Most organisations are solving the wrong AI problem
- aiomniconversation
- 4 days ago
- 5 min read
Boards are approving AI budgets. Chief executives are citing AI in every earnings call. Productivity targets are being set. And yet, across most organisations, the gap between what AI can do and what the workforce can actually deliver with it keeps growing. The bottleneck is not the technology. It never was.
I had a conversation recently on The AI Adoption Podcast with Euan Blair, Founder and CEO of Multiverse, an organisation that works with businesses and public sector bodies to close the AI skills gap through contextualised, on-the-job training. Euan speaks with five to ten chief executives, chief information officers, and chief operating officers every week. The picture they describe is consistent: they understand the opportunity. They are struggling with the reality.
Billions into technology. Almost nothing into people.
The investment landscape tells a revealing story. Billions of pounds have gone into large language models (AI systems trained on vast amounts of text to generate human-like responses). Billions more are flowing into data centres and compute capacity. Nearly every major AI company is well funded. The adoption layer, meaning the capability of leaders, managers, and frontline workers to use these tools effectively, is being left behind.
Euan draws a sharp analogy. Much of what organisations are doing with AI today is the equivalent of owning an smart phone and using it only to make phone calls. The capability exists. The training to unlock it does not. The OBR has estimated that AI could add tens of billions to UK GDP over the coming years. Ken Griffin has suggested the adoption cycle could stretch to 25 or 30 years if workforce training is not treated as a genuine priority. That gap between potential and reality is not a technology problem. It is a training problem.
The consequences are already visible in specific sectors. Entry-level roles in accountancy and law are shrinking. The 40 to 49 age bracket, currently the most productive demographic in the economy, is among the most overlooked when it comes to reskilling. Workers over 50 who lose their jobs face significantly lower chances of re-employment. Euan is direct about the risk: if this is not addressed seriously, the societal consequences will be worse than those seen during deindustrialisation in the 1980s.
[LINK SUGGESTION: Olivia Larkin https://open.spotify.com/episode/0L0vMTxAPsEP3rf0bZvKje?si=0j2_QANUSUKzbof1pgwkEg
Generic training is not the answer
The reflex response from many organisations is to commission a training programme. They purchase licences for an AI tool, send communications encouraging staff to experiment, and report back that a high percentage of employees have logged in at least once. Euan is unsparing about this approach. It is AI as a toy, not AI as a capability.
The challenge is that different workers face entirely different problems. Equipping a nurse with AI skills requires understanding the pressures of patient triage. Equipping a factory worker requires understanding production line optimisation. A council worker dealing with housing cases needs something different again. Generic training addresses none of these contexts. It produces engagement metrics. It does not produce outcomes.
The evidence from Multiverse's work makes this concrete. A purchasing assistant at a retailer built an automated stock-out alert that saves £150,000 in lost revenue. A council worker built an AI triage process that identifies invalid eviction notices in seconds, preventing homelessness and reducing the council's administrative burden. An NHS worker at the Royal Free Hospital digitised the patient journey, doubling the daily department caseload and cutting patient waiting times by a third. None of these individuals works in a technology role. All of them were trained on specific, contextualised use cases tied to their actual responsibilities.
Euan's conclusion is unambiguous: generic training will always be insufficient. On-the-job learning, built around real roles and real outcomes, is the only route to productivity gains that organisations can actually measure.
Boards are asking the right question but measuring the wrong things
Most boards now understand that AI capability is a strategic imperative. My concern, reinforced by Euan's observations, is that awareness is not translating into action at the right level. Handing the challenge to a learning and development team and hoping for business-wide impact is a structural error. It requires business leaders, HR, and learning functions to work together on solutions that are as beneficial to the employee as they are to the employer.
The measurement problem compounds this. Many organisations have settled on hours saved as their primary metric for AI adoption. Euan acknowledges that hours saved is a reasonable starting point, but argues it is too blunt a measure on its own. The more useful frame is to translate activity into one of three buckets: cost avoided, cost reduction, or revenue generation. The specific metric will differ by company and by team, but the discipline of connecting AI activity to tangible business outcomes is what separates organisations that are genuinely building capability from those that are performing it.
There is also a challenge that boards rarely discuss openly. The workforce will bifurcate between those with genuine AI skills and those without. Organisations will face the same divide. Those with an AI-capable workforce will pull ahead. Those that have substituted licence purchases for real training will find the gap increasingly difficult to close. The question for any board is not whether to invest in AI skills. It is whether they will act before the divide becomes irreversible.
The window to act is narrower than it looks
The pace of AI capability development is accelerating faster than at any previous point in the technology's history. Euan's observation, one I share, is that in Europe and Britain specifically, AI adoption will be defined less by the most advanced model capabilities and more by what organisations can practically achieve within their own operating constraints: their data policies, their regulatory environment, their sector-specific requirements.
That context-dependency makes the case for investment in training stronger, not weaker. It also makes the case for urgency. Workers who develop genuine AI capability become demonstrably more valuable to their organisations. Workers who do not face material displacement risk. The right to reskill, as Euan frames it, may be the most important workers' right of the coming decade.
For those of us working with organisations on AI adoption, the message is clear. Stop measuring adoption by the number of licences activated. Stop outsourcing the challenge to a single team. Start connecting AI training directly to the outcomes the business needs to deliver. The organisations that do this well will not just be more productive. They will be better places to work, better at retaining talent, and better positioned to grow.
The technology is not waiting. Neither should we.
Listen to the full conversation with Euan Blair on
Spotify: open.spotify.com/show/296zibtjU3w4ANuUsPSu2D and
Comments