LRQA has joined forces with Coval, a specialist in AI simulation and evaluation, and Krew, a provider of AI agents for financial services, to pilot a more practical way to assure agentic AI systems in the real world.
The project focused on Krew’s AI-powered voice agent – technology that interacts directly with customers in sensitive financial situations. To build a complete picture of risk, LRQA and Coval assessed the system from complementary angles, with LRQA focusing on how it is governed, and Coval simulating and evaluating how it behaves in practice.
5,000 simulations. Single digit basis point error rates.
As part of the pilot, LRQA conducted an independent limited-assurance review of Krew’s AI governance framework – an external check that appropriate controls and safeguards were in place, including risk management, fairness and oversight. In parallel, Coval ran more than 5,000 simulations, stress-testing the AI agent across a wide range of consumer engagement scenarios.
Together, the results pointed to strong performance across both governance and simulated testing. LRQA confirmed that Krew’s governance framework meets recognised best-practice principles, with any identified findings addressed, while Coval’s testing showed 100% pass rate across compliance-critical metrics and low single-digit basis point error rates.
The market is moving fast
As AI adoption accelerates, particularly in customer-facing roles, expectations are rising just as quickly. Large organisations are no longer buying AI capability alone, they are looking for visible governance, clear accountability and evidence that systems perform responsibly in practice.
At the same time, frameworks such as ISO/IEC 42001 are beginning to formalise what ‘good’ looks like, raising the bar for providers across the market.
Voices from the team
Ian Spaulding, CEO of LRQA, said:
“We believe this project with Coval and Krew represents a genuine first in the AI space. Together, our teams have combined market-leading technology with human expertise to deliver real value for clients and greater assurance to the wider market. By bringing together independent auditing and technical evaluation, we’ve been able to provide a more comprehensive assessment – one that we believe can act as a blueprint for organisations developing and deploying AI voice agents.”
Brooke Hopkins, Founder & CEO at Coval, said:
“Understanding how AI agents behave in real-world scenarios is critical, particularly in high-stakes, customer-facing environments like financial services. Simulation at scale allows us to test, stress and capture behaviours that are difficult to capture through design or manual QA alone. When combined with governance assurance, it creates something so much deeper and more dependable.”
Michael Goh, Cofounder at Krew, said:
“As a market leader in AI-powered servicing and collections, Krew is proud to be the first and only provider in the servicing and collections industry to have had our AI voice agents externally audited. Coval and LRQA’s external audit sets a new standard for trust, transparency, and accountability for AI in financial services. We are excited to leverage our audited AI platform and proven operational expertise to continue supporting Fortune 1000 firms, federal student loan servicers, and other organizations operating in highly regulated environments.”
What comes next?
This pilot points to where the market is heading.
It shows that voice AI systems can be assessed not only on how they are designed, but on how they perform in practice. By combining LRQA’s governance assurance with Coval’s real-world simulation and evaluation, market-leading technology with human expertise, auditing with technical evaluation – enterprises can have a more complete and credible view of risk.
As expectations continue to evolve, this kind of integrated approach is likely to become increasingly important, not as an added layer, but as a core part of how agentic voice AI systems are validated and trusted.
