Even the best BA process won't remove complexity entirely. But it can turn hidden chaos into manageable work — and that's often the difference between constant firefighting and controlled delivery. Part 2: from diagnosis to algorithm.

EU AI Act ᐉ What a business analyst should consider for medical projects ≡ Blog ArtofBA

Keep asking "Any new data yet?" — or get notified the moment something happens? We break down how webhooks work, how they differ from REST API and polling, and when to use them in your integration projects.

No real data. Can't collect it — too rare, too risky, too expensive. That's when synthetic data generation kicks in. But someone has to define what to generate and how. That's the BA and PO's job. Here's how the role changes in AI training projects.

The model works in the lab. It fails in production. Not because the model is wrong — but because the data is messy, the pipelines don't exist, and nobody owns the governance. Here's what the 5% who succeed do differently.

In 1628, a king ordered engineers to build a warship they knew would sink. It sank in twenty minutes. Klarna, Salesforce, and dozens of boards just did the same thing with AI. The ship has a name. So does the pattern.

On a project without a BA, requirements drift between PM, product owner, and dev. On a project with a BA but no process — the same thing happens. Here's what BA processes actually are, which patterns exist, and what they must include.