Delivery
The infrastructure, the capital, and the ambition are arriving. What is still missing is the capability to convert that into productivity inside the businesses where most Welsh people work. This is the execution layer: forward-deployed AI engineering that works inside Welsh SMEs and is accountable for measured results.
The implementation gap
Wales already has useful building blocks: applied research hubs that have supported hundreds of SMEs, Business Wales for signposting, universities running applied programmes, and an active AI community. These matter. But none of them is an accountable mechanism that can enter a small firm, make sense of how it operates, fix its data foundation, build and embed one high-value AI capability, and then measure whether productivity actually changed.
That is the gap into which good intentions and adoption funding tend to fall. Awareness campaigns, generic training, and short pilots leave little behind. For a ten-person engineering firm in Wrexham or a care provider in Ceredigion, there is a chasm between "here is your readiness assessment and a training catalogue" and "here is a working scheduling assistant connected to your systems, your staff trained, and a measured baseline showing what changed." Closing that chasm is the whole task.
The model
The proposal has two layers, designed to work together. A front door makes the national offer navigable. An engine does the delivery. Most thinking stops at the front door; the engine is what turns a Growth Zone from a set of sites into an outcome for the wider economy.
A single, bilingual national entry point for SME AI adoption, anchored in the Growth Zones. It offers a lightweight diagnostic of readiness and opportunity, a consolidated view of skills and training routes, help navigating funding, and matchmaking across the Welsh provider base and into brokered access to compute. None of this needs new institutions. It wires existing assets into one journey.
Behind the door, small multidisciplinary teams — delivery leadership, data and platform engineering, AI engineering, change support, and graduate associates — work directly inside Welsh SMEs. Each engagement delivers exactly one prioritised, high-value intervention, with a baseline taken before and the same indicators tracked after. One measured change per business, not a sprawling transformation programme.
How an engagement works
Every business goes through the same disciplined path. It is deliberately narrow: find the one change worth making, make it work, prove whether it moved the numbers, and leave the firm able to stand on its own.
Stage 1
Identify firms with committed leadership and a real, measurable problem, and agree how the work will run from the start.
Stage 2
Understand the operating model and score candidate opportunities against a published matrix, so the chosen intervention is the highest-value one.
Stage 3
Fix the data the solution depends on — usually the most underestimated part of SME AI — and design the intervention around how the business actually works.
Stage 4
Build the capability, integrate it with existing systems, and train the staff who will use it every day, so it becomes part of normal operations.
Stage 5
Track the same indicators set at baseline, confirm whether productivity changed, and stabilise the solution so the gains hold.
Stage 6
Hand over with full documentation, or offer light-touch or managed support — the firm's choice. No forced dependency, and a clear exit from the outset.
What makes it different
Other national programmes assess, train, introduce, and signpost. All of that is useful, and Wales should offer it. The difference here is a capability whose output is a proven change inside a real firm — and one that leaves the Welsh economy stronger each time it runs.
Every engagement carries a baseline and a post-implementation outcome report. The question is not "did the firm use AI?" but "what changed in the business because it did?"
A diagnostic only matters if it leads somewhere. Here it feeds directly into a funded delivery pathway, rather than ending as a report on a shelf.
Graduate associates and firm staff build real capability through live delivery inside businesses, not classroom-first courses disconnected from the work.
Delivery runs through an accredited network of Welsh consultancies, boutiques, and universities working to one shared playbook and standard, building local supply.
Access to compute is brokered through the front door as a negotiated community benefit from Growth Zone and hyperscaler commitments, not a separate hurdle.
Each engagement feeds anonymised templates, integration patterns, and sector playbooks back into a shared toolbox, so the cost of helping the next firm falls.
Where it applies
The focus mirrors the sectors where most Welsh firms actually operate, with rural participation and Welsh-language provision built in rather than bolted on. The best interventions are deliberately unglamorous and high-return.
Learning from others
Small countries can move quickly when government, infrastructure, and ecosystem are wired together behind one front door. Luxembourg, comparable to Wales in scale, built a national one-stop shop organised around the company journey and has seen adoption climb well into the majority of organisations. Wales cannot copy that programme, but it can assemble the equivalent from the assets the Growth Zones are creating — and go further, by adding the forward-deployed delivery engine that turns a front door into measured change inside real firms.
How the public interest is protected
Public value is protected through transparency and measurement rather than through adding another institution. The discipline is the point: everything that is funded should produce evidence.
How firms are chosen is set out openly, with sector balance, rural participation, and fairness visible from the start.
Every firm has a baseline and an outcome report, and aggregate results are published — productivity, quality, cost, and workforce effects, not activity counts.
Impact and additionality are assessed independently, so the programme is judged on what genuinely changed rather than on what was promised.
Participation terms, ownership, and a clear exit are agreed up front. Firms can operate the capability themselves, take light support, or buy a managed service.
Any Welsh provider that meets the standard can deliver, keeping the network open, competitive, and rooted in the local economy.
It gives Growth Zones the measure they currently lack: SME productivity diffusion, not just capital spend and headline jobs.
How this fits the framework
Forward-deployed AI engineering is not a separate idea. It is the practical way to meet the tests at the heart of this campaign. It is the delivery mechanism for structuring adoption funding around real SME projects rather than pilots. It operationalises local benefit by routing delivery through accredited Welsh providers and converting data-centre community benefit into SME compute access. And it strengthens the case on grid and energy, because a visible pipeline of AI-adopting firms gives planners a concrete demand signal beyond hyperscale sites.
The route from here is evidence-led: a short co-design phase to confirm standards and selection, then a Pathfinder of around twenty SMEs across both Growth Zone geographies, with published per-cohort results before anything scales. Prove it small, in the open, then grow what works.