The right way to adopt AI has not changed, despite the confusing news.September’s reports of high adoption with low strategic benefit are not a surprise. They are the predictable result of firms rolling out tools without redesigning the work first. The fundamentals still apply. AI changes tasks, then roles, then teams. All the talk of worker anxiety is noise. The real work remains the same methodical, task level redesign. We would advise focusing on one or two roles and getting the sequence right.
Employed people in eurozone using AI at work in 2026, ECB Consumer Expectations Survey
ecb.europa.euAnnual out-of-pocket spending by UK workers on generative AI tools for work, Deloitte UK (2026)
deloitte.comShare of UK working adults (18-70) who say they use generative AI at work, Deloitte UK survey (2026)
deloitte.comIncrease in competitive performance for companies with well-prepared leaders and high AI maturity (vs peers without such leadership advantage)
bcg.com4 briefings this month, with 3 new figures that passed our source checks.
subscribers read every briefing in full, and get each one on WhatsApp. subscribe free · how we research and check sources
| stage | what changes | what to plan |
|---|---|---|
| Tasks | Drafting, searching and checking move to AI | Which tasks move, and how quality is checked |
| Roles | Time shifts to judgement, review and relationships | Role redesign and reskilling paths |
| Teams | Structures and spans of control change | Capacity, headcount and how the gains are used |
Awareness training does not change how work gets done. Capability comes from people using AI on their own live work, with playbooks they keep and roles redesigned around the new way of working.
Workforce transition should be planned with the same rigour as the technology: which tasks move, which roles change, and what support and reskilling people get along the way.
Scaling AI is an operating-model exercise more than a technology one. Hand-offs move, decision rights shift and new roles appear, such as agent supervision, so the workflow, the role descriptions and the career paths have to be rewritten together.
how we help: education. →our frameworks: the Lumo method · the value framework
AI replaces tasks faster than it replaces whole roles. Most roles will change shape, with some reduced and new ones created. Organisations that redesign roles deliberately see less disruption and keep more of the gains.
Train by level. Leaders need to set direction and judge spend, teams need hands-on practice in their own workflows, and everyone needs practical literacy and clear guardrails. Tie it to real work, not certificates.
Because tools were deployed without changing the work. People default to old habits unless workflows, targets and expectations change too, and unless they have time and support to learn.
The tasks and roles affected, the new skills needed, the reskilling path, how productivity gains will be used, and how the change is communicated. It should be owned jointly by the business and HR.
Workflows designed for people and AI working together from the start rather than retrofitted, explicit decision rights for what AI can do alone and what needs a person, continuous data and feedback loops, governance built into delivery rather than run as a separate gate, and workforce changes planned in waves with funded transition support.
All of them, which is why it cannot be delegated to an AI council. The CFO reallocates capital, the CHRO redesigns the workforce plan, the CIO makes platform calls that keep options open, and the COO redesigns the operating cadence. Only the CEO can settle the trade-offs between them.