How to Move an AI Pilot Into Production
What changes between a working demo and a workflow people depend on, and the eight things that block the crossing.
Insights
About this archive
Each piece here answers one question properly rather than surveying a topic. Sources are linked next to the claim they support, limitations are stated rather than omitted, and the basis of each passage is labeled: first-hand experience, interpretation, or third-party evidence. Nothing is published as fact unless the claim register clears it. Written byMartin Zialcita.
12 published RSS feed
Start here
What changes between a working demo and a workflow people depend on, and the eight things that block the crossing.
Most organizations have more AI ideas than capacity. This is the method for choosing what to do first, with scoring scales, a worked example, and the mistakes that send teams in the wrong direction.
Fifty-two checks across eight phases, from scoping to handover. Printable.
Forty-seven checks across eight phases for agentic AI, because agents plan and act, not just respond.
All insights
12 published
What the document contains, section by section, and how to review one someone hands you.
The role from the perspective of someone hiring or contracting one, concretely and critically, with the cases where you should not bother.
Once you know you need the function, the decision is how to get it. A structured guide to hiring, contracting, and internal development, with the organizational conditions each requires.
What changes between a working demo and a workflow people depend on, and the eight things that block the crossing.
Most organizations have more AI ideas than capacity. This is the method for choosing what to do first, with scoring scales, a worked example, and the mistakes that send teams in the wrong direction.
Fifty-two checks across eight phases, from scoping to handover. Printable.
Forty-seven checks across eight phases for agentic AI, because agents plan and act, not just respond.
What the document should contain, why most policies fail, and how to write one staff will actually use.
Sequence, not training volume, is what determines whether a team actually adopts AI. It starts with the conversation most leaders skip.
Three acronyms, a crowded vendor market, and a lot of repackaged SEO. Here is what the terms mean, who coined them, and how to tell real practice from noise.
The retrieval-then-generation pipeline in plain terms, with an honest map of what is documented, what is inferred, and what nobody outside these companies actually knows.
What Hawaiʻi's economic structure (small businesses, lean teams, a relationship-based market) means for who benefits from AI adoption and who bears the risk.
Nothing matches that filter combination.
Next
The connected areas these articles sit within.
FrameworksTwo documented frameworks, versioned, with stated boundaries.
Verify a factThe claim register: what is confirmed, what is not, and the evidence behind each.
Discuss a projectIf a workflow is not working, that is enough to start a useful conversation.