Build a More Capable Workforce, Not a Smaller One
Guest Blog: Chris McKibbin, Principal Consultant, Avencera
By the numbers, the untapped potential of agentic AI seems enormous. Screaming headlines proclaim huge efficiencies; in January 2025 the UK government declared a "£45bn jackpot" in annual savings from digitising the public sector, although the Science, Innovation and Technology Committee has since declared that figure to be unreliable. The wider growth claims are bigger still: AI is poised to boost UK GDP by an estimated £550 billion by 2035.
This growth will purportedly also bring staggering job losses, with the Tony Blair Institute (TBI) estimating AI could displace between one and three million UK jobs over the coming decades, raising unemployment by up to 180,000 by 2030.
Such estimates serve a purpose, informing investment decisions and building momentum. But the problem is their reductive nature. The analysis is constrained by the question "does this save us money?"
For anyone involved in automation, the more valuable framing should always be "what could these systems release?"
The greatest opportunity presented by agentic AI is not automation for its own sake, or pure paper efficiency, but the flourishing of human skill, judgement and imagination. People can be freed from tasks that never really deserved human attention. The prize, in every case, is not necessarily a smaller workforce, but a more capable one. To realise this, government must encourage human growth - skills, attitudes and imagination that can, in turn, drive business and economic growth.
In developing policy
The good news for policymakers is that the underpinning feedback loop is already well established: challenge – policy – programmes – evaluation – evolution. The opportunity in adopting AI is to speed that loop up, leveraging better data to make better decisions. We're delighted to see this happening more, like greenfield public consultation synthesis, the Test Learn and Grow team, and the growing importance placed on data and insight in major service delivery procurements. AI can help further still; for example with digital twins and scenario simulation.
In enabling change
Within government operations, senior buy-in and aircover is critical, and the environment has never been more conducive to reimagining how outcomes get delivered. Permanent Secretaries and senior leaders giving their teams licence to "show up differently" can genuinely unleash brilliant people to transform public services. The "Newco" approach is showing huge promise at HMRC, while DSIT's growing influence is evident in every major tender brought to market. The raw ambition of the Customer First team is also admirable. The challenge now is to protect it, and let it build, prove and do.
In commissioning delivery
The strongest gains lie wherever skilled people spend disproportionate time on coordination, paperwork and information-chasing rather than improving the outcomes they universally want.
Three areas are particularly close to our hearts at Avencera:
Education is perhaps the most talked about opportunity space, and the scale of possibility is virtually boundless. But schools remain bounded by real financial and capacity constraints, and so still focus on efficiency. The TBI's Generation Ready analysis found nearly half of schools had not officially implemented AI as of 2025, and even among those experimenting, the rationale is narrow: reducing teacher workload (44%) is the most common motivation. Far smaller shares cite improving pupil outcomes (14%) or supporting assessment (9%) – highlighting both the untapped opportunity to reshape teaching and learning, and the need for policy to push beyond incremental gains.
Employability programmes, including the huge £1.2bn Future Employment Support contract which is now in early market engagement, are part of a broader government drive. Here agentic AI can make delivery more effective: identifying gaps in the market; matching individuals to them; spotting gaps in their skill base; helping define robust, meaningful qualifications to fill those gaps – consistently and far more rapidly than today. Responses to shifting demand can and should happen in days and weeks, not years.
Business support to seed AI adoption at the grassroots is hard, but compelling: small businesses account for over
. Nationally, we've seen mixed success here. The landscape is loud, the technology is moving fast, and business owners need targeted support explicitly relevant to them. Again, the real opportunity is to scale what works – and from our work with the Thames Freeport, we believe that's locally, not nationally, driven.
Across these areas, the open questions AI can help us solve include: how do we better understand individuals, their behaviours and motivations, not just their skills? How do people learn best? And how can we access better, more granular data locally?
What it takes to deliver
None of this lands on enthusiasm alone, and the hard lessons of earlier automation waves still apply.
Three sets of levers need to move together.
Technically, agents work best grounded in authoritative data rather than left to improvise. They need to be integrated carefully into the legacy systems where work happens, and observable enough that we can see what an agent did and why, with a person kept in the loop wherever stakes and regulatory imperatives call for it.
Organisationally, the binding constraint is rarely the technology. It tends to be skills, not just literacy, but the confidence to direct, question and overrule AI well; alongside the will to redesign broken workflows rather than simply make them faster. Trust is equally as important: people embrace tools they believe in and quietly resist those they don't.
On policy, scaling responsibly calls for clear accountability for what agents do, procurement routes that let public bodies adopt fast-moving technology without locking themselves in, and proportionate regulation. Most of all it needs honest measurement, so claims of saving and growth are evidenced rather than asserted.
Handled well, agentic AI can clear away the drudgery that has long crowded out meaningful work: the diagnosis, the teaching, the judgement, the creative leap. The UK's advantage – and the greatest opportunity – will not come from deploying AI fastest, but from deploying it in service of people, to unleash their imagination.
That is an opportunity worth getting right.
