Career Development
AI Isn’t the Problem. Your Knowledge Layer Is.
56% of CEOs report no revenue or cost gains from AI. The problem may not be the model—it may be the missing enterprise knowledge layer beneath it.

PwC just published its 29th Global CEO Survey — 4,454 CEOs, 95 countries and territories. The AI number getting quoted everywhere: 56% say they’ve seen neither higher revenue nor lower costs from AI in the past year.
Read that as “AI doesn’t work” and you’ll miss what PwC actually found. Only 12% of CEOs got both revenue gains and cost reductions. What separated them wasn’t the vendor — it was AI foundations: Responsible AI frameworks and technology built for enterprise-wide integration. That group was two to three times more likely to have embedded AI across products, services, demand generation, and strategic decisions, and three times more likely to report meaningful returns.
Here’s my read, past what the survey says. Those “foundations” have a name: the enterprise knowledge layer. A written record of how the company actually operates, in a form software can read — which source is authoritative, who approved it, which version is current, and who’s allowed to act on it in a given context.
Most companies don’t have a trusted enterprise knowledge layer. Ask an agent what the discount approval threshold is, or which severance terms apply to someone hired in 2022, and it pulls from whatever mix of old email threads, outdated docs, and half-updated wikis it can find. It hands back an answer that reads well and traces to nothing. The business can’t verify the result or reproduce it, so the pilot gets shelved — not because the model was wrong, but because there was no governed ground truth underneath it.
A new employee solves that gap by asking around until they find the person who knows. An agent has no one to ask.
Our answer: we spent a few weeks ingesting our own enterprise knowledge layer — policies, business logic, definitions, and more — into a form our agents can read. It’s the reason what comes out for humans and agents is consistent and something we can trust, verify, source, and govern.
If your AI spend hasn’t moved the numbers, I’d check your enterprise knowledge layer before I checked the model. Ask your AI what your promotion criteria are, then ask a manager. See how far apart the answers are.
