Column: Silicon systems design


Documentation and knowledge retrieval can also deliver value when teams need faster access to methodology guidance, project history, or design intent. Review assistance may help improve consistency, especially across large teams and repeated artefact checks. Te riskier areas are those where AI


Figure 3: From isolated AI tools to measurable verification capability


processes, debug environments, review gates and sign-off flows. AI is easier to apply where inputs, ownership and review processes are clearly defined. Governance, security and IP protection are


also central maturity dimensions. Te NIST AI Risk Management Framework highlights the need for transparency, accountability and risk control in AI systems. In verification, these concerns directly affect traceability and confidence in sign-off. Figure 1 shows a capability-orientated


maturity benchmark for “AI in DV” adoption across strategy, governance, data readiness, workflow integration, pilot measurement, engineering skills, ecosystem readiness and scalability.


Measuring AI impact in verification A common mistake is to measure AI activity instead of verification impact – the number of generated tests, summaries, or classifications doesn’t prove capability. More meaningful measures include


coverage closure efficiency, regression turnaround time, debug cycle time, root- cause clustering quality, review consistency and reduction in repeated failure analysis. Engineer productivity should also be measured carefully, not as a vague claim but as time saved on repeatable engineering tasks. Figure 2 shows the distinction between


task-level AI optimisation and broader verification outcomes. While AI can accelerate activities such as log analysis, clustering, debugging assistance and coverage


14 June 2026 www.electronicsworld.co.uk


exploration, the broader objective is to improve measurable verification capabilities across coverage closure, regression efficiency, review consistency and sign-off confidence. Industry examples already show AI


and machine learning being applied to regression optimisation, debug automation and coverage closure. Tese are promising areas because they involve large datasets, recurring patterns and measurable workflow pain points. Even so, the metric should not be “AI was used”. Te metric should be whether the verification process became faster, more consistent, more traceable, or more effective at reducing risk.


Where AI works and where it struggles to scale Te most practical near-term uses of “AI in DV” are areas where the task is bounded and reviewable. Regression triage is a good example. AI can help cluster similar failures, identify likely duplicate issues, summarise logs and direct engineers towards probable root causes. Log summarisation and failure classification are useful because they reduce manual search effort while leaving the engineering decision visible. Coverage gap analysis is another


promising area, particularly where AI helps engineers identify patterns in unhit coverage, prioritise investigation, or suggest areas for directed stimulus.


output can be mistaken for verified intent. Unsupervised test generation is one example. Generated tests may increase activity, but without a clear linkage to verification intent and coverage goals, they can create noise. Automated interpretation of ambiguous specifications is another. Specifications oſten contain context, exceptions and architectural assumptions that are difficult to infer safely. Te highest risk case is AI involvement


in sign-off decisions without traceability. A recommendation that can’t be reproduced, reviewed or linked to evidence shouldn’t become part of the sign-off chain.


From experimentation to adoption Te practical shiſt is to move from asking “Where can we use AI?” to “Where does AI improve verification capability?” Figure 3 shows the progression from


specification and design through verification workflows enhanced by AI-assisted analysis and automation. Te emphasis is not on isolated AI features but on integrating AI into measurable verification capability across engineering flows. Instead of selecting tools first, teams


should assess verification maturity, data readiness, governance needs, workflow bottlenecks and measurable pilot opportunities. Good pilot definitions include the workflow being improved, the baseline metric, the expected improvement, the data used, the review process and the scaling condition. Without these elements, a pilot may demonstrate technical possibility but not adoption readiness. Te next phase of AI in design verification


will not be defined by who experiments first, but by who can measure, govern and scale it inside real verification flows.


Tis column continues in next month’s edition of Electronics World


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44