9 posts on Artificial Intelligence.
A scoring method for choosing the first enterprise AI use case, what a pilot costs and how long it takes, why most pilots stall before production, and what hardening a pilot for daily use actually involves.
Read moreUse a hosted model, build the workflow around it, measure accuracy on real examples before you commit, and keep the budget for the product rather than the research.
Read moreFive questions that separate AI features that create value from AI features that create demos, and the cases where the honest answer is a rules engine.
Read moreBuild costs for the three common approaches, the running costs that scale with users, the hidden costs of data and evaluation, and how to keep the number in proportion.
Read moreThe AI features that reliably deliver value in a first release, the ones that reliably disappoint, and what separates them.
Read moreWhen a conversational interface is right, when it is the wrong shape for the job, and the four kinds of AI feature that are not chatbots at all.
Read moreWhich AI features need your data and which do not, what counts as enough, how to get it if you have none, and the data hygiene that makes AI possible later.
Read moreWhat AI coding tools genuinely do for a non-technical founder in 2026, where the generated app stops being safe to launch, and how to use the tools without being misled by them.
Read moreMost AI projects fail on integration, not on models. Our approach starts with the process, the data and the decision, and only then with the technology.
Read more