Why AI Makes Governance More Important, Not Less
Governance should not be the thing that slows delivery down. At its best, governance gives leaders the confidence to move faster because they can see the risks, understand the decisions, and trust the information being used to manage delivery.
That becomes more important as AI starts influencing how teams plan work, summarise information, identify risks, prepare updates and support decision-making.
My view is simple: AI does not make governance less important. It makes governance more important.
If the underlying delivery environment is weak, AI will not create control. It will only help teams produce unclear information faster. AI can help teams move faster, but if governance is weak, it will only help them create confusion faster.
The practical issue leaders are missing
A lot of organisations are still looking at AI through a productivity lens. They are asking whether AI can summarise meetings, draft status updates, prepare actions from transcripts or help project managers write steering committee papers.
Those are useful use cases, but they are not the real governance question.
The practical issue is whether AI is improving delivery confidence or simply producing more content. In enterprise delivery, governance is not just a set of templates or approval forums. It is how leaders know whether work is under control. It is how risks are understood, decisions are recorded, dependencies are managed, and controls and mitigations are tested.
This is where I think many organisations are missing a beat.
AI should not sit outside the delivery control environment as a general productivity tool. It needs to become part of how delivery information is checked, challenged and governed.
Governance should create confidence, not bureaucracy
There is a version of governance that slows teams down. Most delivery leaders have seen it: too many templates, too many forums, duplicated reporting, status packs that take more effort to prepare than they are worth, and approval processes that are unclear or disconnected from the actual risk.
That kind of governance deserves to be challenged.
But that is not an argument for less governance. It is an argument for better governance.
Good governance should create confidence that risks are visible, actions are owned, decisions are being made at the right level, dependencies are not being ignored, and teams are working within the organisation’s standards, controls and risk appetite.
When AI enters the delivery environment, that confidence becomes even more important. If AI is helping prepare delivery information, leaders need to understand what the output is based on, whether the source information is complete, and who remains accountable for the judgement.
The technology may produce the draft, but the delivery leader still owns the position.
AI will expose weak delivery practices
In many organisations, the issue is not that leaders have no information. The issue is that they cannot always tell which information they should trust.
Across financial services and regulated delivery environments, one pattern I have seen is that delivery teams are often working hard, but the underlying information does not always give leaders confidence. Risks may be weakly structured. Controls and mitigations may not be clear. Decision registers may be overlooked. Dependencies may be discussed in meetings but not reflected in the delivery plan. Status updates may look polished but still fail to explain whether the work is genuinely under control.
That is where AI becomes interesting, but also risky.
If AI is applied to a weak delivery environment, it may simply produce a better-written version of poor information. A steering update can sound clearer, a RAID summary can look more professional, and a meeting summary can capture actions more neatly. But if the risks are incomplete, the decisions are missing and the mitigations are weak, the organisation has not improved governance. It has improved the presentation layer.
That is not transformation. That is acceleration without assurance.
The real test is not whether AI can write the report
The better question is whether AI can help test the report.
For example, can AI review a status update against the underlying RAID information and identify inconsistencies? Can it check whether a risk has a clear cause, event and impact? Can it identify whether a mitigation is actually a mitigation, or just a general statement of intent? Can it check whether decisions discussed in meetings have been captured in the decision register?
These are more valuable questions than simply asking AI to create another project update.
The question I would be asking is: can AI help us understand where delivery confidence is weak?
That is where governance matters.
AI needs to sit inside a controlled delivery environment
A mature organisation should be able to take its PMO standards, delivery framework, RAID templates, decision logs, reporting expectations and governance guidelines, and use them inside a controlled AI environment.
The purpose is not to let AI make delivery decisions. The purpose is to help teams and leaders test whether delivery information is complete, consistent and aligned to organisational expectations.
In practice, this could mean using AI to help identify risks without clear causes, impacts or owners; mitigations that do not properly address the risk; decisions that have been discussed but not formally captured; dependencies without clear ownership or due dates; and status updates that do not align with the actual delivery evidence.
It could also help identify repeated actions that appear in meetings but do not close, steering papers that hide unresolved delivery risks, or projects operating outside the organisation’s delivery framework.
That is delivery assurance.
I have seen AI create practical value when applied to structured delivery communication, including reducing the effort required to prepare release-related communications. But the value did not come from AI operating in isolation. It came from having source information, a defined process, review points and human accountability.
That is the lesson leaders should take seriously. AI can reduce manual effort, but trust still depends on the quality of the inputs and the discipline of the control environment around it.
Communication tools matter, but control matters more
A lot of important delivery information does not start in a template. It starts in email, Teams, Slack, meetings and informal stakeholder conversations. Often, the first sign of a risk or dependency appears in a discussion before it ever appears in a formal report.
This is one reason AI could become valuable in enterprise delivery. It can help teams identify risks, actions, decisions and dependencies earlier, especially when relevant communication can be connected safely to delivery artefacts.
But this needs to be handled very carefully.
In regulated and risk-heavy environments, organisations should not connect AI to everything and assume the governance problem is solved. They need clear rules around access, confidentiality, data boundaries, retention, review and accountability.
The point is not unrestricted AI access. The point is controlled AI assurance.
AI should help teams identify what may have been missed, what needs review and where delivery confidence is weak. The delivery leader still owns the judgement.
The PMO role needs to evolve
This is also where the PMO conversation needs to mature.
A PMO that only produces status reports is going to struggle, because leaders do not need more slides. They need assurance.
The future PMO should not simply ask whether the report was submitted on time. It should help answer whether the delivery position is reliable, whether the risks are real and well controlled, whether decisions are being made at the right level, whether dependencies are visible and actively managed, and whether the steering committee is seeing the truth early enough to act.
That is a more valuable role than producing another pack.
AI can support that shift by helping PMOs review more information, identify inconsistencies and highlight areas where human judgement is required. But accountability cannot be outsourced to AI. The PMO, delivery leader and executive sponsor still need to own the decisions.
What leaders should stop doing
Leaders should stop treating AI as a shortcut around weak delivery discipline.
AI will not compensate for unclear governance. It will not replace ownership. It will not remove the need for strong delivery practices. It will not make a poor RAID process effective simply because the output is easier to read.
If the process is unclear, AI will not fix it. If accountability is unclear, AI will not fix it. If the delivery information is unreliable, AI may only make the unreliability harder to see.
That is the risk.
What leaders should start doing
Leaders should strengthen the delivery system AI will operate within. That means improving the quality of core delivery practices before scaling AI across them.
Start with the basics: strengthen RAID quality, improve decision recording, clarify delivery accountabilities, define what must remain human-owned, make controls and mitigations more explicit, and embed PMO standards into delivery workflows.
Then use AI to challenge delivery information, not just summarise it. AI outputs should be reviewable and evidenced. Governance should be connected to the real conversations where delivery risk emerges, not just the formal reporting cycle.
This is not about adding bureaucracy. It is about making governance more useful.
The organisations that get this right will not be the ones with the most AI tools. They will be the ones that understand how AI fits into their operating model, governance rhythm and accountability structure.
The leadership question
AI will make parts of delivery faster. The harder question is whether it will make delivery more reliable.
That depends on the governance environment around it.
If risks are poorly captured, AI will not create risk discipline. If decisions are not recorded, AI will not create accountability. If dependencies are hidden, AI will not create delivery confidence. If PMO standards are disconnected from how teams work, AI will not magically close that gap.
AI should support judgement, not replace it. It should help leaders see earlier, challenge faster and govern with better evidence.
The goal is not less governance. The goal is better governance: more continuous, more evidence-based and more connected to how work actually happens.
That is how organisations can use AI to move faster without losing control.
Comments