Case Study — Bringing Fragmented AI Activity Back Under Programme Control

1. Organisational Problem

A large, regulated organisation found that emerging AI activity was becoming increasingly difficult to manage through a structured programme view.

What had begun as isolated experimentation was becoming harder to see, align and govern coherently at an organisational level. AI activity was increasing through departmental initiatives, local operational pressures, growing interest in generative AI tools and external expectations for visible progress.

This was not because the organisation lacked governance or delivery discipline. It already had:

• Mature programme management arrangements

• Established governance and assurance frameworks

• Defined delivery standards and controls

The problem was that AI activity was moving faster than it could be consistently brought within those structures.

As a result:

• Initiatives were progressing at different speeds and with different levels of control

• Some activity was emerging outside a single programme view

• Use of AI tools was increasing ahead of formal oversight

• Executive and programme leaders had incomplete visibility of what was happening across the organisation

This created a pressing challenge for programme leadership: how to bring fragmented AI activity back under structured control without stopping useful progress.

This situation is not unusual. It reflects a broader pattern seen across many organisations, where AI activity begins in a fragmented way before any overarching structure is in place.

This case study illustrates how fragmentation can be brought back under control — and how structured programme management plays a critical role in restoring clarity, alignment and delivery confidence.

In Orr Consulting’s AI Transformation Process, situations like this often reflect AI activity moving into Deliver before enough Discover and Design discipline is in place. The full Process provides a practical structure for improving line of sight across AI activity, strengthening governance and assurance and reducing delivery risk across the organisation.

2. Situation

The organisation was not starting from zero.

It was attempting to bring fragmented AI activity back under control while that activity was already underway.

This meant the challenge was not simply one of defining a new AI programme. It was one of recovering visibility, structure and confidence in a live environment.

Several programme-level tensions had become clear:

• AI activity was expanding faster than it could be consistently seen, governed and coordinated

• Department-led initiatives were progressing unevenly, with different levels of control and maturity

• Programme leaders were accountable for oversight, but did not yet have a complete line of sight across emerging AI activity

• Existing governance arrangements needed to be applied and adapted to reflect the particular uncertainty and risk profile of AI

• Delivery teams were being asked to move at pace in an environment where requirements, assumptions and risks were less stable than in more traditional change initiatives

At executive level, the risks extended beyond programme discipline alone. Left unmanaged, this pattern risked fragmented investment, unmanaged exposure, weak prioritisation of delivery capacity and executive accountability without reliable visibility or control.

3. Background

The organisation already had credible programme, governance and delivery disciplines in place.

This was not a case of an organisation with no structure suddenly attempting to adopt AI. Rather, AI was emerging through multiple local pressures faster than existing structures could consistently absorb, coordinate and direct it.

Some upstream weaknesses had contributed to the situation. In several areas, AI activity had moved forward without enough early clarity around:

• The problem being solved

• The outcome being sought

• The data and dependencies involved

• The complexity and risk of the proposed solution

However, once the situation had become a live reality, the priority was not to assign blame or simply point backwards to missed Discovery and Design activity.

The immediate need was to recover control in a way that:

• Reduced unmanaged risk

• Improved leadership visibility

• Strengthened decision-making

• Allowed useful AI activity to continue under more disciplined conditions

This required a combination of retrospective process intervention and active AI programme management leadership.

4. Action Taken

A structured response was introduced to reassert programme control while allowing AI activity to continue.

The approach combined targeted retrospective application of the Orr Consulting AI Transformation Process with broader programme management actions suited to a higher-uncertainty environment.

4.1 A Programme View of Activity

The first step was to create a reliable programme-level picture of what was happening across the organisation.

This included:

• Rapid identification of active, proposed and informal AI initiatives

• Consolidation into a single view of AI-related activity

• Initial categorisation by purpose, capability type, delivery status and risk profile

• Identification of activity taking place outside normal approval and oversight routes

This created the visibility needed for programme leaders to begin re-establishing structured control.

4.2 Reasserting Governance

With visibility improved, attention turned to restoring clear governance oversight.

This included:

• Clarifying programme leadership accountability for AI activity across the organisation

• Defining how existing governance and assurance forums would oversee AI-related decisions

• Introducing proportionate controls for data, risk, assurance and acceptable use

• Setting clearer decision rights for which initiatives could proceed, pause, scale or stop

The aim was to ensure that AI activity sat within a credible and structured control environment.

4.3 Retrospective Work

Where activity had moved forward without enough clarity, targeted retrospective work was carried out to strengthen the foundations for delivery.

This included:

• Clarifying the underlying business problem and intended outcome

• Reassessing alignment to strategy and value

• Reviewing data readiness, dependencies and delivery complexity

• Testing assumptions that had been allowed into delivery too early

In effect, selected elements of AI Use Case Discovery, prioritisation and Design-stage shaping were applied retrospectively to improve the quality of decisions already in motion.

This helped distinguish between activity that should continue, activity that should be reshaped and activity that should not proceed further.

4.4 Managing Higher Uncertainty

A central lesson was that AI delivery could not be managed in exactly the same way as more predictable change initiatives.

Programme management therefore adapted by:

• Structuring work into shorter controlled cycles

• Treating assumptions as items to be tested, not facts to be relied upon

• Reviewing delivery confidence more frequently

• Using evidence from early cycles to inform subsequent decisions and investment

This improved control while giving leadership earlier evidence and a stronger basis for investment decisions.

4.5 Strengthening Prioritisation

As demand continued to grow, stronger prioritisation was needed.

This included:

• Creating a more disciplined route for new AI ideas and requests

• Filtering activity based on strategic fit, readiness, risk and likely value

• Focusing leadership attention on the initiatives that most warranted coordinated support

• Reducing fragmented effort across too many disconnected activities

This helped contain the spread of AI activity and redirect momentum into a more manageable programme shape.

4.6 The Human Dimension

The response also recognised that programme control depends partly on organisational behaviour.

This included:

• Clearer communication of how AI activity should be progressed

• Support for leaders and teams moving from informal experimentation to structured delivery

• Targeted education and guidance to build confidence and consistency

• Active attention to adoption, readiness and stakeholder expectations

This reduced the likelihood that AI activity would continue to emerge outside programme line of sight.

5. Outcomes

The organisation was able to regain control of AI activity without disrupting progress.

This resulted in:

• A single, coherent view of AI initiatives replacing fragmented and opaque activity

• Improved programme leadership visibility and accountability across live and emerging AI work

• Governance and assurance were re-established in a way better aligned to AI-specific risks

• Weak or misaligned initiatives being filtered out earlier, reducing downstream issues and avoidable investment

• More consistent and controlled delivery approaches, despite inherent uncertainty

• Increased confidence across executive, risk and delivery communities

At a broader level, this improved the organisation’s ability to:

• Focus investment and leadership attention on the most valuable opportunities

• Reduce exposure to uncontrolled or poorly justified AI activity

• Make better-informed decisions about which initiatives to scale

• Move from fragmented experimentation toward a more coherent enterprise approach to AI

Importantly, AI activity was not stopped or reset. It was brought back under structured control and shaped into a more manageable programme.

6. Recommended Next Steps

First, it would continue strengthening earlier Discover and Design activity so that new initiatives entered delivery with clearer problem definition, stronger prioritisation and better understanding of dependencies and risks.

Second, it would further embed proportionate AI governance and assurance within existing programme structures.

Third, it would continue developing a more disciplined route for identifying, assessing and prioritising emerging AI opportunities, reducing the likelihood of future fragmented or uncontrolled activity.

Fourth, it would build organisational capability and confidence through targeted education, clearer guidance and stronger expectations around how AI activity should be initiated and progressed.

Together, these steps would help move the organisation from reactive containment toward a more mature model of AI programme leadership.

7. Final Thoughts

As AI adoption accelerates, this pattern is likely to become increasingly common.

AI does not always enter organisations through formal programmes. It often emerges organically, driven by local need, curiosity and external pressure.

For senior leaders, this creates a distinctive challenge: they remain accountable for risk, investment and outcomes even when AI activity evolves faster than existing structures can fully absorb.

This case study shows that the response is not simply to pause everything or look backwards. It is to restore visibility, re-establish control, strengthen upstream thinking where needed and apply disciplined programme leadership across a changing AI landscape.

This Case Study is part of the Orr Consulting AI Insights Library — structured thinking for AI transformation leaders and decision makers.

8. Call to Action

If AI activity in your organisation is growing beyond existing programme visibility and control, it is possible to bring it back under structured control without losing momentum.

The Orr Consulting AI Transformation Process provides a structured way to stabilise, prioritise and govern AI activity across Discover, Design and Deliver.

If this case study reflects the situation your organisation is currently facing, Orr Consulting would be pleased to discuss your next AI steps.

Discuss your next AI steps

Subscribe to Orr Consulting to receive occasional emails with practical AI Insights and updates.

Subscribe to Orr Consulting