16-62 Insights https://16-62.co/insights Field notes on data architecture, reporting and applied AI. en Is your AI workflow ready for operations? Five tests. https://16-62.co/insights/ai-workflow-ready-for-operations https://16-62.co/insights/ai-workflow-ready-for-operations Sun, 25 Oct 2026 00:00:00 +0000 Most AI pilots look convincing in a demo. Far fewer survive contact with real documents, real exceptions and real consequences. Before a workflow touches operations, it should pass five tests. Why every system shows a different revenue number. https://16-62.co/insights/why-every-system-shows-a-different-revenue https://16-62.co/insights/why-every-system-shows-a-different-revenue Sun, 25 Oct 2026 00:00:00 +0000 The webshop says one figure, the payment provider another and the accounts a third. Usually all three are correct. They answer different questions. Here is where the differences come from, and how to settle them for good. Search, retrieval, or RAG? Start with the question. https://16-62.co/insights/search-retrieval-or-rag https://16-62.co/insights/search-retrieval-or-rag Wed, 21 Oct 2026 00:00:00 +0000 Retrieval-augmented generation is useful when a system needs to find relevant material and write an answer from it. It is not the right default for every knowledge problem. The design should follow the question being asked. A warehouse is only useful when its metrics mean something. https://16-62.co/insights/warehouse-metrics-that-mean-something https://16-62.co/insights/warehouse-metrics-that-mean-something Tue, 20 Oct 2026 00:00:00 +0000 Connecting systems to a warehouse is the easy part. The work that matters is agreeing what a customer, an active account or net margin means, and keeping those definitions consistent as new data arrives.