What 99.5% accuracy actually means.
Vendors throw around accuracy numbers like marketing copy. Here's what the math looks like at field, record, and document level — and the QA process that gets us there.
Quality · OperationsEssays on data quality, archives, the human craft behind AI-ready datasets, and what happens when you take ordinary work seriously.
Three years into the LLM boom, the businesses winning with AI are the ones who took data entry seriously when nobody else was. A look at where humans-in-the-loop still wins, what it costs to skip the layer, and three industries where the work is heating up.
Vendors throw around accuracy numbers like marketing copy. Here's what the math looks like at field, record, and document level — and the QA process that gets us there.
Quality · OperationsWhat duplicated customer records, mis-keyed claim amounts and broken catalog data really cost — line by line, on the P&L.
Finance · RiskThe boring, unfashionable answer to the question every AI-first company hits in year two: who reviews the model's mistakes, and at what scale?
AI · OperationsOne thoughtful piece every 3–4 weeks. No marketing email, no nonsense, no link tracking.