01 / Articles
Technical content reviewed before publication.
AI Act transparency rules now apply, Italy publishes an implementing decree, and OpenAI, Anthropic and Google ship new models at lower prices. The AI news of September 2026 and why it matters to a business.
Using AI models on data is easy in a demo and hard in day-to-day work. How I keep AI model calls reliable in the tools I use to process data: errors, retries, idempotency, tracing and costs.
What the KuramaLab network contains, where each service stands and how the tools use KLAPI Hub. With a concrete example: MeetCraft, which transcribes locally and sends only the text you choose to a model.
Sometimes a scanner and a shared folder are enough. These are the signs that document digitisation really pays off, the costs to plan for and the questions to ask before starting.
Technicians' job reports as a worked example of automated data entry: five criteria to choose what to automate, how to handle exceptions and how to measure the result.
An AI agent is not a smarter chatbot: it uses tools and takes actions. How I pick the first use case, which permissions I grant, and where a person still makes the call.
