AI & executives: the 2026 state of play

The real issue with AI in business is no longer the technology

The 2026 state of play fits in one sentence: AI adoption is massive, individual gains are proven, organizational gains remain rare — and the variable separating the two is not technological, it is managerial. The studies converge: where the CEO steers governance, where managers champion usage and where processes are redesigned, impact follows. Everywhere else, AI produces saved time that evaporates.

Exact wording, because the nuance matters: organizational gains are not absent — they are rare, concentrated in a minority of companies, and systematically associated with a transformation of management, not with a choice of tool.

Everyone is adopting. Almost no one is banking the gains.

No technology has ever been adopted this fast: McKinsey measures in 2025 that 88% of organizations use AI in at least one function — and that nearly two thirds have yet to start deploying it at scale. Investment is keeping pace: global corporate investment in AI more than doubled in 2025 (Stanford HAI, AI Index 2026).

The returns, however, are not — and this finding runs through every reference survey:

The bluntest synthesis comes from MIT, cited by Gallup's CEO in the State of the Global Workplace 2026: despite roughly $40 billion of corporate investment, 95% of AI pilots produced no measurable impact on the P&L. And an NBER survey of American, British, German and Australian executives, cited in the same report, backs it up: 89% of executives see no effect of AI on their company's labor productivity over the past three years — while anticipating +1.4% for the next three.

Individual gains, for their part, are proven — under controlled conditions

The paradox does not stem from a disappointing technology. Experimental studies — the most methodologically robust — measure clear, reproducible individual gains:

Two caveats from the same literature, worth knowing before extrapolating. First, the “jagged frontier” (Dell'Acqua et al., Harvard/BCG): on a task outside the model's zone of competence, AI-assisted users were 19 points less likely to produce a correct answer — AI amplifies error too. Then aggregation: by late 2024, weekly usage already reached 23% of American workers, yet the time actually saved amounted to about 1.4% of hours worked (NBER, Bick/Blandin/Deming); and the NBER's longitudinal Danish study detects, after two years, no significant effect on wages or hours across the 11 occupations studied. Micro-level gains are real; their macro-level diffusion has not happened yet. A transmission belt is missing.

The variable that separates the two: management

This is the most convergent result of the 2025-2026 corpus — and the least discussed in executive committees. The transmission belt between individual gains and company performance is the management line, and it is today the most weakened link in the organization.

At the top of the organization, McKinsey establishes that CEO oversight of AI governance is one of the attributes most correlated with bottom-line impact — ahead of most technical choices. And its Superagency survey (3,613 employees, 238 executives) overturns the received wisdom: employees are ready; it is leaders who are not steering fast enough. 92% of companies plan to increase their AI investments — 1% of executives consider their deployment mature. Microsoft/LinkedIn's Work Trend Index reaches the same diagnosis from the other end: organizational factors weigh more than twice as much as individual factors in AI's reported impact.

This intersection — a decisive managerial lever, a disengaged managerial population — is precisely where 2026 will be decided. Gallup puts the global cost of disengagement at around $10 trillion in lost productivity, or 9% of world GDP. In the benchmark organizations it studies, 79% of managers are engaged — nearly four times the global average. High managerial performance is not a foregone conclusion of the market; it is an organizational choice.

2026: the year AI lands on the CEO's desk

The steering is changing hands. In 2026, BCG (2,360 executives, including 640 CEOs) measures that 72% of CEOs identify as their organization's primary AI decision-maker — and 50% believe their job is on the line if AI fails to deliver. Budgets are following: AI investments are expected to rise from 0.8% to about 1.7% of revenue in 2026.

The dark side of this acceleration is documented by the same literature: 56% of users acknowledge work errors caused by AI, and 66% rely on its answers without checking their accuracy (KPMG). At Deloitte, the most cited barrier is no longer technical but regulatory and risk-related. Steering AI in 2026 is first a matter of governance and posture — not a choice of tooling.

In France, a quiet revolution — carried by the CEO alone

The French business fabric is moving, but in its own way: cautious, pragmatic, and heavily dependent on the CEO's personal conviction. The reference survey is Bpifrance Le Lab's (1,209 CEOs of SMEs and mid-market companies):

The wide gap between global surveys and the French field comes down to definition and size: INSEE counts 10% of French companies with 10 or more employees using AI in 2024 (6% in 2023) — 9% of those under 50 employees, 33% of those with 250 or more — while France remains below the European average (13%). On individual usage, BCG measures in France 82% regular use among executives versus 45% among employees, 28% of employees who consider themselves adequately trained and 12% receiving clear guidance. And the human backdrop is the most degraded in Europe: 8% of employees engaged in France, versus 12% in Europe and 20% worldwide (Gallup, 2026).

For a mid-market CEO, the reading is direct: AI there is today a matter of the CEO's personal conviction, applied to optimization, with no formalized strategy and no equipped managerial relay. That is exactly the configuration global studies associate with plateauing — and that is where the room for maneuver lies.

What the studies do not say with one voice

An honest state of play also documents the contradictions — they are instructive.

On adoption: the spectacular gaps (88% at McKinsey, 10% at INSEE) come down to definitions — “reported use in at least one function” on one side, effective integration measured across the entire business fabric on the other. Neither figure is wrong; they do not measure the same thing.

On ROI: declarative company surveys (McKinsey, PwC CEO Survey, Deloitte) tell a story of majority non-impact; the PwC AI Jobs Barometer, which observes real-world data (nearly one billion job postings, company accounts), measures on the contrary productivity growth 40% higher in 2026 in the companies most exposed to AI, and revenue per employee growing three times faster in exposed sectors. Both are compatible: the gains exist, but they are captured by a minority — and the sorting happens precisely on the quality of the organizational transformation.

On the time horizon: controlled trials measure immediate gains at the task level; longitudinal labor-market data (Denmark, NBER) do not yet register any effect on wages and hours. The micro→macro lag is a classic pattern of general-purpose technologies — electricity took twenty years.

Methodological caveat: the MIT figure (95% of pilots with no P&L impact) and the NBER executives figure (89%) are reproduced here as cited by the primary Gallup 2026 source in our possession; the original MIT NANDA report is no longer publicly accessible at the time of publication.

Five implications for an Executive Committee

1 · Posture before tooling. AI does not need a CEO who codes; it needs a CEO who decides — on use cases, on governance and on what to do with the time saved.

2 · Governance is steered from the top. CEO oversight is the attribute most correlated with impact. Delegating AI three levels down is choosing the plateau.

3 · The management line is the multiplier. The top predictor of usage is the direct manager's support — yet this is the population whose engagement is collapsing. Equipping and re-engaging managers is not a side HR topic: it is the condition for ROI.

4 · Saved time needs a destination. Without direction on how freed-up time gets used, the individual gain evaporates. The question “to do what?” is a leadership decision, not an afterthought.

5 · Measure, or walk away. Between the 95% of pilots with no effect and the companies capturing real gains, the documented difference is discipline: tight use cases, redesigned processes, tracked impact.

The questions executives ask

What is the AI adoption rate in business in 2026?

It all depends on the definition. 88% of organizations report using AI in at least one function (McKinsey, 2025); but INSEE counts only 10% of French companies with 10 or more employees actually using it in 2024 (33% above 250 employees). The gap comes down to the definition used and to company size.

Does AI really increase productivity?

At the individual level, yes — it is measured: +12 to +35% according to controlled studies (Harvard/BCG, NBER). At the company level, rarely: more than 80% of organizations see no tangible impact on EBIT (McKinsey, 2025). The difference plays out in governance, process redesign and managerial sponsorship.

Why do most AI projects fail?

Not because of the technology. The documented factors are organizational: no steering from the top, processes left unredesigned, managers left unsupported. The top predictor of an employee's AI usage is the active support of their direct manager (Gallup, 2026).

What is the CEO's role in AI transformation?

Central and measurable: CEO oversight of AI governance is one of the attributes most correlated with bottom-line impact (McKinsey, 2025); 72% of CEOs now identify as their organization's primary AI decision-maker (BCG, 2026); in French SMEs and mid-market companies, 73% of projects are driven by the CEO personally (Bpifrance, 2025).

Where do French mid-market companies stand on AI?

58% of SME and mid-market CEOs see AI as a matter of survival within 3-5 years, but 57% have no formalized strategy and 94% of use cases serve to optimize what already exists rather than grow the business (Bpifrance Le Lab, 2025). The room for progress lies in moving from tool to strategy.

Primary sources and method

Method: this report is a meta-synthesis of studies published between 2024 and 2026 — no primary data collection. Each figure is tied to its original source, with the year and, where available, the sample. The data were cross-checked across several documentary search engines, then verified on the publishers' pages or PDFs. Two figures (MIT, NBER executives) are cited via the Gallup 2026 report, their primary source no longer being publicly accessible.

1. McKinsey — The State of AI (2025, n=1,491, 101 countries)
2. McKinsey — Superagency in the Workplace (2025, 3,613 employees + 238 executives)
3. BCG — AI Radar: Closing the AI Impact Gap (2025, >1,800 C-suite)
4. BCG — AI at Work (2025, >10,600 respondents, 11 countries)
5. BCG — As AI Investments Surge, CEOs Take the Lead (2026, 2,360 executives including 640 CEOs)
6. Accenture — Pulse of Change (2026, n=3,650 executives, 20 countries)
7. Deloitte — State of Generative AI in the Enterprise (2024-2026, n=2,773)
8. PwC — 29th Global CEO Survey (2026, 4,454 CEOs, 95 countries)
9. PwC — AI Jobs Barometer (2025-2026, ~1 billion job postings)
10. Gallup — State of the Global Workplace 2026 (263,810 respondents)
11. KPMG / University of Melbourne — Trust, Attitudes and Use of AI (2025, 48,340 respondents, 47 countries)
12. NBER — Generative AI at Work (2024, 5,179 agents)
13. NBER — The Rapid Adoption of Generative AI (Bick, Blandin, Deming, 2024)
14. Harvard/BCG — Navigating the Jagged Technological Frontier (Dell'Acqua et al., 758 knowledge workers)
15. Stanford HAI — AI Index 2026
16. Microsoft / LinkedIn — Work Trend Index (31,000 respondents, 31 countries)
17. Bpifrance Le Lab — AI in French SMEs and mid-market companies (“L'IA dans les PME et ETI françaises”) (2025, 1,209 CEOs)
18. France Num Barometer (2025, 11,021 companies)
19. INSEE — Artificial intelligence in French companies (2025)

By Ludovic Baumgartner — executive coach, executive sparring partner, founder of Wegartner. He led before he coached: twenty years in business, a seat on an Executive Committee, a P&L he owned. Full profile · the founder. Published and updated on July 9, 2026.