Articles
Your backtest knows more than your system did A chronological split does not establish that historical inputs were available at prediction time. Replay arrival, publication and revision boundaries before using a backtest to approve the proposed service. Define client trust for MCP metadata documents Client ID Metadata Documents give MCP authorisation servers a resolvable client identifier. Define trusted applications and metadata change controls before treating that identifier as approval to access internal systems. The output cannot tell you whether the mechanism was ever there A plausible output does not establish how a supplier produced it. Where a mechanism matters to the purchase, define direct evidence and a test that could expose its absence before accepting the service. Agent memory is a new system of record, and it arrived as a feature flag Persistent agent memory creates retained records that can affect later decisions. Define what may be stored, how its source and scope are preserved, and how entries are corrected or erased before expanding its use. An AI-drafted control profile describes your documents, not your controls An AI-drafted control profile is bounded by its source evidence. Distinguish policy, self-report, observations and test results, then review source mappings and evidence gaps before using the profile for an assurance decision. Only the questions where they disagree are deciding your model choice A paired model comparison needs more than two average scores. Record item-level disagreements, uncertainty and the smallest difference worth acting on, then assess whether the evaluation can resolve that decision. Review the recovery boundary of every agent action Review an agent’s permitted actions, their consequences and recovery paths. Reversibility gives the architecture review a concrete question about who can correct an effect and within what time. The architecture outlives the team that built it Design an AI service so a successor team can operate, evaluate and change it. Retain decision records, quality criteria and reproducible dependencies, then test the handover instead of assuming documentation is sufficient. Specify the documentation profile before assessing conformity A documentation standard helps buyers only when its scope and required evidence match the decision. NIST’s July 2026 draft separates base templates from profiles that buyers can make more demanding. Keep clinical purpose and confirmation in change control MHRA guidance distinguishes administrative ambient scribing from products with a medical intended purpose. Review changes to clinical recommendations, confirmation and promotional claims before extending the workflow. Audit the benchmark before procuring against it Annotation corrections changed agent rankings in two text-to-SQL benchmarks. Use public scores to inform a shortlist, then inspect the answer key and test candidates against cases the organisation can review. The ICO's agentic AI timetable moved. The design work remains. Future agentic AI guidance does not remove current design responsibilities. Define the processing purpose, permitted tool access and routes for people to challenge significant automated decisions, while tracking the ICO’s publication status.
Bring us the question
Something here already on your desk?
That is the conversation we are best at. Thirty minutes, a written summary, no obligation.