When a digital transformation programme stalls, the inquest follows a familiar script. The technology was poorly chosen. The vendor overpromised. The team lacked capability. The change management was inadequate.
These verdicts feel satisfying because they are specific and because they point at someone. They are also, in the majority of cases, wrong.
The technology works. The team is capable. The vendor delivered. And the programme is still slower than it should be, still consuming governance overheads that dwarf the cost of the tools themselves, still producing the dispiriting spectacle of modern analytical tools running at full speed through a process designed for a slower age.
The real suspect has not yet been interviewed.
A brief history of misplaced trust
In the 1960s, Douglas McGregor identified two competing assumptions about human motivation. Theory X holds that workers are reluctant, require close supervision, and will avoid responsibility. Theory Y holds that people seek meaningful work and accept – even seek – responsibility when the conditions are right.
The private sector spent six decades, haltingly and imperfectly, moving toward Theory Y. Delegation, devolved accountability, and outcome-based performance management became the dominant grammar of modern management.
Government did not make the same journey. It adopted the vocabulary – empowerment, agile, transformation – while keeping the underlying architecture of Theory X intact. The result is a system that hires capable, frequently brilliant people, places them in director and director general roles, and then requires them to convene committees of peers before renewing a £30,000 software support contract.
This is not accountability. It is its impersonation.
The indulgence market
Multi-signatory approval processes are routinely described as quality controls. They are more accurately described as liability distribution mechanisms.
When twelve officials sign off on a decision, no individual bears meaningful responsibility. If the programme succeeds, credit is shared. If it fails, the audit trail demonstrates due process. The Public Accounts Committee sees a procedure followed, not a decision made.
What looks like governance is closer to what Cyril N. Parkinson observed in 1955: committees gravitate toward decisions where everyone feels qualified to contribute, and away from those requiring visible personal accountability. The approval chain does not improve the decision. It insures the approvers against it – at a cost, in senior time and deferred delivery, that is rarely calculated and rarely even visible.
The evidence hiding in plain sight
In September 2025, the Government Digital Service began piloting a different approach across HMRC, NHS and DSIT: outcome-based portfolio funding, where each level of the hierarchy optimises against an explicit objective function and holds delegated authority within defined parameters.
A portfolio director who knows their objective – maximising service-value per pound of discretionary spend, within defined risk and reversibility thresholds – does not need a committee. They have a criterion. They can decide, defend their decision, and be held accountable for the outcome.
This month’s NAO report on the government shared services programme shows what happens without this architecture. Twenty-five digital change programmes across government ran in parallel with inadequate portfolio-level coordination, each with its own benefits claims and design assumptions, until the accumulated interdependency failures forced a reset of the Applicant Tracking System programme at an estimated cost of £26–38m. By that point, £459m in costs had already been incurred by the broader shared services transformation. The question is why the governance architecture – lacking a single owner with the mandate to enforce coordination – did not surface the need for intervention earlier. Without level-specific objective functions and work in progress constraints, no one had a criterion to call it.
The real suspect
A director general did not reach that position because they cannot make decisions. They reached it because they can. The approval chain does not add to their judgement – it substitutes process for it, signalling something corrosive: that the organisation does not trust the people it has appointed to lead it.
Theory X encoded in governance architecture produces exactly the outcomes Theory X predicts: process compliance is rewarded; outcome failure is punished. The Agile coaches cannot fix this. The AI tools cannot fix this. A faster horse is still a horse.
What departments can do differently
Define the objective function before delegating the decision. Delegation without criteria is not empowerment – it is exposure to blame without the means to defend the decision. A director who can articulate the objective function for their portfolio, and the parameters within which they hold authority, can decide quickly, transparently and accountably. Without that articulation, escalation is the only rational response.
Make the cost of consensus visible. The approval process for any significant decision should carry an estimated cost alongside the decision brief: senior time required, number of stages, elapsed time. When that figure approaches the value of the decision itself, the governance architecture has failed – regardless of whether the correct box has been ticked.
Treat work in progress as a governance variable. Organisations running more concurrent initiatives than their governance capacity can oversee will systematically underperform – not because the initiatives are wrong, but because the portfolio has no constraint mechanism. Portfolio-level work in progress limits are not a management fashion. They are the structural precondition for accountability at scale.
The Monday morning test
Take any digital decision worth less than £100,000 taken within your directorate in the last six months. Write down, in two sentences, the criteria against which you could have made that decision independently – without convening a committee or seeking peer sign-off.
If you cannot write those sentences, your directorate does not have delegated authority. It has the appearance of it.
The suspect was never the technology.
Vsevolod Shabad, FBCS, advises boards and senior leadership teams on governance failure, decision architecture and digital accountability, drawing on executive experience across critical infrastructure and regulated sectors in eight countries