A joint study by QualiZeal and HFS Research finds that enterprise AI adoption is moving faster than the frameworks and accountability needed to validate AI systems in production.

The gap is clear in deployment: 57% of surveyed enterprises run generative AI in production and 36% have agentic systems live, yet only 15% use a purpose-built AI testing approach; 52% test probabilistic AI with frameworks designed for deterministic software.

While 88% measure model accuracy, far fewer assess other risks: 47% measure bias and fairness, 37% evaluate explainability, and 35% can provide evidence of regulatory compliance.

Accountability is also missing at many companies, with 37% running AI in production without a named function responsible for its quality.

The study says Quality Engineering is evolving into “trust engineering,” with its remit expanding from software functionality to AI behavior, fairness, explainability, and compliance; 82% expect the QE role to grow significantly.

But the expanded mandate is not matched by funding: only 24% of enterprises say investment in the role has increased proportionally.

QualiZeal and HFS Research call for clear ownership of AI quality, validation methods suited to probabilistic systems, and budgets that reflect QE’s broader responsibilities.