Mapping the gap
Why place-based AI adoption might demand a new model for change
The Department for Science, Innovation and Technology's 2026 AI Adoption Research confirms what many regional and place-based practitioners have long suspected: AI uptake across the UK economy is modest, uneven, and concentrated in sectors and geographies already advantaged by capital, connectivity, and skills. Only one in six UK businesses currently uses AI, and the majority have no active plans to adopt it. For England's Mayoral Combined Authorities, Freeport zones, Investment Zones, and Further and Higher Education institutions operating as ecosystem anchors within these economic geographies, the policy and practice implications are profound.
This thought-leadership green paper argues that the dominant models of change management and technology adoption are insufficient for the place-based challenge. It introduces the PACE Matrix, a diagnostic framework distinguishing Institutional Readiness from Workforce Orientation, and proposes an ecosystem model of AI adoption in which training interventions at every level, from executive leadership to operational staff to supply chain partners, constitute the primary lever for territorial economic development.
1. The policy moment
The DSIT AI Adoption Research, published this month and drawing on fieldwork conducted between February and May 2025 across 3,500 UK businesses, provides the most granular picture of AI adoption patterns yet available to policymakers and practitioners. Its headline finding is stark: 80% of UK businesses neither use AI nor have plans to do so. Only 16% are current adopters, with a further 5% planning future adoption.
That finding alone would be notable. What renders it urgent for place-based economic development is the distributional pattern beneath it. Adoption is significantly higher in London (20% versus the national 16%) and in sectors concentrated in metropolitan cores: information and communication (43%), finance and real estate (21%), and business services and administration (23%). Scotland registers the highest proportion of businesses with no AI plans at all (84% versus 80% nationally). Construction, retail, transport and storage, and hotel and catering, the sectors that form the productive backbone of many Combined Authority economies, show adoption rates of 12% or below.
The AI Opportunities Action Plan, published in January 2025, set out a vision of AI embedded across the UK economy, supporting productivity, competitiveness, and inclusive innovation. It identified three pillars: enabling infrastructure (compute, data, skills); transforming public services; and securing technological sovereignty through homegrown frontier capabilities. The DSIT research reveals the distance between that ambition and current reality. The skills pillar is the most acutely underdeveloped: limited AI skills and expertise is cited as the second most common barrier to adoption (60%) across all businesses, and the most common barrier among those who are actively planning to adopt but cannot yet do so (68%).
This paper is a contribution to the more specific and more tractable question: what does an effective AI adoption strategy look like when the unit of analysis is a place, a city-region, a Freeport zone, an Investment Zone, a Further Education catchment area, rather than a firm or a sector? I suggest that the question requires a different model of change, a different theory of intervention, and a different relationship between the institutions that anchor economic geographies and the businesses and communities they serve.
2. What the change management canon tells us
As an associate visiting professor in leadership and management, with a long history of running complex organisations and coaching, I have found change models to be, in many ways, the backbone of my approach. The academic literature on organisational change is both extensive and convergent on a set of foundational claims. Those that have most influenced my teaching and practice share common features that can be applied in this context.
Lewin's three-stage model, unfreezing, changing, refreezing, established the principle that change requires destabilising existing equilibria before new behaviours can be embedded. Kotter's eight-step model extended this into a sequenced programme emphasising urgency, coalition-building, and the consolidation of gains before attempting institutionalisation. Both frameworks share an assumption that change is something done to an organisation from a position of leadership, and that resistance is a predictable and manageable feature of the process.
More recent formulations have complicated that picture. Bridges's transition model distinguishes between the external event of change and the internal psychological transition individuals must navigate, insisting that organisations frequently manage the change without attending to the transition. The Satir Change Model captures the nonlinear dynamics of transformation, noting that performance typically deteriorates before it improves and that the period of chaos between the old status quo and integration is frequently the point at which change programmes fail.
Technology adoption specifically has been theorised through Rogers's Diffusion of Innovations, which identifies innovators, early adopters, early majority, late majority, and laggards as distinct population segments with different motivations and risk tolerances.
What unifies this body of work is its primary focus on the organisation as the unit of change, and on individual psychology as the primary mechanism of adoption. But these models are insufficient for the place-based challenge for three reasons.
First, they assume a relatively bounded organisation with clear leadership authority. Mayoral Combined Authorities, Freeport governance structures, Investment Zones, and Further Education consortia are not organisations in this sense. They are ecosystems: overlapping assemblages of institutions, businesses, community organisations, and regulatory bodies with distributed authority and contested legitimacy. Change in these contexts is not directed from the top; it is negotiated across a horizontal field.
Second, the canonical models treat the environment in which change occurs as context rather than content. The economic geography of a place, its industrial composition, its transport connectivity, its skills infrastructure, its history of deindustrialisation or growth, shapes what AI means for the businesses and institutions within it. A firm in Teesside operates in a different environment from a firm in Shoreditch, even if they are nominally in the same sector.
Third, and most directly relevant to the DSIT findings, the canonical models address adoption within organisations that have already made a decision to change. The more fundamental challenge revealed by the research, that 80% of businesses have not identified a use for AI, or lack the skills to evaluate whether they should, is a pre-adoption problem. It is a problem of awareness, confidence, and capability that precedes any change management process.
3. The place-based challenge: MCAs and Investment Zones
England's devolution architecture has created a set of institutions with both the mandate and the incentive to address territorial productivity gaps. Mayoral Combined Authorities now hold significant powers over skills, transport, and economic development. Freeports and Investment Zones offer tax incentives and simplified planning regimes designed to attract high-value activity to specific geographies. Together, these institutions constitute the primary place-based policy infrastructure through which central government ambitions, including AI adoption, are expected to be translated into local economic reality.
The challenge is that the DSIT data reveals a structural mismatch between where AI adoption is occurring and where place-based policy instruments are targeted. Freeports are largely located in port and coastal areas, including Teesside, Humber, Solent, and Freeport East, where the dominant industries are manufacturing, logistics, energy, and construction. The DSIT research shows these are precisely the sectors with the lowest AI adoption rates and the highest rates of non-identification of AI need. The same pattern holds for many Investment Zones, which are concentrated in post-industrial areas seeking to attract advanced manufacturing and life sciences activity: sectors where AI adoption barriers are highest, not lowest.
The policy implication is not that AI is irrelevant to these places. Quite the opposite. Manufacturing, logistics, and construction are among the sectors with the greatest potential productivity gains from AI adoption, in predictive maintenance, supply chain optimisation, autonomous site monitoring, and workforce scheduling. The implication is that realising those gains requires a different model of support than the one currently available. Off-the-shelf AI tools and generalist skills programmes will not reach businesses where the primary barrier is not cost or access to tools, but the absence of a visible, credible connection between AI and the firm's specific operational context.
For Combined Authority mayors and their economic development teams, this creates a specific and tractable mandate: to build the connective tissue between AI capability and sectoral need within their economic geography. That connective tissue has three components. The first is intelligence, a clear and locally grounded picture of where AI adoption is occurring, where the gaps are, and what barriers are specific to the place rather than generic to the technology. The second is a convening and intermediary function that connects businesses to training, tools, and trusted use cases from firms in comparable contexts. The third is institutional leadership, the signal sent when anchor institutions within the geography, including colleges, universities, NHS trusts, and local authorities themselves, demonstrate credible and visible AI adoption.
4. Further and Higher Education as ecosystem anchors
Further Education colleges and universities occupy a structurally distinctive position in the place-based AI adoption challenge. They are simultaneously institutions that must manage their own AI adoption, with all the governance, workforce, and ethical complexity that entails, and institutions whose primary social function is the development of human capability within their catchment geography.
The DSIT research underscores the centrality of skills to the adoption gap. Across all barriers to AI adoption, limited skills and expertise ranks second in frequency (60%) and is the modal barrier for businesses actively planning adoption. When businesses were asked what would most help reduce or remove barriers, training and education was among the top responses: not generic digital literacy, but sector-specific, applied, and contextually grounded capability development.
Further Education colleges and private training organisations are, in principle, uniquely positioned to provide exactly this. Their vocational and technical mission aligns directly with the applied AI capability that employers say they need. Their geographic footprint places them within the economic catchments of Freeport zones, Investment Zones, and Combined Authority skills plans. Their relationships with local employers, through apprenticeship provision, T-level industry placements, and employer advisory boards, provide the trust and contextual knowledge that generic training providers lack.
In practice, however, many FE colleges face the same adoption barriers they are expected to help others overcome: governance structures designed for financial compliance rather than innovation agility; workforce cultures shaped by decades of curriculum stability; senior leadership teams whose primary experience is pedagogical rather than technological. The DSIT research finds that mid-sized businesses, the size category into which many colleges effectively fall when considered as employing organisations, are more likely to cite ethical concerns (32%) and limited skills (64%) as hindrances to wider AI adoption. This may require new training solutions, and new entrants to the market, to address in the short term.
Universities face a different but related challenge. Research-intensive institutions have greater technical capacity and stronger AI literacy among academic staff but frequently struggle to translate that capability into institutional adoption at scale, or to connect their research outputs to the productivity needs of local business communities. The gap between the university as a site of AI research and the university as a model of AI adoption is itself a symptom of the broader challenge: technical sophistication does not automatically produce organisational or territorial change.
What is needed is a change model that accounts for the specific dynamics of institutional ecosystems, and a training model that addresses capability at every level simultaneously, from the principal or vice-chancellor to the curriculum team to the employer partner to the learner.
5. Introducing the PACE Matrix
The PACE Matrix is a diagnostic framework developed in response to the inadequacy of existing models for the place-based AI adoption challenge. It draws on the change management canon described above, on the AI adoption literature, and on the specific empirical picture provided by the DSIT research. It is designed to be used at the level of an institution, a college, a university, an NHS trust, a Combined Authority, a Freeport governance body, and at the level of an ecosystem of institutions within a defined economic geography.
The framework is structured around two axes that the DSIT data and the wider literature converge on as the primary determinants of adoption outcomes.
Institutional Readiness describes the degree to which an organisation has the governance, infrastructure, data capability, and strategic mandate to implement AI in a sustained and accountable way. It is not primarily a measure of technical sophistication. A highly technically sophisticated organisation with fragmented governance and no clear accountability for AI outcomes will score low on Institutional Readiness. A less technically advanced organisation with clear leadership ownership, a coherent data strategy, and board-level accountability will score high. This distinction matters because the DSIT research consistently shows that the barriers perceived as most significant, ethical concerns, unclear regulation, and data complexity, are governance and accountability barriers, not technical ones.
Workforce Orientation describes the degree to which the institution's staff and leadership actively engage with AI as a tool that is relevant, manageable, and beneficial to their work. It is distinct from readiness because an organisation can have excellent governance structures and still have a workforce that is passive, resistant, or disconnected from the adoption agenda. Bridges's concept of psychological transition is directly relevant here: organisations frequently invest in technical readiness without attending to the human transition that must accompany it. The DSIT data shows this pattern clearly; even among businesses currently using AI, only 56% report an increase in employee productivity, suggesting that tool access does not automatically produce behavioural change.
These two axes produce four named states, each with a distinct intervention logic.
Pioneering
The institution has the governance infrastructure to implement AI and a workforce actively engaged with the agenda. The intervention logic is acceleration and knowledge transfer: these institutions should be leveraged as proof points and peer-learning partners within their economic geography. The DSIT finding that organisations in information and communication are most likely to feel ready to scale (76%) identifies a Pioneering cohort whose peer influence has been under-utilised by policy.
Aspiring
Staff and leadership want to move, but the institution cannot yet support them. This is the most common failure mode in FE and HE: enthusiastic champions embedded in institutions whose governance, procurement, and data infrastructure cannot keep pace with their ambition. The risk is burnout and disillusionment. The intervention is governance architecture before further adoption activity, building the institutional infrastructure around the intent that already exists.
Compliant
Systems and governance exist, but staff are not orientated or skilled. Adoption is surface-level: tools are deployed but not embedded; policies exist but are not understood or owned by the people they govern. This is the pattern associated with top-down mandates without accompanying cultural change. The intervention is meaning-making: leadership modelling, internal communication that connects AI to the values of the workforce, and investment in psychological safety around experimentation and error.
Stalled
Neither the governance infrastructure nor the workforce culture supports adoption. This is not a technology problem and is not solvable by deploying better tools. It is the condition the DSIT research captures when it observes that 51% of UK businesses do not see AI as relevant to their organisation. The intervention is foundational: leadership development, change-management capacity, and the deliberate construction of the conditions within which adoption can eventually become possible.
6. An ecosystem model of AI adoption
The PACE Matrix is diagnostic. The ecosystem model that follows from it is prescriptive. Its central claim is that AI adoption in a defined economic geography cannot be achieved institution by institution. It requires a coordinated, multi-level intervention that addresses capability gaps simultaneously across the full range of actors within the ecosystem: anchor institutions, employer partners, supply chain businesses, community organisations, and the individuals at every career stage and role level who will be the ultimate agents of change or resistance.
This claim is grounded in the DSIT research's own evidence on the social nature of adoption. Businesses that had adopted AI cited competitive pressure and the perception that peers were already using it as primary motivators. Businesses that had not adopted cited the absence of visible, trusted use cases in comparable firms as a significant barrier. The adoption dynamic is not primarily driven by the quality of available tools or the clarity of cost-benefit calculations. It is driven by social proof and institutional legitimacy, precisely the mechanisms that place-based anchor institutions are positioned to provide.
Level one: executive and board-level capability
The first and most frequently neglected level of intervention is executive and board-level AI literacy. The DSIT research notes that large businesses are significantly more likely to report senior leadership engagement with AI strategy, and correspondingly more likely to report readiness to scale. The converse, that limited board-level understanding of AI produces risk aversion, governance gaps, and adoption inertia, is equally well evidenced. For FE colleges, universities, Combined Authority executive teams, and Freeport governance boards, the foundational intervention is structured executive development that addresses not technical AI capability but strategic AI literacy: the ability to ask the right questions of AI proposals, to evaluate governance frameworks, to identify ethical risks, and to set the institutional conditions for responsible adoption.
Level two: middle leadership
The middle layer of institutions, heads of department, curriculum managers, service leads, and team managers, is simultaneously the most important and the most overlooked in AI adoption programmes. Rogers's diffusion research and its successors consistently identify this layer as the primary determinant of whether innovations move from early adoption to mainstream use or stall at the enthusiast fringe. It is at this level that adoption either becomes embedded in processes and practices, or remains the preserve of individual champions whose influence cannot survive staff turnover. Intervention here requires applied, role-specific training that connects AI tools to the concrete decisions and responsibilities of middle leaders.
Level three: front-line workforce readiness
The DSIT research shows that the dominant AI use pattern among adopters is off-the-shelf generative tools, specifically natural language processing and text generation (85% of adopters). This is the surface layer of AI adoption: individual staff using AI tools in largely informal, uncoordinated ways, with limited oversight and limited embedding in organisational processes. The intervention at practitioner level is not to regulate or restrict informal use, but to develop the capability for informed, critical, and contextually appropriate use, addressing confidence and judgement alongside tool competence, including how to recognise the limitations of these tools in specific professional contexts.
Level four: employers, SMEs and supply chain businesses
The ecosystem model extends beyond individual institutions to the businesses that constitute the economic geography. For FE colleges, this means the employers who engage with apprenticeship and T-level programmes. For universities, the businesses in their knowledge exchange and innovation networks. For Combined Authorities, the SME base that constitutes the majority of regional employment. The DSIT research reveals a significant asymmetry: large businesses are significantly more likely to have adopted AI than small and micro businesses, and more likely to feel ready to scale. The supply chain dynamics of major employers in Freeport and Investment Zone geographies create an opportunity to accelerate adoption through the value chain.
Level five: community capability
The final and longest-horizon level of the ecosystem model is community and learner capability: the AI literacy of the future workforce and the communities within the economic geography, whether the residents of a place, including those currently in employment, those returning to learning, and those who have not engaged with formal education for decades, have the access, confidence, and capability to participate in an AI-enabled economy. The 80% of UK businesses with no AI plans employ people whose AI literacy is likely to be low. The risk is that AI adoption, like previous waves of technological change, accelerates existing inequalities. Addressing this requires FE colleges and Combined Authorities to take an active role in community AI literacy that extends beyond their immediate institutional remit.
7. From diagnosis to action
The PACE Matrix and the ecosystem model together constitute a framework for place-based AI adoption strategy. The framework provides a basis for sequencing interventions intelligently, addressing the right barriers at the right level for the right institutions at the right time, and for measuring progress against dimensions that matter: not simply the number of businesses that have purchased AI tools, but the quality of governance, the depth of workforce capability, and the credibility of the institutional ecosystem as a whole.
For Mayoral Combined Authority leaders, the immediate practical application is a systematic diagnostic of the institutions within their economic geography: mapping FE colleges, universities, NHS trusts, large employers, and key supply chain businesses against the PACE Matrix to identify where they sit and what the appropriate intervention is for each. For institutions themselves, the framework provides a basis for honest self-assessment and for prioritising the governance and capability investments that will make adoption sustainable rather than episodic.
The DSIT research identifies, with notable clarity, what businesses themselves believe would help: government support in the form of funding, training, and incentives; clearer regulation and industry standards; sector-specific training and upskilling; and tried-and-tested use cases from analogous businesses. These are demands for exactly the intermediary, intelligence, and ecosystem-building functions that place-based advisory practice is positioned to provide.
8. Now is the time to act
The DSIT AI Adoption Research is the most comprehensive empirical account yet available of where UK businesses stand on AI. Its findings are sobering but not pessimistic. They describe a country at the beginning of a long adoption curve, with significant capability concentrated in a narrow band of sectors and geographies, and the majority of businesses yet to engage substantively with AI at all.
The institutions that will determine the AI adoption trajectory of England's economic geographies are not primarily technology companies or central government departments. They are the colleges, universities, Combined Authority teams, and anchor employers and SMEs that have the geographic presence, the community trust, and the sector-specific knowledge to make AI relevant, accessible, and governable for the businesses and people they serve. What happens next depends on whether the institutions with the mandate to close the gap act with the intelligence, the framework, and the results-orientation that the moment requires.
References and notes
- IFF Research and Technopolis Group for DSIT, AI Adoption Research (2026), Figure 1 and regional breakdowns. gov.uk/government/publications/ai-adoption-research.
- DSIT, AI Adoption Research (2026), Barriers to AI adoption chapter.
- Lewin, K. (1951). Field Theory in Social Science. Harper and Row.
- Kotter, J.P. (1996). Leading Change. Harvard Business School Press.
- Bridges, W. (2009). Managing Transitions: Making the Most of Change, 3rd edition. Da Capo Press.
- Satir, V., Banmen, J., Gerber, J. and Gomori, M. (1991). The Satir Model: Family Therapy and Beyond. Science and Behavior Books.
- Rogers, E.M. (2003). Diffusion of Innovations, 5th edition. Free Press.
- Moore, G.A. (2014). Crossing the Chasm, 3rd edition. Harper Business.
- Davis, F.D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3). Venkatesh, V. et al. (2003). User acceptance of information technology: toward a unified view. MIS Quarterly, 27(3).
- DSIT, AI Adoption Research (2026), readiness, budgeting, sector and size breakdowns (Figures 8, 15, 19, 21, 23, 25–27).
Civentis is an advisory, intelligence and ventures practice operating across four areas: Advisory & Strategy, Place-Based AI Adoption, Thought Leadership & Intelligence, and Venture & Enterprise Support. We work with Further and Higher Education institutions, Combined Authorities, NHS bodies, and public sector organisations.
enquiries@civentis.org