The AI Talent Gap Is a Build Problem, Not a Buy Problem

The conventional response to a talent shortage is to recruit harder. Post more jobs, raise salaries, and wait for the market to deliver. That logic is about to fail a large number of organizations, and the data is specific enough to sting.

The UK alone faces a shortage of 160,000 AI professionals, with Personnel Today (2025) reporting that skills gaps could leave half of all AI roles unfilled by 2028. Simultaneously, demand isn't softening — Personnel Today (2025) cites PwC analysis showing that the number of job postings for specialist AI roles has surged by two-thirds in the past year alone. The gap between supply and demand is not narrowing. It is widening faster than the hiring pipeline can compensate.

This is the structural trap: organizations are competing for a pool of external talent that is categorically insufficient to meet aggregate demand. Every CHRO who believes they can simply out-recruit their way to AI capability is bidding against every other organization for the same 50% of roles that will actually be filled. That is not a talent strategy. That is a lottery ticket.

The internal build case is stronger than it looks

Bank of America has taken a different position. Bernard Hampton, head of the bank's Academy, has described how the learning and development organization is centering its strategy on workforce agility — preparing a massive global workforce for AI not by replacing people, but by reskilling them at scale. The framing from MIT Sloan Management Review (2025) is instructive: the Academy treats learning not as a benefit but as an operational capability, one that must be continuously maintained across a workforce of that size and complexity.

Most organizations are nowhere near that model. A TalentLMS report cited by HR Dive (2025) found that "speed-to-skill" has become a critical metric in a market where skills become obsolete far faster than before — a dynamic that makes the typical annual training cycle look like infrastructure built for a slower era. The question is not whether your current workforce can learn AI-adjacent skills. It is whether your L&D infrastructure is built to deliver that learning at the pace the market now demands.

The Bank of America model is not replicable verbatim across a 200-person company. But the underlying logic is. Workforce agility does not require a dedicated academy; it requires treating skill-building as a continuous operational process rather than a periodic HR program.

The anxiety variable organizations are underestimating

There is a force working against internal build strategies that rarely appears in workforce planning decks. HR Dive (2025) reports that both job seekers and hiring managers are expressing growing concerns about reliance on AI, based on research from Express Employment Professionals. AI anxiety is not just a communications problem for your change management team — it is a drag coefficient on adoption rates, learning engagement, and ultimately the ROI of every upskilling dollar you spend.

If your workforce believes that AI is coming for their roles regardless of how much they learn, their motivation to invest in reskilling collapses. The internal build thesis only holds if employees believe the organization is building with them, not building toward replacing them. That requires a level of organizational transparency that most executive teams are not yet comfortable delivering.

What this means for the board conversation

Senior leaders looking at AI capability gaps tend to frame the problem as a budget question — how much do we need to spend on hiring or tooling? The more clarifying frame is a make-versus-buy decision with a hard constraint: the buy market is insufficient to solve the problem at scale by 2028. That changes the calculus entirely.

The organizations that will have functioning AI-capable workforces in three years are the ones that started treating internal reskilling as a supply chain problem today. Not a feel-good investment in people. A supply chain problem, with lead times, throughput constraints, and quality control requirements.

The external AI talent market is not going to save you. The only question left is whether your L&D function is built to handle that reality — or whether it's still optimized for a world where you could buy your way out of a skills gap.