BusinessMarkets, money & strategy
CEOs take over AI decisions while firm data show about 0.2% annual productivity gain
Nearly three-quarters of CEOs now say they lead their company's AI decisions, double last year's share. Yet a large firm survey finds AI has lifted labour productivity only about 0.2% a year, and agentic tools raise new governance questions.
From the Financial Times · AI article of October 30, 2026 and cited sources: BCG's 2026 AI Radar surveyed 2,360 executives, including 640 CEOs. [2] [9]
SOURCES[2] BCG: As AI Investments Surge, CEOs Take the Lead (press) bcg.com[9] BCG AI Radar 2026 bcg.com
AT Pulse research deskPublished · Updated · 4 min read · AI-assisted analysisAI-assisted, not human-edited. Methodology Corrections Who owns this site
- Original source
- Financial Times · AI article, independent perspective
- Brief by
- AT Pulse research desk AI-assisted; publisher and contact on the desk page
- Studies
- arXiv 2607.07612 (Jul 2026) [7] date from arXiv ID
- When
- October 30, 2026 article published
- Covered
- Oct 30 edition
- Evidence
- 10 sources across 7 sites
- 1AI and Leadershipft.com
- 2BCG: As AI Investments Surge, CEOs Take the Lead (press)bcg.com
- 3NBER WP 34836: Firm Data on AI (RePEc)ideas.repec.org
- 4Outlook Business: Gartner on agentic AI cancellationsoutlookbusiness.com
- 5BCG: Agentic AI Blurs Line Between Tool and Teammatebcg.com
- 6BCG: AI Isn't Keeping CEOs Up at Night, but It Shouldbcg.com
Reach: not measured yet
BCG's 2026 AI Radar surveyed 2,360 executives, including 640 CEOs. It found that 72% of CEOs now call themselves their company's main AI decision-maker, about double the year before. Half believe their job depends on getting AI right. BCG, which sells AI services, also reports that about 90% of CEOs expect agents to deliver measurable returns in 2026. That figure is an expectation, not an outcome. [2] [9]
An NBER survey of nearly 9,000 executives estimates that AI added about 0.2% a year to labour productivity over the past three years, with essentially no net effect on employment. Executives expect about 0.6% a year ahead. They report using AI about 2.1 hours a week. [3]
In an MIT SMR/BCG survey of 2,102 executives, 76% said agentic AI works more like a coworker than a tool, and 45% expect fewer middle-management layers. Gartner forecasts that more than 40% of agentic AI projects will be cancelled by the end of 2027. It cites cost, unclear value and weak risk controls. [5] [8] [4] [10]
Accountability is still loose. BCG says only about a quarter of CEOs face high stakeholder pressure to show near-term value, and it expects boards to start demanding revenue, cost and engagement metrics. The FT argues that agents free executive time for strategy but add governance risk. An academic review argues that agentic systems need their own governance frameworks. [6] [1] [7]
CEOs now own AI decisions, but measured productivity gains remain small and expected returns are largely forecasts.
Read the full assessment
Implication: leaders should define revenue, cost and engagement metrics and redesign roles before boards demand proof and agent projects stall.
Executive brief
Executives say they own AI, but the measured payoff is small. About 72% of CEOs now call themselves their company's main AI decision-maker, double the share a year earlier (BCG). Yet an NBER survey of nearly 9,000 executives finds AI added only about 0.2% a year to labour productivity over the past three years (NBER via RePEc). The FT's 30 October report argues that executives must rethink their roles. It says agentic tools free up time for strategy but carry governance risks (FT). Only the FT's feed excerpt is available, so this dossier relies on surrounding evidence.
What changed and event timeline
Gartner predicts agent project failures
Gartner forecasts that over 40% of agentic AI projects will be cancelled by end-2027. It cites cost, unclear value and weak risk controls ().
Executives call agents "coworkers"
An MIT SMR/BCG survey of 2,102 executives finds 76% see agentic AI as more like a coworker than a tool. 45% expect fewer middle-management layers ().
CEOs take the AI decision
BCG's AI Radar (2,360 executives, 640 CEOs) finds 72% of CEOs lead AI decisions. Half believe their job depends on getting AI right ().
The accountability "grace period"
BCG argues AI's importance has grown faster than the tools used to judge it. Only about 25% of CEOs report high pressure from stakeholders to show near-term value ().
Academic governance review
Raji and Bashir publish a systematic review arguing that agentic systems need their own governance frameworks ().
FT "AI and Leadership" report
The report frames executive roles and skills as needing a rethink. It says agentic tools free time for strategy while raising governance risks ().
Capabilities and access
The reviewed sources name no specific model or version. "Agentic AI" here means systems that plan and carry out multistep tasks on their own (BCG press).
Read the full section
The reviewed sources name no specific model or version. "Agentic AI" here means systems that plan and carry out multistep tasks on their own (BCG press). Adoption is early:
Technical analysis for researchers and developers
- Nearly all of it comes from executive surveys of self-reported attitudes and intentions, not measured agent performance.
- It is a multi-country survey of firms run by central banks and academics (Bloom, Davis and others).
- The arXiv paper is a literature review with no empirical testing. It sets out priorities, mechanisms and the roles of different stakeholders (arXiv).
Read the full section
- Evidence base. Nearly all of it comes from executive surveys of self-reported attitudes and intentions, not measured agent performance.
- NBER method. It is a multi-country survey of firms run by central banks and academics (Bloom, Davis and others). It is stronger on firm-level outcomes than consultancy polls (RePEc).
- Governance research. The arXiv paper is a literature review with no empirical testing. It sets out priorities, mechanisms and the roles of different stakeholders (arXiv).
- What builders need. The sources call for governance to be a cross-functional effort (IT, HR, finance, operations), not an IT add-on (MIT SMR).
Claims and evidence
- Agents free executive time for strategy: this is the FT's framing (FT). There is no independent confirmation that this time goes to strategy.
- About 90% of CEOs expect agents to deliver measurable returns in 2026: this is an expectation reported by BCG, a consultancy that sells AI services, not an outcome (BCG).
- Productivity: the NBER survey finds about 0.2% a year so far and executives expect about 0.6% a year ahead (RePEc).
Read the full section
- Agents free executive time for strategy: this is the FT's framing (FT). Independent data show executives use AI about 2.1 hours a week, up from 1.5 in winter 2026 (RePEc). There is no independent confirmation that this time goes to strategy.
- About 90% of CEOs expect agents to deliver measurable returns in 2026: this is an expectation reported by BCG, a consultancy that sells AI services, not an outcome (BCG).
- Productivity: the NBER survey finds about 0.2% a year so far and executives expect about 0.6% a year ahead (RePEc).
Context and prior work
- This is BCG AI Radar's third year. The CEO share leading AI decisions roughly doubled from the year before (BCG).
- MIT SMR/BCG's ninth annual AI study describes agents as both tool and coworker. It says this dual nature creates tensions in how work, workforce planning and investment are organised (MIT SMR).
Limitations, safety and contested findings
Read the full section
- Optimism vs measured output. BCG reports 80% of CEOs more optimistic about AI returns, but NBER finds little measured productivity gain and essentially no net effect on employment (BCG; RePEc).
- Failure risk. Gartner names inadequate risk controls as one cause of expected cancellations (w.media).
- Governance changes ahead. 58% of agentic AI leaders expect their governance structures to change within three years (BCG press).
Business and practitioner implications
- BCG expects boards to start demanding proof of impact through revenue, cost and engagement metrics (BCG).
- Fewer middle managers and more generalists are expected, so redesigning roles matters as much as buying tools (BCG press).
- Fund agents only where value is clear. That is Gartner's advice, given its cancellation forecast (Outlook Business).
Read the full section
- Set success measures now. BCG expects boards to start demanding proof of impact through revenue, cost and engagement metrics (BCG).
- Plan the org changes. Fewer middle managers and more generalists are expected, so redesigning roles matters as much as buying tools (BCG press).
- Fund agents only where value is clear. That is Gartner's advice, given its cancellation forecast (Outlook Business).
The source trail.
Sources (10)
AI and Leadership
Feed excerpt only.
ft.com