What the 2026 Research Reveals About UK Transformation Failures
The headline numbers from ILX Group's 2026 survey of 600 UK IT and project leaders are striking. 49% of UK organisations report that their transformation initiatives are misaligned with strategic priorities. 46% still classify change management as a 'nice to have' rather than a core delivery requirement. Meanwhile, 75% cite project, programme, and portfolio management skills as a top priority yet the organisational muscle needed to embed change is absent in nearly half of those same businesses.
The gap between building capability and embedding adoption is where AI investment disappears.
In CPG and FMCG specifically, this looks like a demand forecasting model that works and a commercial planning team still running the process in Excel. Not because the model failed. Because nobody designed the adoption.
Why AI Programmes Deliver Technology and Not Results
Most AI delivery projects focus on the build: the data pipeline, the model, the dashboard. These are measurable milestones with clear deliverables. Change management is harder to define, slower to evidence, and easier to defer to the next phase.
The result is a recognisable pattern. The technology performs. Adoption does not follow. The metric that tracks this most clearly is the override rate: the proportion of AI-generated recommendations a human team overrides in favour of a manual decision. It is a cleaner signal than model accuracy, and it tells the story within weeks of go-live.
A high override rate means the system is working and the business is not using it. The ROI calculation begins to collapse not in the technology, but in the habit.
In one CPG business assessed through FlightCheck™, the override rate on AI-generated demand signals was running above 70% six months post-deployment. The model was accurate. The forecast was better. But the commercial team had not been given a reason or a structure to trust it. That is a change management failure, not an AI failure.
The Alignment Gap: When Strategy and Execution Pull in Different Directions
The ILX Group finding that 49% of UK businesses have transformation initiatives misaligned with strategic priorities is not a project management observation. It is a governance observation.
In FMCG and logistics, misalignment typically surfaces in one of three ways:
- The AI programme is owned by the technology function but the expected benefits sit in commercial or operations. There is no shared definition of success, so accountability belongs to nobody.
- The board has approved AI investment but it has not filtered into KPIs, S&OP cycle design, or incentive structures. Senior leaders are publicly supportive; the middle layer has not changed what it measures.
- The pilot delivered results. Nobody redesigned the process to make AI the default. The workaround became permanent.
None of these are failures of AI. They are failures of alignment between the technology investment, the operating model, and the leadership behaviour needed to sustain it.
What Change Management Actually Means in a CPG or Logistics Context
Change management in AI programmes is not workshops. It is not a culture deck. It is a set of deliberate decisions about ownership, measurement, and incentive that make the AI-enabled process the path of least resistance for the people doing the work.
In practical terms, for a CPG or logistics business, this means:
- A named owner for each AI-enabled decision area — not the data team, but the operational function whose behaviour needs to change.
- An override rate tracked from week one as a leading indicator of adoption health, not as a lagging indicator reviewed in quarterly business reviews.
- Leadership behaviour that models the new process visibly. If the Supply Chain Director is still requesting the manual version, the team will not trust the automated one.
- Training delivered in the context of real workflows — not generic AI literacy sessions, but specific guidance on what the system does, what it does not do, and where human judgement remains essential.
- A clear consequence structure: what happens when an AI recommendation is overridden, who reviews it, and what it feeds back into the programme.
These are operational decisions. They belong in the programme plan from day one, not in a separate workstream six months after go-live.
AI Navi Insight: From the FlightCheck™ Files
AI Navi Insight Across AI Navi's FlightCheck™ diagnostic assessments, the Leadership dimension of the SCALE AI™ framework consistently scores lowest. The benchmark across mid-market CPG businesses sits at 18% — the weakest of the five SCALE dimensions — against an average AI confidence score of 4.1 out of 10. What drives that score is not a lack of leadership interest in AI. What drives it is the absence of the structural conditions that make AI adoption the default: clear ownership, explicit performance measures tied to AI-enabled decisions, and visible senior behaviour modelling the new process. For a £400M CPG client, the Flight Risk Index™ score dropped from 7.2 to 4.1 in 60 days. The intervention that moved it most was not a model retrain. It was the introduction of override rate tracking, a named AI ownership structure, and a revised S&OP agenda that made AI-generated forecasts the starting point rather than the reference point. The technology had been in place for months before that decision was made. |
Change Management Built In, Not Bolted On: The SCALE AI™ Approach
The SCALE AI™ methodology addresses this directly. The five dimensions — Strategy, Capability, Applied AI, AI Leadership, and Data Architecture — are designed to be assessed and addressed together, not sequentially.
Leadership (L) covers the ownership, governance, and behavioural conditions for adoption. Capability (C) covers the skills and training needed for humans to work effectively alongside AI systems. Both are assessed in the FlightCheck™ diagnostic before any build begins.
This means change management is part of the delivery design, not a separate track that follows it. By the time a FlightPath™ Sprint delivers working AI in production, the ownership structure, the adoption measures, and the stakeholder alignment are already in place.
The alternative — delivering the technology and planning adoption afterwards — is how 46% of businesses end up with AI programmes that work technically and underdeliver commercially.
Change Management In vs Out: What the Programme Looks Like
| Factor | Without Change Management | With Change Management Built In |
|---|---|---|
| Override rate at month 6 | Typically 60-80%. Unknown and untracked. | Measured from week one. Intervention triggered early. |
| Ownership | Technology team holds accountability. Commercial team disengaged. | Named operational owner per AI-enabled decision area. |
| S&OP integration | AI output is a reference, not a starting point. | AI forecast is the S&OP default. Manual override is the exception. |
| ROI realisation | Technology works. Business value is not captured. | Business value tracked from deployment against defined baseline. |
| Programme risk | Adoption gap discovered at six-month review. | Adoption risk identified in FlightCheck™ before build begins. |
Three Questions to Ask Before Your Next AI Programme
Before committing resource to any AI initiative, these three questions determine whether adoption is likely:
1. Who owns the outcome?
Not the build — the decision this AI system is designed to improve. Is that person actively involved in the delivery?
2. What is your adoption metric?
Not model accuracy. Not data quality. The specific metric that tracks whether human behaviour has changed in response to AI output. The override rate is the most useful starting point.
3. What happens when someone overrides the AI recommendation?
Is that logged? Is it reviewed? Does it feed back into the system? If the answer to any of these is unclear, the programme has a change management gap that technology will not close.
Frequently Asked Questions
Why does AI fail in FMCG when the technology is working?
The most common cause is adoption failure rather than technical failure. AI that produces accurate recommendations has no business value if the operational team continues to make decisions without reference to them. Change management specifically ownership structure, override rate tracking, and behavioural incentives determines whether the technology delivers ROI or sits unused.
What is the override rate and why does it matter?
The override rate is the proportion of AI-generated recommendations that a human team overrides in favour of a manual decision. It is the clearest leading indicator of adoption health in an AI programme. A high override rate does not necessarily mean the AI is wrong it usually means the conditions for trust have not been established. Tracking it from go-live gives programme leaders something specific to act on.
How long does it take to embed change in an AI programme?
Structural change ownership, metrics, revised governance can be implemented within the first 30 to 60 days of deployment if planned in advance. Behavioural change takes longer but is measurable throughout. The businesses that struggle most are those that treat change management as a downstream activity after deployment rather than a parallel workstream from the beginning.
What does the SCALE AI™ Leadership dimension assess?
The Leadership dimension of SCALE AI™ assesses whether the organisational conditions for AI adoption are in place: named ownership, explicit KPIs linked to AI-enabled decisions, visible senior behaviour modelling the new process, and a consequence structure for overrides. In AI Navi's FlightCheck™ diagnostics, Leadership consistently scores as the weakest of the five SCALE dimensions across mid-market CPG businesses.
How does AI Navi address change management within its delivery model?
Change management is embedded from the FlightCheck™ diagnostic stage. The Leadership and Capability dimensions of SCALE AI™ are assessed before build begins. The FlightPath™ Sprint delivers working AI in production with adoption infrastructure already in place: ownership structure, override rate tracking, and stakeholder alignment. This is not planned as a post-go-live workstream.
| If your AI programme has the technology but not the adoption, a FlightCheck™ diagnostic will identify exactly where the gap sits and what to address first. Book a free 30-minute discovery call with the AI Navi team, or take the AI Readiness Scorecard at ai-readiness-scoring.scoreapp.com. |