Field report W-19 / L4 · sheet /strategy
Corporate StrategyFiled from the territory
80% of staff feel more productive; only 37% of firms see EBIT gains
Eighty percent of employees report higher productivity from AI, yet only 37% of organisations see EBIT impact. The largest obstacles are incentive conflicts, managerial habits and fear of role change.
- Grid
- W-19 / L4
- Surveyed
- November 16, 2026
- Reading
- 6 min read
- Correspondent
- Amara Osei
Field notes
- 80% of employees say AI makes them more productive.
- Only 37% of organisations report an EBIT impact; nearly two in three bank none of the perceived value.
- An aviation solver took 18 months to move from half to almost all recommendations accepted.
- Maersk reported $3.5 billion in operating profit for 2025.
- 69% of executives use AI for less than one hour a week, according to Stanford SIEPR research.
By Amara Osei
The AI productivity-to-profit gap is a governance failure: 80% of employees report greater productivity from artificial intelligence (AI), but only 37% of organisations see an earnings-before-interest-and-taxes (EBIT) impact. McKinsey’s State of AI data frame that split, and nearly two in three organisations have banked none of the value workers feel. The central constraint is organisational willingness, not data availability.
Data can identify the optimal move before an organisation accepts it: an Irish rugby victory shifts advertising, an Italian summer extends a drinks season, and an aircraft waits at the gate. The central governance problem is blunt: a model can recommend a move, but only the bonus plan can approve it.
What turns past decisions into present constraints?
Many binding constraints began as choices rather than rules, then hardened as policies with established owners. Those owners can revise them once the original assumptions no longer hold.
During my work leading a global industry team and later Google’s strategic business transformation team, I met customers that had committed substantial spending without yet seeing value. One large aviation partner, accustomed to leading its industry, had each team defend its own part.
A tenured executive vice president sponsored the programme, backed by millions in multi-year capital and the best teams from both sides. The resulting optimisation solver (software that compares operational choices against constraints) weighed crew, aircraft rotation, passengers and ground control simultaneously.
Only legislation, capacity and safety entered as fixed constraints; every other factor entered as an input. When one team member said, “I just know,” the team returned to the underlying rationale and coded the judgment only if it remained valid for future conditions.
The solver demonstrated value, yet adoption took 18 months to rise from half its recommendations accepted to almost all. Controllers carried responsibility for passengers’ lives, so each accepted recommendation challenged a judgment built over a career.
The airline became one of the most profitable in the world, but expansion across the network stalled on decision hierarchies and fear that AI would overrule decades of service. The binding constraint was human, not computational.
In the week before 16 November 2026, an advisory client described its newly implemented dynamic pricing model (a system that adjusts prices) with pride, then listed the constraints built into it. It had bought a model to find the future and fenced it with the past.
AI does not erase corporate politics; it moves those conflicts from project conversations into business-defining decisions. The analysis can become automated while those politics remain intact.
Why do incentives keep value unbanked?
More than a decade ago at Diageo, I combined point-of-sale sell-out data, customer sentiment, media spending and weather data by brand and country. Ireland’s rugby team victory moved resources toward television and Guinness, while a prolonged Italian sun extended Tanqueray’s season.
Every euro reallocation changed which country manager hit the target they had staffed for, who gained a bonus and who had to tell their family that a summer vacation was at stake. The tool became a lagging indicator (a measure that records outcomes after the event), rather than identifying the value the model could have created.
That incentive structure sits directly between the 80% productivity figure and the 37% EBIT result. People use AI to complete tasks faster and are often asked to do much more with the time freed.
When managers cite AI as the reason for headcount reductions, showing how much it saves can feel like drafting one’s own redundancy. Organisations remove tasks from old roles, while no one is paid to imagine the new ones.
Who owns the next constraint?
Every employee is now an incumbent, confronting a familiar choice: protect the past or take ownership of what comes next. At Maersk in 2016, I encountered the same dilemma at corporate scale.
Margins were thin, value lay in customers’ wider supply chains, and a shipowner had to prove it would choose the carrier best for the customer rather than its own fleet. The industry remained unprofitable in 2016, while integrated logistics (coordinated transport and supply-chain services) later paid off through the pandemic.
Maersk reported $3.5 billion in operating profit for 2025, demonstrating the value of capabilities beyond ship ownership. Leaders must help employees see how the right capabilities contribute to the next business model.
After a session I led recently, an executive assistant approached me following a discussion of AI’s effects on business and individuals. She used AI often, more than her boss knew, and feared she would be replaced.
When I asked where she last felt incredibly valuable and proud, she described bringing a leadership team of strong personalities behind one common goal. She saw each individual and caught fears before colleagues voiced them.
Her role was to create an environment where leaders could lower their guard enough to align. She said that, if an AI model could perform that work, she would train it and use it to scale her contribution without limit.
Her immediate barrier was not technology but knowing her own value. Multiply that individual constraint across a workforce, and the distance between 80% productivity and 37% EBIT impact becomes understandable.
The economy now organises around its next constraint. It cannot be raised, built or bought as before; this time, it is us.
What must leaders change first?
Four moves address the human barrier in advisory work.
- Start from the ideal ambition, not from today. An outcome set too small invites people to optimise the present; an ambition gives them a future position to imagine and work toward.
- Audit every constraint, including those in our heads. Curiosity reveals what drove each rule and whether it still fits; honesty requires naming the narratives people continue to defend.
- Steer the system, not just the budget. Tie part of every target to the shared result. Pay signals how people are coping this year; ownership signals whether they join the next ten. Promotions reveal what the organisation truly values.
- Live the change before asking others to do it. If every saved hour becomes a headcount cut, employees will hide the gain. Stanford Institute for Economic Policy Research (SIEPR) research reports that 69% of executives use AI for less than one hour a week.
That usage leaves leaders executing a transformation they ask others to inhabit while quietly negotiating how much of themselves to change. People feel the shift long before they can voice it.
My question to leaders is direct: “What are you doing differently this week?” Organisations move quickly once their leaders move first. Momentum begins with the first honest move, and leadership must make it.
via fortune.com/feed/ (Source)
More from Amara Osei
Show full bio
Market editor covering people, launches and day-to-day practice in corporate strategy and chief-executive leadership at CEO Frontier.
452 articles
Adjacent sheets
- IBM warns British firms cannot prove their AI is under control
- AI Agents Are Eroding the Moats Built on Consumer Laziness
- Meta Weighs 20% Workforce Cut as AI Infrastructure Costs Climb
- Companies replace blanket AI access with budgets and controls
- Consultancies Pour Billions Into AI as ROI Question Goes Unanswered