Resaro, Partnership on AI further AI quality assurance discussion at ATxSummit

Organisations are now applying structured, evidence-based methods to verify that AI systems perform as required. Comprehensive AI quality assessment is now a necessary complement to trustworthiness and mitigating AI risks. To advance this further, Resaro and Partnership on AI (PAI) hosted a closed-door working session at ATxSingapore that
focused on the key questions at the centre of AI assurance: how to define, measure, and verify that an AI system is actually fit for deployment, and how to decide which AI solution is best suited in a given context.

Source: Resaro. Fireside conversation between Singapore Minister of State (MoS) for Digital Development and Information, Jasmin Lau (centre), PAI CEO Rebecca Finlay (right) and Resaro Co-CEO April Chin (left).

A fireside conversation with Singapore Minister of State (MoS) for Digital Development and Information, Jasmin Lau, PAI CEO Rebecca Finlay and Resaro Co-CEO April Chin was a session highlight. MoS Lau encouraged the cross-pollination of ideas among early AI adopters, emphasising the importance of knowledge-sharing beyond sector boundaries, where diverse perspectives can help reframe familiar problems.

The event also hosted the publication of the 4th report on strengthening the AI assurance ecosystem from Resaro and PAI. Titled Pathways for Operationalising AI Assurance, the report responds to a common challenge: Organisations deploying AI today typically base procurement decisions on vendor-supplied benchmark scores and product demonstrations, even if these signals do not reliably predict how a system will perform under real operational conditions. 

The report argues this is a structural feature of the current AI market, not a failure of individual due diligence: no independent, standardised quality framework has existed for AI, equivalent to the metrics in more mature sectors (e.g. crash-test ratings, engine power, fuel consumption, boot space etc. when deciding which car to procure). 

The report identifies six reasons why closing this gap has proven difficult: 

- The absence of a shared language for quality across technical, operational, and governance stakeholders

- The challenge of translating abstract principles into measurable criteria

- The systematic divergence between benchmark performance and deployment performance

- The cost and limited scalability of bespoke evaluation

- The context-dependence of what counts as "good enough"

- A persistent tendency to treat trustworthiness and performance as competing rather than complementary properties 

The report presents the AI Solutions Quality Index (ASQI) methodology as a practical approach to these problems. ASQI structures AI quality assessment around a set of orthogonal, use-case-specific quality indicators, scored on a five-level scale rather than a pass-or-fail binary. The framework is designed to be automatable and repeatable, to produce results meaningful to non-technical decision-makers, and to support ongoing assessment rather than one-time certification. 

Case studies in the report cover deployments in public administration, public safety, and legal services (also in connection with Singapore's IMDA Global AI Assurance Pilot). Across all cases, the framework produced structured, evidence-based assessments that vendor benchmarks could not: a clear basis for deployment decisions, and specific direction on where systems needed improvement before they were ready for use.

Strengthening the AI Assurance Ecosystem is a multiphase initiative by Partnership on AI and Resaro examining what independent AI assurance requires in practice. The 2026 series covers ecosystem architecture, standards for assurance providers, demand and incentives for independent assurance, and—with this new report—the operational challenge of defining and verifying AI quality. Further reports will address post-deployment accountability.

Explore

Download Pathways for Operationalising AI Assurance at https://resaro.ai/insights/whitepapers/pathways-operationalising-ai 

Hashtags: #ATXSummit, #ATXSummit2026