INSIGHTS · FREE TO READ
The number behind a model choice. The definition behind a metric. Each piece follows a product decision through the evidence and the trade-offs.
14 resources to explore
A dated review of employer guidance and candidate accounts, with practice for model platforms, offline AI and enterprise product decisions. See what the evidence supports and what to confirm with your recruiter.
Read for freePractise customer choice, estimation, metric diagnosis and distribution with original AI product cases. Strengthen the PM reasoning behind your technical interview answers.
Read for freePractise technical judgment by tracing an unread-message summary from permissions and context to model behavior, evaluation and cost. Includes a study route and a whiteboard exercise.
Read for freeFollow a maintenance assistant from a promising demonstration to a defensible investment. Test demand, data readiness, operating cost and ownership with original practice cases.
Read for freePrepare a project story a reviewer can inspect: your decisions, permitted artifacts, metric definitions and the limits of your evidence. Includes a practical claim-audit worksheet.
Read for freeUse official interview guides and AI PM role descriptions to choose useful practice. Six original exercises cover writing, evals, model behavior, voice, experiments and agent recovery.
Read for freeTurn an AI quality score into a launch decision. A worked contract-summary exercise covers sample size, failure severity, segment floors and production monitoring.
Read for freeWhy dividing an AI bill by requests misleads product decisions, with a worked comparison of a small and a mid-tier model once failures and human fallback are counted.
Read for freeKlarna's support figures and disputed quality narrative show why AI PMs need source checks, segment metrics and a measure of resolution that holds.
Read for freePer-resolution pricing turns a metric definition into a billing term. Using Fin's $0.99 model as the example, what a PM must decide about abandonment, re-contact and multi-issue conversations.
Read for freeWhen Salesforce widened what counts toward Agentforce ARR, the headline nearly doubled in two quarters. How AI PMs should define, change and report metrics so the trend means something.
Read for freeEU AI Act milestones after the 2026 Digital Omnibus, the US state laws taking effect, and India's DPDP rules, laid out as a calendar a product manager can plan against. Not legal advice.
Read for freeWhy a launch you stopped, delayed or narrowed is the most useful behavioural story an AI PM can bring, how to structure it with evidence, and the follow-ups that test whether it is true.
Read for freeA $1.5 billion settlement over books obtained from pirated sources excludes model-output claims from its release. What that leaves for AI PMs to decide about data provenance, indemnities and output risk.
Read for free