Neither Adam nor Angel

The saviour story has numbers behind it. Gartner has forecast the data-gathering, the status-chasing, the report-assembling will be automated by the end of the decade. That equates to approximately four-fifths (80%) of today’s project-management tasks being performed by AI. Read quickly, that sounds like a profession dissolving. The assassin story borrows the very same figure and points it at your salary decreasing, reduced job opportunities and nowhere to go for professional advancement. And yet the actual adoption numbers are those of an industry still lacing its boots: barely a fifth (20%) of project managers report AI genuinely in use, and while roughly three-quarters (75%) say they are optimistic about it, at the same time admitting they have had nowhere near enough training to know what they are optimistic about. The costumes are vivid but the wardrobe is mostly still on the rail.

The reality sits somewhere in the middle. AI is good at the administration of projects. It will draft your weekly report, turn a scrawl of notes into minutes and actions, produce a first-cut plan, keep a log tidy, and find the pattern in a year of budget data faster than you can find the spreadsheet. What it is poor at is everything that made the job hard in the first place: the judgement call with half the facts missing, the stakeholder who says yes and means no, the decision that can only be defended by a person willing to be blamed for it. It eats the paperwork. It cannot eat the responsibility and accountability.

There is even a pleasing wrinkle in who benefits. It is for those new to the project profession. It seems that AI is less a replacement for competence than a leg-up towards it. In a controlled field experiment run in 2023 by researchers at Harvard Business School and Boston Consulting Group, the people whose work improved most when handed an AI assistant were not the experts but the relative novices. This was a lift of around forty-three per cent for the lower-performing half, against seventeen for the top. For the new project manager, the small team, the person running something that matters without a PMO down the hall, that is genuinely good news, and worth saying plainly amid the noise: AI raises the floor faster than it raises the ceiling, but not without issues.

Before any champagne is popped, let’s be clear, the AI outputs are only as good as the data that is fed into it and the discipline of the framework that data flows through. The same tool that drafts your report will, with total composure, invent a figure that is wrong, cite a risk that does not exist, and present both in prose so tidy you will be tempted not to check. It is only ever as good as the data you feed it, and most projects are fed on scraps. It does not know the office politics, the tired and under resourced team, the way the market moved leading to a strategic shift. It will not rescue a failing project; it will simply produce the wrong status report more quickly.

“AI does not supply discipline. It amplifies whatever discipline it finds.”

Which is rather the point, and the reason this is neither a rescuer nor a robber. What changes is not the nature of the job but the location of your hours. The time you once spent producing documents you will now spend reviewing them; the time you save assembling the pack, will now be spent deciding what the pack should say. The centre of gravity shifts from doing the admin to owning the outcome. This was always the part that mattered, and is now simply the part that is left.

Now for the unapologetic promotion of the SSLM framework. The shift to an outcome focus is exactly why a light, disciplined framework earns its keep in an age of automation rather than losing it. If the ordinary work is about to become nearly free, the scarce and valuable thing becomes the structure that makes it trustworthy: one project reference that threads every artefact, one source of truth for the machine to read, and the plain SSLM habit of doing the least sufficient thing well rather than the most impressive thing badly. Give a machine a clean, consistent shape to work in and it will hand the boring half back finished; keep that shape simple, and you will actually be able to check what it hands you. A simple framework is not made redundant by the automation. It is what lets you trust it.

So let us put down both costumes. AI in project management is not our Adam and not our fallen Angel. It is a very fast pair of hands that still, always, needs assurance and someone to decide. It makes the ordinary parts of a project nearly free — and in doing so it does not shrink the job, it concentrates it onto the judgement, the accountability and the light structure that were the point all along. That is the thread we will follow through this “Loving the AI-lien” series, and it begins here, with the plainest version of it: the machine can do the work; it cannot be the one who is answerable for it.