PerspectiveAugust 1, 20267 min readYour First AI Agent: Pick the Right Use CaseThe first agent should be narrow, measurable, and close to clean data. A filter for choosing — and the use cases to avoid.Read →
Deep DiveJuly 11, 20266 min readAgents Aren't the Hard Part. Your Data Is.Most firms fail at agents for the same reason they failed at analytics: the data underneath is siloed, unprocessed, or ungoverned. The fix is engineering — not another model demo.Read →
Deep DiveJuly 11, 20267 min readWhat It Takes to Ship a Production Invoice AgentSeven months of agent tuning produced mediocre results. One month of data engineering plus a two-week build outperformed it. A deep dive on the sequence.Read →
