Nine Industries, One Set of Constraints: London 2026 Themes

"Ten thousand engineers in a city" does not mean ten thousand applicants. That line surfaced in the talent intelligence discussion at our London conference on 17 June, and it was the one I kept returning to all day. The point behind it was that the report is rarely the problem. Leaders do not grasp what talent intelligence can do for them, and analysts hand over findings without knowing how to have the conversation that would make them matter.

The day ran to thirteen sessions at The Savoy with more than fifty talent, recruitment and workforce leaders. The room drew from a multinational FMCG company, a generics pharma company, an industrial distributor, an industrial conglomerate, a UK banking group, an energy engineering firm, an industrial-tech manufacturer, a capital-markets operator, a payments company and a biopharma company. What follows is not a recap. It is the read-across — the patterns that surfaced repeatedly in businesses that have almost nothing else in common.

Three lenses organise most of what was said: talent intelligence, recruitment, and work design. Sources here are referenced by industry rather than by name, which is how the synthesis was written.

Talent intelligence is trying to move upstream

The clearest shift is in mandate. Talent intelligence teams are no longer content to help recruiters recruit; they want to shape business strategy. The recurring failure mode, described by the multinational FMCG business, is being pulled in too late — invited to the table only once a hiring problem already exists, at which point the available moves are tactical.

The advice from those who have made progress was consistent and modest. Pilot with one supportive business unit, prove value there, and let the advocates tell the story on your behalf. Understand where HR planning starts and show up there with the right insight, rather than waiting to be asked. One team named a specific disconnect worth fixing early: the peer companies you benchmark pay against should be the same companies you source talent from, and in most organisations those two lists are maintained separately by people who never speak.

Maturity is lower than the ambition suggests. Teams are small, refreshes happen every couple of months, and organisational awareness of the function is thin. The stated ambition is continuous, living intelligence rather than periodic reports. That is also the direction we are building in: generative, self-serve interfaces that put intelligence in a practitioner's hands without query skills, dynamically generated dashboards, and living documents that refresh themselves.

Location strategy was the most-cited single problem

Across a payments company and a biopharma company, location strategy came up as the number-one problem, and the fix described was structural rather than analytical. Start from talent access, then layer cost, real estate and diversity onto it — in that order — and govern the decision through a cross-functional committee. The word used for it was a unified model, and the sequencing is the substance of it: what you put first is what the decision ends up optimising for.

The argument for involving talent intelligence early was commercial. Early involvement protects negotiating leverage with sites and authorities, and can win millions in tax incentives. It is a rare instance of this function pointing at a number a CFO already recognises.

Recruitment is being rebuilt for a world it was not designed for

The industrial-tech manufacturer put it most bluntly: the world TA was built for no longer exists. The move is from post-and-pray fulfilment to intelligence-led hiring, and from asking how many people we need to asking what capability we need.

The most immediate operational problem was AI on both sides of the funnel. Candidates volume-apply using AI while TA screens using AI, and fraud and authenticity have become central concerns rather than edge cases. The response leaders described is deliberately human: referrals, proactive outreach, and verification treated as a judgment task nobody delegates to a model.

That reshapes the recruiter's capability profile. Data interpretation, deep sourcing, business acumen, AI fluency, change leadership and human skills came up repeatedly as the new baseline, with the human share of the work moving toward premium hiring. Internal mobility was framed as a mandate rather than a programme: redeploy rather than replace, with skills architecture taking over the job headcount planning used to do.

Two numbers stuck. Citing a May 2026 study of 350 executives, the industrial-tech manufacturer noted that around 80% of companies funding technology investment through layoffs miss their expected return; the returns come from investing in skills, roles and operating models instead. And the biopharma company reported saving close to $8M in the first year after moving off RPOs, starting with executive search. Four forces were named as demanding a response now: skills-based hiring as the dominant paradigm, pay-transparency legislation, candidate experience as a brand risk, and fast-converting TA functions. The standing request from that group was to get TA a seat at the workforce-planning table and stop measuring it on speed alone.

Work design was the newest idea in the room

Work design drew the most unfinished thinking, which is usually a sign that a theme is real. The starting proposition is to understand the work before the org chart, and the supporting evidence is uncomfortable: job descriptions capture under 60% of the tasks people actually perform. You hire because work needs doing, so a plan built on documents that miss more than two-fifths of that work is built on sand.

The method discussed was to decompose roles into tasks and score each task for automation, augmentation or human ownership. The banking group was firm that judgment layers such as credit risk stay human, and added a point that gets lost in efficiency arguments: customers sometimes want a human even when automation is demonstrably faster. Skills were split three ways — sunset, critical or human, and emerging or sunrise — mapped by function, role and business unit so reskilling lands where the organisation is actually trailing.

The framing that landed hardest came from the bank and the capital-markets operator: perhaps 10 to 15% of jobs disappear, but 100% of them change. The real questions are which skills, by when, and how fast you can build them. Both observed that work-design thinking currently lags AI deployment. The related complaint was job architecture itself. Workday job families serve compensation and recruiting well enough, but they go stale — the capital-markets operator was still working from 2013 data it described, accurately, as dinosaur. A dynamic work architecture lets you redesign continuously, and capability maps do more than siloed career paths.

Regulation, fear, and the argument for a single source of truth

Three threads cut across every session. The first is regulation: the EU AI Act treats hiring as high-risk, which means explainability, auditability and the ability to justify a rejection are now table stakes rather than roadmap items. Fairness-by-design, counterfactual testing and a zero-trust, zero-retention data architecture are the responses we discussed.

The second is narrative. The consistent reframe from the pharma, capital-markets and biopharma leaders was human amplification rather than replacement — taking out the tasks people never wanted in the first place. They were equally clear that this needs active change management, because employees carry two distinct fears: being replaced, and being made invisible. Sense-making was the word used for addressing both.

The third is data. Fragmented workforce data breaks AI systems, and the manual reality behind the slideware is still giant Excel files. The consolidation case is integration plus interoperable, headless access — running your own vendor's agents, whether Microsoft, IBM or Salesforce, on top of one trusted set of workforce data rather than accepting lock-in.

Where a leader should start

If your talent intelligence function is still called in after the hiring problem appears, change that first, and do it through a pilot with a willing business unit rather than a reorganisation. If your job architecture predates your AI programme by a decade, treat refreshing it as prerequisite work, not as a tidy-up. And if you are measuring TA on time-to-fill alone, you are measuring the part of the job that is being automated fastest.

The pattern I take from the day is that the leaders furthest along were not the ones with the most sophisticated technology. They were the ones who had done the unglamorous work first: fixing the data, rewriting the architecture, and learning to have the conversation that turns an analysis into a decision.