Executives put the spotlight on AI’s reliability issue

A majority of business leaders don't trust AI to work reliably. Only 35% say their AI systems consistently deliver business outcomes, pass regulatory scrutiny, and remain under organisational control at the same time, according to research from HFS Research and TCS.

This gap matters because it exposes a central tension in how organisations deploy AI right now. They're rushing implementation without the governance infrastructure to match. Regulators are tightening requirements – the EU AI Act, sector-specific rules, proposed US frameworks – but most boards haven't aligned their AI strategy with those emerging standards.

The finding also flags a credibility problem. If two-thirds of executives admit their AI doesn't reliably hit all three pillars – performance, compliance, control – then either their AI programmes are genuinely fragile, or their governance and monitoring are weak enough that they can't verify it works. Neither is reassuring.

For ESG teams, this signals a wider issue: AI governance sits outside traditional compliance lanes. It's not purely IT risk, not purely legal, not purely operational risk. It needs board-level ownership and real-time transparency on model performance, training data bias, and regulatory alignment. Most organisations don't have that yet.

The question isn't whether AI works. It's whether boards can prove it works within the constraints their stakeholders – regulators, investors, customers – increasingly demand.