Topics
AI & Data
11 articles in the AI & Data topic.
Why a semantic layer isn't enough for AI
A semantic layer defines your metrics. AI needs meaning that is governed, reaches the operation, and stays yours.
Connect once, not once per agent
Every agent you add has to be identified, authorised, networked and trusted, and then kept that way. You solve that once, centrally, or you turn it into a permanent job that grows with every agent you run.
Start with the spreadsheet that runs your company
A company brain does not need a finished data platform to prove its worth. It needs the one file your business already runs on, and the meaning around it.
You do not own what you cannot take with you
Sovereignty decides where your intelligence runs. Ownership decides whether you can leave with it. Most AI platforms hand you neither, and the gap only shows on the day you try to go.
An agent's bill is set before it runs
A customer interaction that cost four cents in 2023 now costs about $1.20. The token bill is the visible part; what an agent really costs is decided by the foundation it runs on.
Start with one number
Two departments, two figures for the same thing. How to turn that one contested number into an agreement a team dares to steer on, in a single afternoon.
Everyone is buying the same AI
Almost everyone buys the same models from the same vendors. Your edge is not in the model but in the layer beneath it: your definitions, your owners, your meaning.
Data quality has two owners
Every test passes and still nobody trusts the number. Data quality holds two questions: is the data technically correct, and is the right thing being computed?
Technology only creates value once people work differently
The platform runs, the numbers add up, and still nothing changes. Value shows up when the work around the system changes, in small steps.
A data platform finds the fault. It does not change behaviour.
A test reports the deviation reliably, day after day. Still the alert sits there. Why data quality only moves when ignoring it hurts, and what agentic AI changes about that.
AI needs your own knowledge. The industry now agrees.
Databricks and Microsoft renewed their partnership around one principle: AI delivers value only once it is grounded in your own knowledge. The principle is settled. Two questions are not: where does it start, and whose is it?
Related
Start with one decision that has to get better
Pick a recurring question where teams lose time or contradict each other. Make the meaning behind it explicit and bring it straight into daily operations.