The Last Programme Director
We free the graduates. Then we close the school.
The school no one recognised
I have been trying to follow the programme director role forward by ten or fifteen years.
Not by asking whether AI will replace it. That question feels too blunt. Programme work is not one thing, and neither is the role. I am more interested in what happens when the parts begin to separate.
The first part to move is obvious.
Plans. Dependencies. RAID. Status. Steering packs. The weekly machinery through which a programme tries to keep a shared version of reality alive.
I do not say that dismissively. Much of this work matters because organisations are not naturally coherent. Different functions hold different facts. Suppliers see different risks. Sponsors hear different versions of progress. Somebody has to turn all of that into a picture people can act on.
Today, a large amount of human effort goes into manufacturing that picture.
I think AI will make much of that effort unnecessary.
And if that were the whole story, this would be a cheerful article. Programme directors would lose the administration and keep the judgement. Less time chasing updates, more time leading. Less governance theatre, more consequence.
I can see that future. I just do not think it is where the story ends.
Because some of the work we are about to remove was doing something else at the same time.
It was teaching us.
The plan was a school.
Not the Gantt chart itself. Nobody became wise by moving a bar three weeks to the right.
The education was in building the plan with people who disagreed about what had to happen first, who owned what, how long it would take and whether the promise was credible in the first place.
After enough years, you stopped hearing only the answer.
You heard the hesitation before it.
You noticed the date nobody wanted to own. You learned that an argument about sequence was sometimes an argument about trust. You learned that a dependency described as technical could be a commercial dispute, an incentive problem or an executive disagreement that had travelled down the organisation until it was wearing a scheduling costume.
The RAID review was a school too.
A junior practitioner learned to log a risk. An experienced programme director learned to wonder why the same risk had been sitting unchanged for four months.
The terrible steering meeting was a school.
The one where the sponsor was present but no longer sponsoring. The one where Finance asked a question that was not really about money. The one where everybody agreed to a decision and nobody behaved afterwards as if a decision had been made.
None of this was efficient.
Some of it was pointless.
But the waste was also a habitat. People spent years inside it and, occasionally, emerged able to read an organisation’s actual behaviour rather than its reported behaviour.
“We thought we were maintaining the programme. We were also learning the organisation.”
That is the part I think we have not thought hard enough about.
What happens when we remove the work, but still need the person that work used to produce?
The coordination layer goes first
I would bet quite heavily on the coordination layer becoming machine work.
Not tomorrow, and not everywhere at once. But follow the direction of travel.
A programme plan is a set of commitments, sequences and constraints. A dependency model is a network. A RAID log is structured uncertainty. A status report is a compression of change into a narrative for decision-makers.
The hard part is often not inventing these things. It is keeping them consistent while the programme moves.
That is exactly where machines should become useful.
Imagine a programme in which the plan is continuously reconciled against delivery evidence. Dependencies are inferred from changes across workstreams rather than rediscovered in meetings. Decisions are traced back to the conversations and evidence that produced them. Contradictions between reported status and observable behaviour are surfaced before somebody turns a box amber.
The Thursday status meeting begins to look strange.
Why are twenty people synchronising information that is already synchronised?
Why is a programme director asking whether a milestone is still credible when the consequences of every upstream change have already been recalculated?
Why are three analysts producing a steering pack that was effectively current yesterday?
This is not science fiction. It is the natural destination of a coordination system once enough of the programme’s evidence becomes machine-readable and the models become good enough to reason across it.
And I think most programme directors will welcome it.
I would.
There is no virtue in protecting administrative work because professionals happen to have built careers around it.
The capable programme director will say: finally. Now I can spend my time on the real job.
Judgement. Intervention. Politics. The difficult conversations.
For the existing generation, that may be exactly what happens.
We free the graduates.
Then we close the school.
What remains may not belong to us
The reassuring version of this future says that judgement remains human.
I am less sure.
AI will not stop at producing better plans and cleaner status. It will move into interpretation.
A system that can see programme evidence across meetings, messages, budgets, decisions, actions and delivery data may become very good at spotting patterns that today we call political instinct.
It may notice that a sponsor’s commitment is weakening before the programme director does. It may detect that a supplier relationship is becoming adversarial. It may see that two executives are behaving as if they have made different decisions even though the minutes record one.
So I would not stake the future of programme direction on humans retaining a monopoly on political sensing.
That boundary looks temporary.
The more interesting boundary comes later.
Observation.
Inference.
Recommendation.
Commitment.
Consequence.
A machine can observe.
It can infer.
It can increasingly recommend.
But at some point a recommendation becomes an organisational act.
Stop the workstream.
Move the money.
Abandon the date.
Choose one constituency over another.
Tell the regulator the commitment will not be met.
Overrule the executive who wants to continue.
That is where intelligence stops being the whole problem.
Someone needs the right to make the choice. Someone needs enough legitimacy for the organisation to accept it. Someone remains there when the consequences arrive.
I suspect this is where delegation will prove hardest: not at judgement itself, but at commitment under consequence.
And that creates an awkward problem for the programme director.
Because formal authority does not naturally sit with us.
It sits in the line.
The role splits
If coordination moves down into machines and consequential commitment moves up towards formal authority, the programme director is squeezed from both directions.
I can imagine two things happening at once.
For a large part of the market, the role simply erodes.
Not with an announcement. There will be no memo declaring the end of programme direction.
A programme director leaves. The organisation looks at the new coordination tooling, looks at the SRO, looks at the functional leads and decides not to refill the post.
The programme was, in truth, a large project wearing a programme label. Coordination was most of the job. Once coordination is cheap, the justification goes with it.
At the other end, something much more powerful survives.
The genuinely difficult transformations do not become easy because the information becomes clean. Executives still have different incentives. Suppliers still defend their economics. Regulators still impose constraints. Sponsors still lose attention. Functions still protect themselves.
Somebody still has to make those systems move together.
But look carefully at that person.
If they can settle an executive trade-off, commit resources, kill a major initiative, override a function and remain answerable for the outcome, are they still a programme director?
Or have they become a temporary executive?
A mission leader.
An outcome owner.
A general manager for a piece of institutional change.
This is where I think the profession may misunderstand its own future.
The attractive story is that AI removes the bureaucracy and reveals the programme director in their purest form.
My suspicion is almost the opposite.
Once the coordination layer disappears, we discover that the remaining work either does not justify a senior standalone role, or requires more authority than the role traditionally possesses.
The work survives.
The profession may not.
The organisation that can see everything
Here is the future I find more unsettling than simple automation.
Imagine an organisation in 2036 that has become extremely good at programme coordination.
The plan is current.
The dependencies are visible.
The forecasts move continuously.
The system knows which commitments conflict, which risks are correlated and which decisions are being avoided.
There is very little informational darkness left.
And the programme is still failing.
Everybody can see the trade-off that has to be made.
The system has modelled the options. It has identified the least damaging route. It may even predict which settlement the organisation is most likely to accept.
And nothing happens.
One executive will not absorb the loss.
Another will not surrender the date.
The sponsor will not spend the political capital.
The analytically correct answer is institutionally unacceptable.
The machine has not failed.
It has done exactly what we asked.
The failure is that knowing what the organisation should do is not the same as making the organisation do it.
“Coordination can become abundant while commitment remains scarce.”
That distinction may turn out to be the real delegation boundary in programme work.
It is also where the apprenticeship problem returns.
Because the person who knows how to move an organisation through that moment did not learn it from a framework.
They learned it by spending years in the machinery we just removed.
The last graduate
This is where the title starts to mean something different to me.
The last programme director may not be the final person whose job title contains those words.
They may be the last person formed by the old route.
Picture them around 2040.
They learned the craft before the coordination layer disappeared.
They built the thirty plans.
They sat through the steering meetings that should have been emails and the steering meetings that changed everything.
They heard hundreds of plausible explanations and slowly learned which ones deserved belief.
They saw sponsors remain sponsors in title after they had stopped sponsoring in behaviour.
They watched technical dependencies become political fights and political fights disguise themselves as delivery problems.
Then the machinery changed.
Their working week now looks almost nothing like programme management.
They do not maintain the plan.
They do not run RAID.
They do not assemble the status pack.
The programme already knows its state.
Their week is a sequence of interventions.
A sponsor whose attention has moved elsewhere.
A supplier dispute that will become contractual unless somebody changes the conversation now.
A choice between protecting the business case and protecting the date.
A workstream that needs to die even though hundreds of people have spent two years building it.
A recommendation from the machine that is analytically sound and organisationally impossible.
They are valuable because they know the difference between what the programme says and what the institution will actually do.
The organisation wants another one.
And here is the problem.
Where does the next one come from?
The route that produced the first person has been dismantled.
Not because anyone wanted to destroy it. Because the work looked inefficient, and much of it was.
Perhaps we will create a better school.
We may use simulations, rotations, deliberate exposure to live decisions, AI-supported shadowing, new forms of apprenticeship.
I hope we do.
But that is not what we have today.
Today the profession is preparing to automate an apprenticeship it has barely recognised as an apprenticeship.
That creates a possibility I had not considered when I started thinking about this piece.
The last programme director may not be replaced by AI.
For a while, they may become more valuable because of it.
Not because human judgement is mystical. Not because machines can never learn politics. Not because the old ways were better.
Because the old system produced something accidentally, and the new system has not yet decided how to produce it deliberately.
The last programme director is therefore not the last person capable of running a programme.
They are the last graduate of a school we closed before we knew it was a school.
What I am staking
By 2031, AI systems will be capable of maintaining programme plans, dependency structures, risk and issue records, decision histories and routine reporting to a standard equal to or better than a competent human coordination layer.
As that capability matures, programme organisations will become materially smaller. Roles whose principal value is maintaining coherence will disappear faster than roles whose value lies in consequential intervention.
The delegation boundary will move beyond administration and into judgement. AI will increasingly observe political behaviour, infer organisational conditions and recommend consequential actions. The harder boundary will sit at commitment: the point where an assessment becomes an organisational act requiring legitimate authority and someone who remains answerable for the consequence.
By 2035, a substantial share of roles currently called programme director will either have disappeared or will describe a fundamentally different job. At the top end, the surviving role will move closer to temporary executive authority than to programme direction as we know it now.
The coordination work being automated is also part of the apprenticeship through which experienced programme directors develop tacit organisational judgement. Organisations that remove that exposure without creating a replacement formation system will create a capability gap: abundant programme information, fewer practitioners able to convert it into institutional commitment.
The work survives. The profession may not.
— The Fringe makes claims. The Lab tests them. Verdict to follow.