few minutes.
That difference is becoming less important, what matters is whether the people reading it understand how the output was produced, whether they have tested what it says and whether they are prepared to make a decision based on it.
AI can help PMOs work through large amounts of information, spot patterns, prepare summaries and explore different options. It can also produce something which sounds confident while being incomplete, based on the wrong assumptions or simply incorrect.
The challenge for PMO professionals is therefore moving beyond knowing how to use an AI tool – we also need to know how to judge what it gives us and how to govern the decisions it influences.
The Skills England annual skills report 2026 describes practical AI literacy as the ability to use, verify and safely integrate AI tools. It also highlights judgement, problem-solving, digital fluency and responsible AI capabilities as skills employers increasingly want.
That gives us three useful areas to think about in a PMO context.
1. Using AI effectively
Using AI effectively does not begin with finding as many tasks as possible to automate, it begins with understanding the work and deciding where AI can genuinely help.
There are plenty of useful applications in a PMO. AI might help to summarise project updates, compare information across reports, identify recurring risks, prepare the first draft of a governance paper, turn meeting notes into actions or explore the effect of a change in priorities.
The value comes from reducing the effort involved in processing and organising information, which gives PMO professionals more time to interpret what it means, speak to the people involved and focus on the decisions which need attention. Imagine, for example, that a portfolio office receives 25 project updates at the end of the month. An AI tool could review them and identify common themes such as resource pressure, delayed decisions or supplier issues.
That could be very useful, but only if the task has been set up properly. What information has the tool been given? Are all 25 updates using the same definitions? Is the tool being asked to summarise, to analyse or to recommend? What information should not be entered because it is sensitive or confidential?
A vague request will normally produce a vague or overly general response. A more useful request includes the context, the task, the information to use, the intended audience and the form the output should take.
A useful question for the PMO is: What part of this work are we asking AI to do, and what still requires human judgement?
This distinction matters because AI is well suited to some activities and much less suited to others. It can organise information quickly; it does not automatically understand the politics surrounding a delayed decision, the history behind a stakeholder relationship or the reason a project manager has chosen cautious wording in a status update.
Effective use therefore involves more than prompting. It includes choosing the right task, providing suitable information, protecting data, understanding the limitations of the tool and keeping a person involved where judgement or accountability is required.
2. Critically evaluating AI output
The second skill begins when the AI has finished its work.
One of the difficulties with AI-generated output is that it can be very easy to read. The language is usually tidy, the structure looks sensible and the conclusion may sound assured. That presentation can encourage us to accept the answer before we have properly considered whether it is right.
PMO professionals already use critical evaluation in their day-to-day work. We question a forecast which does not appear to match progress or look for the assumption hidden behind a completion date. We might compare a green status with the issues being discussed in meetings or ask why two reports contain different figures for the same thing.
The same habits are needed when working with AI output.
Suppose an AI tool reviews a programme risk register and concludes that supplier dependency is the greatest threat to delivery. Before that conclusion is included in a board paper, the PMO needs to understand how the tool reached it.
Did it focus on the number of supplier-related risks, their scores, the strength of the wording or how often suppliers were mentioned? Did it overlook a single high-impact regulatory risk because that risk appeared only once? Was the register current and complete? Did the tool introduce information which was not in the source material?
The answer may still be useful, but the polished wording is not evidence that the analysis is sound.
The question is not simply, Does this look right? It is, What would give us enough confidence to rely on it?
That might mean checking statements against the original source, testing calculations, asking for the reasoning or evidence behind a conclusion, comparing the response with another source and involving someone who understands the subject. It also means looking for what is missing. AI may summarise the information it has been given very well while saying nothing about information which was absent, outdated or recorded in a different system.
This is particularly important when AI is used to produce insights or recommendations rather than straightforward summaries. The further the tool moves from organising information towards interpreting it, the more scrutiny the output requires.
Critical evaluation is not about distrusting everything AI produces. It is about applying an appropriate level of challenge before the output becomes part of a report, conversation or decision.
3. Governing AI-enabled project decisions
The third area is broader than the way an individual uses AI. It’s about how the organisation makes decisions when AI has played a part in producing the information, analysis or recommendation.
A decision does not become an AI decision simply because an AI tool helped prepare the paper. Equally, calling the final choice a human decision does not tell us whether the person making it had enough information to challenge what the tool produced.
Good governance should make the role of AI visible and proportionate to the significance of the decision. For a low-risk task, such as creating an initial summary of meeting notes, a quick human review may be enough. If AI is being used to recommend which projects should lose funding, predict whether a programme will succeed or identify which suppliers should receive closer scrutiny, stronger controls will be needed.
The PMO can help by asking practical questions. Where has AI been used? Which data and instructions shaped the output? Who checked it? What assumptions or limitations should the decision-maker understand? Who is accountable for the recommendation and the final decision? Can the reasoning be revisited later if the decision is challenged?
These questions are familiar territory for PMOs because they concern decision rights, assurance, evidence, risk, traceability and accountability.
A useful governance test is: If this decision is questioned in six months, could we explain what information was used, how AI contributed and who exercised judgement?
If the answer is no, the governance is probably not strong enough for the significance of the decision. This does not mean surrounding every use of AI with a new committee and a large amount of paperwork. Governance should be proportionate. The aim is to make responsible use easier, provide clear boundaries and ensure that higher-impact uses receive the attention they deserve.
The PMO may not own the organisation’s AI policy, but it is well placed to translate that policy into the reality of projects, programmes and portfolios. It can help define where AI use should be declared, what evidence should accompany AI-supported recommendations and when additional review or assurance is required.
The three skills work together
These three areas should not be treated as separate stages which belong to different people. Using AI effectively without critical evaluation can produce poor decisions more quickly. Critical evaluation without governance may identify concerns which are never recorded or acted upon. Governance without practical capability can result in rules which look reassuring but do little to improve how AI is actually used.
The strongest approach combines all three:
- Use it well by choosing suitable tasks, providing the right context and understanding the limits of the tool.
- Evaluate it carefully by checking the evidence, assumptions, omissions and conclusions.
- Govern it proportionately by making the use of AI visible and keeping accountability for decisions clear.
For PMO professionals, this is less of a departure from existing practice than it might first appear – we already work with imperfect information, challenge assumptions, maintain governance and help people make decisions under uncertainty.
AI changes the speed, scale and apparent confidence with which information can be produced, it does not remove the need for those PMO capabilities, if anything it makes them more important.
A role for the PMO
There is a lot of attention on which AI tools people should learn and which tasks they can automate, the more important question for the PMO is how AI can be used to improve project decision-making without weakening the scrutiny, transparency and accountability on which good decisions depend.
That is where PMO professionals have something valuable to contribute, we understand how information moves through delivery environments, how governance operates, where assumptions sit and what happens when a decision is made using information which has not been properly tested. The opportunity is not simply to become confident users of AI, it is to help the organisation become a more capable, critical and responsible user of it.
Perhaps the simplest place to begin is with three questions:
- Are we using AI for the right part of the work?
- Have we challenged the output enough to rely on it?
- Can we explain and stand behind the decision which follows?
If we cannot answer all three, we are probably not ready to act on what the AI has told us.
The House of PMO Insight Conference 2026
These themes sit at the heart of the PMO Insight Conference in Edinburgh on 16th and 17th November 2026. Across the two days, we will explore how PMOs can work with data, insight and AI while strengthening the human judgement and governance needed to turn information into better decisions.
Sources and further reading
Skills England, Annual Skills Report 2026: https://www.gov.uk/government/publications/skills-england-annual-skills-report-and-sectoral-skills-needs-assessments-2026/skills-england-annual-skills-report-2026
Skills England, AI foundation skills for work benchmark: https://www.gov.uk/government/publications/ai-foundation-skills-for-work-benchmark
House of PMO, PMO Dashboards: 5 Questions to Ask Before Making Decisions: https://houseofpmo.com/blog/2026/08/13/pmo-dashboards-5-questions-to-ask-before-making-decisions/
PMO Learning at the House of PMO: Practical AI Skills for the PMO: https://training.houseofpmo.com/specialist-pmo-courses/practical-ai-skills-for-the-pmo/
House of PMO: Inside PMO Report – AI and the PMO: https://houseofpmo.com/library/inside-pmo-ai-pmo-threat-or-opportunity/
