Boston Consulting Group · Survey
The Formula for Agentic AI Value (Applied AI Index 2026)
Almost half of companies now get some value from AI. The new bottleneck is governing agents, not proving ROI.
- Published
- Reading time
- About 20 min
- Authors
- Nicolas de Bellefonds, Vladimir Lukić, Amanda Luther, et al. (18 BCG authors)
What is it about?
BCG surveyed 1,330 CxOs and senior leaders across more than 20 sectors. It classifies 7.5% of companies as “future-built” and 41% as “scaling”, so almost half now generate at least some value from AI. Its formula is strategic clarity plus applied AI: companies strong on both report five times the AI value of companies with neither.
The sharpest finding is about control. BCG expects agentic AI to deliver close to 40% of AI value by 2030, and 42% of companies expect to give agents real autonomy by then. Only 5% have all six recommended controls in place today: governed data and memory access, human-in-the-loop for high-risk decisions, testing with shut-off and rollback, hand-off interfaces, a named owner for every agent, and ongoing security.
Key numbers
- AI spending reached 3.3% of revenue, and about 80% of it sits outside the enterprise IT budget.
- 42% expect autonomous agents by 2030; 5% have all six controls today.
- Companies with all six controls get three times the agentic value of those with only one.
- Companies that track AI value directly in the P&L realize three times the value of those that don't (3.6% vs 1.2%).
- Future-built firms are 3.5× more likely to fund AI as a single multiyear program (61% vs 17%).
Why read it
- It gives you a concrete control checklist for agents that risk, security and audit teams can sign off.
- It reframes governance as an accelerator, which helps when control functions are seen as blockers.
- The “80% of AI spend is outside IT” number tells a CIO how much AI they are not seeing, and who to partner with.
Who should read it
- CIOs and AI leads designing an agent governance model
- CIOs who suspect they do not see the full AI budget
- Risk, security and internal audit leaders
About the author: Nicolas de Bellefonds
Managing Director & Senior Partner, BCG; leads AI at BCG X
An industrial engineer from École Centrale Paris who started in financial audit at Ernst & Young and at a wireless-device start-up before joining BCG, where he later ran its data-science unit, BCG GAMMA. He describes that work growing from problems that “did not fit into an Excel” into a global organization of about 3,000 people.
He is known for blunt advice: “If you see a roadmap with 200 use cases, walk away.” He was the first-listed author of BCG's 2024 study “Where's the Value in AI?”, which found only a small minority of companies creating substantial value.
Read it carefully: common misquotes
- “Almost half generate value” includes the 41% who get only some value. Significant value is the 7.5%.
- The 2025 and 2026 figures come from different samples, not from tracking the same companies.
- BCG describes spending growth in two ways (doubled in one year, tripled in two). Quote one, with its baseline.
How it shapes the CIO's AI Playbook
- Step 1.1: Find all AI spend, including what sits outside IT · Diagnose before you build
- Step 1.4: Inventory pilots and stop most of them · Diagnose before you build
- Step 2.1: Pick three to five domains, not a use-case list · Choose where to play
- Step 3.5: Budget for people, not just models · Redesign the work, not the task
- Step 4.1: Install the six controls before granting autonomy · Govern agents like a workforce
- Step 5.1: Measure outcomes against a non-AI baseline · Measure what the P&L sees
- Step 6.1: Run AI as one multiyear program · Fund and scale
Notes by Pavel Veselov, last reviewed . Summaries are my own; quotes and figures belong to the original authors.
All 11 sources in the CIO's AI Library