Your AI assistant can now chart like an analyst: introducing the Datama skill

Ask an AI assistant why your revenue dropped and you get a confident paragraph. Ask it for a chart your CFO can present and things fall apart.

It writes 300 lines of chart code, differently every time. The growth rates are sometimes right. The waterfall sometimes balances. And when someone in the room asks "can you split that mix effect by country", the answer is another five minutes of token burning and a chart that no longer matches the first one.

This is not a model problem. It is an architecture problem. Language models are built to generate, not to compute. Analysis needs the opposite: the same input must always produce the same output.

What the Datama skill does

The Datama skill plugs our analysis engine into Claude, ChatGPT and agent frameworks. When your assistant needs to explain a KPI move, it stops guessing and calls Datama:

1. You ask a business question: "Why did EMEA margin drop in Q2?"
2. The assistant calls the skill with the metric, the periods and the dimensions.
3. Datama runs the market equation and returns the decomposition plus a rendered chart: mix, price, volume and rate effects, ranked by impact.

The answer is a waterfall with three key points, not a wall of prose. The same waterfall your team already sees in Power BI, Tableau, Data Studio, Qlik, Excel or Google Sheets, because it is the same engine underneath.

Why deterministic matters in an AI workflow

The market equation expresses your KPI as a quantified tree, then splits any variation into independent drivers. No overlap, no double counting. It is math, not sampling.

That gives AI workflows three things they badly need:

• Trust. Every figure is traceable. If the skill says the mix effect is -2.2 points, you can audit the calculation. No hallucinated causes.
• Consistency. Monday's answer matches Friday's answer. The chart in the chat matches the chart in the board pack.
• Efficiency. One skill call replaces thousands of tokens of generated chart code. For agents that run analyses in loops, that is the difference between a usable workflow and an expensive toy.

Fine-tunable by humans, readable by executives

AI output usually comes with a take-it-or-leave-it finish. Datama charts stay editable: relabel a driver, group the long tail, force the order, adjust the growth rate display. The analyst keeps the last word before anything reaches a C-level deck.

That combination, deterministic engine plus hand-tunable output, is what generative charts cannot offer today. And it is why we built Datama to live inside every tool where decisions happen, including the newest one: the chat window.

Try it

The skill is rolling out now for Claude and ChatGPT, with agent framework support next. If you want early access, book a demo or start with the free web app to see the market equation on your own data.