John McIntyre
Founder, HotPMO
Every month I trawl through the research papers, reports, and podcasts so you don’t have to. Here’s what caught my eye this month.
June was the month AI governance properly arrived in our profession, with PMI publishing a full standard for AI in project work. There was also a fine reflective piece on why contracts never quite control projects, a method paper from Cameroon with a lot to teach us about lessons learned, and two days of PMO Conference to digest.
An AI standard that doesn’t read like AI wrote it
The big one this month: on 9 June, PMI published The Standard for Artificial Intelligence in Portfolio, Program, and Project Management, which PMI describes as the first ANSI-approved AI standard for the project profession. It runs to nearly 300 pages, which makes me wonder how many people will drop the PDF into their AI chat tool and ask for a summary. I’m still not sure whether that defeats the object or uses it exactly as intended.
When this came up on the PMO HotHouse news slot, Eileen’s first question was how much AI had been used to write it. So I went looking for the usual tells. I found very little, and it doesn’t read like machine output, which counts for a lot given the subject matter.
The standard works from two angles: using AI within project delivery, and managing initiatives that deliver AI. It sets out eight guiding principles (because you can never have too many principles!):
- Strategic Value: start every AI initiative from the value it should deliver against organisational strategy; the technology is a means to that end.
- Risk: AI brings new threats, from algorithmic bias to opaque model outcomes, whose impact can reach well beyond a single project.
- Governance and Compliance: oversight, accountability and alignment built into every phase of the AI life cycle, rather than a rulebook written once.
- People and Culture: implementations succeed through the people involved and a collaborative culture, so invest there first.
- Ethics and Professional Responsibility: put an actionable ethics framework with clear accountability in place from the start.
- Stakeholder Engagement: plan AI adoption around what stakeholders expect from it, and manage the change that follows.
- Optimisation and Innovation: keep asking how your AI practice aligns with strategic goals and where it can improve efficiency and value delivery.
- Data Quality: outputs are only as trustworthy as the data underneath, so data stewardship comes before clever models.
The eight guiding principles of PMI (2026), ‘The Standard for Artificial Intelligence in Portfolio, Program, and Project Management’. The summaries are mine, and I’ve anglicised their ‘Optimization’.
Five performance domains sit alongside those, and the appendices map the AI life cycle onto familiar project life cycles instead of inventing a parallel universe. Human-in-the-loop oversight runs through the whole thing. PMI’s announcement says it addresses compliance with emerging requirements such as the EU AI Act and ISO 42001, though the standard itself stays principle-led, so don’t expect a clause-by-clause mapping.
Two things stood out for me. First, the strategic-value principle: any AI initiative should start from a hard look at the value it creates for the organisation. Hold that lens up to your own AI activity, because plenty of organisations are throwing licences around and seeing what sticks. Second, a warning of my own about the automation examples. Everyone is letting AI write their meeting minutes now, and I’m watching ownership of minutes and follow-through on actions decline as a result. We used to hold people to account against concise minutes. A transcript dumped into a channel does not do the same job.
Where AI does the work, and where people stay in charge. adapted from Figure 1-2 of PMI (2026), ‘The Standard for Artificial Intelligence in Portfolio, Program, and Project Management’.
And in case you’re wondering: the PMO is mentioned exactly twice in those pages. I searched. Once for AI-driven analytics in resource management and tracking, and once in a reviewer’s job title. Make of that what you will. The practical move is to map your existing assurance and governance against the eight principles and five domains, and decide where your human-in-the-loop escalation points sit in your gating, before AI governance fragments across legal, IT and risk.
What counts as normal clay?
In the Project Management Journal, Stewart Clegg has written Project Studies: Embeddedness in Field and Theory, a reflective piece built around a dispute he witnessed as a young labourer on a construction site in 1971. The contract said dig 600 millimetres into sandy stony clay. A borehole report said two metres. What followed was a long, expensive argument about what the contract meant by sandy stony clay, whether the ground was more muddy than stony, and which areas met the definition at all. The contract could not settle an argument about its own words.
Anyone who has been around projects for a while has lived some version of this. My view, for what it’s worth: contracts are your backstop. If you find yourself back at the contract, something has already failed in the communication around it. Clegg makes the same point at length. Documents plan for normal, and projects are open systems where deviation is the normal condition. Shared meaning and trust between the people involved do the work that paperwork alone cannot. He offers the Sydney 2000 Olympics as the counter-case, where alliance contracting deliberately changed the incentives so every side read the documents in the project’s interest.
One more habit worth borrowing from the paper: be sceptical of tidy explanations for why a project slipped. Clegg argues that most are rationalisations built after the event, which skip over the messy decision points that could have gone either way. So here’s the check to run this week. Pick your most contested governance document and ask whether it works as a starting point people build shared understanding around, or whether it’s quietly waiting to become evidence in an argument.
Don’t let your lessons die
I know what some of you are thinking: all very interesting, John, but what’s been happening in the regular publication of Campbell Systematic Reviews? Worry no more. It remains one of the most painful reads in my monthly pile, but this edition held a method paper that stopped me: Don’t Let the Evidence Die: From Rapid to Living Reviews, Experiences From the Global South, from the team at eBASE Africa in Cameroon.
On the surface it has nothing to do with project management. The authors ran rapid evidence reviews during COVID and found the products went stale almost immediately, so they rebuilt them as living reviews: evidence bases with update triggers, scheduled refreshes and continuous surveillance, where AI does the gathering and sorting and humans validate a sample of everything the machine produces. They are also candid about what went wrong along the way, which is rare in a methods paper and part of why I trust it.
So how does a COVID evidence unit’s process translate to a PMO? Swap the sources. They watched journals, preprint servers and even social media posts; we have meeting transcripts, the project documents scattered across SharePoint, delivery data in the PPM tool, and a steady drip of audit and assurance findings. Swap the update triggers too. Theirs were new studies and policy deadlines; ours would be gate reviews, project closures, incidents and a scheduled quarterly refresh, so the repository gets reopened on a rhythm instead of abandoned at handover.
The AI part of their loop is the piece most PMOs haven’t built yet. Classifiers screen and sort whatever the sources produce, candidate lessons get clustered and surfaced, and people hand-check a sample, roughly one item in ten, with the authority to overrule the machine. The authors are blunt about the failure modes: the models struggled with fragmented grey documents, and badly tuned alerts turned into spam that people learned to ignore. Their answer was people at the decision points and constant tuning of the triggers. A PMO borrowing this loop should expect the same maintenance, and budget for it.
The living review process from Tezok et al. (2026), ‘Don’t Let the Evidence Die’, Campbell Systematic Reviews, 22(2), translated into a lessons loop for the PMO. Diagram: HotPMO.
Most lessons-learned logs are written once at project closure and never reopened. They rot. Run the loop instead and you get a living record that compounds across the portfolio, a far richer seam than a spreadsheet column marked “lessons”.
We’ve already travelled from spreadsheets to tools to retrospectives. AI-supported living lessons look like the next step. If lessons learned isn’t on your list of AI candidates, add it. And if you’re already doing something in this space, I’d like to hear how it’s going.
Two days in Moorgate: the PMO Conference
June also brought the PMO Conference in London, with PMOLearn! on 17 June and the conference itself on the 18th. Lindsay and Eileen walked us through both days on the stream, and a few ideas have stayed with me since.
From the Learn day: this year’s agenda was built subjects-first, with speakers found to fit, and the recurring feedback was “did you write this agenda just for my organisation?”. The session on PMO models landed on a point I’d echo. The organisation decides what it’s ready to let the PMO do, so meet it where it is and add the value that’s available now; trust and credibility buy the mandate for more. The delivery-frameworks session offered a good-enough test worth stealing: does anyone know the framework exists, and would anyone miss it if it went? And the governance session argued that governance earns its keep only when it drives decisions, with stage gates giving you the view through the windscreen while weekly health checks watch the rear-view mirror.
On the conference day, the opening keynote pictured the PMO as an extension cable: useful, portable and temporary, which is exactly the problem. After a year of watching PMOs get disbanded or shrunk, the question of how a PMO becomes part of the furniture is the right one to sit with. The day closed with Bonita Norris, who stood on the summit of Everest in 2010, aged 22, then the youngest British woman to have done so. There were tears in the room, and I’ll admit some were mine.
Two pieces of House of PMO news from the event. The new Inside PMO report on specialist roles in the PMO launched on the 18th. And the PMO Service Catalogue is getting a refresh, with a review group you can join before the next version is published. If the conference left you with a notebook full of intentions, borrow the practice the day ended with: pick one lesson, tell a colleague what you’ll do with it, and do that one thing.
July: worth leaving your desk for
8 July, online: the House of PMO competency-framework series continues with a lunchtime session on the PMO skill of Collaboration, 12:30 to 1:30pm. Lunchtime well spent.
And one for the diary: the pilot of the new two-day Resource Management and the PMO course runs virtually on 5 and 6 October. It’s built on the circle of resource management the community helped create, a course by the profession for the profession.
You can watch the full session, including the news segment all of this came from, right here:
This roundup is based on my monthly PMO News segment from the PMO HotHouse, run by the House of PMO. If you’re not already part of that community, it’s well worth a look.
If June’s news says anything, it’s that AI in the PMO now needs a proper governance home. If you’d like a hand working out what that looks like in your organisation, from assurance mapping to living lessons, get in touch. We’d be glad to help.
