Conversational analytics: ask AI for your own data
Conversational analytics means asking about your own company's data in ordinary speech — in AI chat, as if you were asking an analyst. It only gives reliable answers when AI works on a data warehouse with a semantic layer that determines what each metric means.
What is it
Conversational analytics differs from the normal use of chat AI in one way: instead of general knowledge, the AI answers over your data. You ask what the margin in category X was last month, net of returns, and you get the answer based on the definitions your business has agreed upon — not the model's guesses.
You choose the tool yourself: ChatGPT, Claude, Gemini, Copilot, or whatever comes next. Conversational analytics is not based on a specific model, but on the context you give to the model. You replace the tool. The context remains.
Why AI is guessing over exports
Asking AI about sales without context is like a CEO going out on the street and asking passers-by instead of asking their own data analyst. Passers-by are willing to answer, they just don't know your business. You get general advice, or an honest "I don't know".
When you give a passer-by a box of invoices — upload an export from an online store or accounting to the chat — they will answer confidently and with a specific number. But did you choose yourself what sales are: with or without VAT? Before the return or after it? According to orders in Shoptet or invoices in Pohoda? Including unpaid or just paid?
This step hurts. And people do it every day. The example of sales versions at the beginning shows how 32 different numbers are created from one question.
What do you need for it?
Three things: data in one place, agreed definitions and an AI chat that can access it. For a free consultation, we'll go over your systems and show you where to start.
- A cloud data warehouse into which data is automatically downloaded from all systems — online store, accounting and marketing — and which stores it with history.
- Semantic layer: unified definition of metrics. What is Sales, Margin, or Active Customer — Names, Descriptions, Calculations, and Relationships Between Data.
- AI chat you already use, connected to data with explicit permission. No inserting exports and no fancy definitions.
What you may ask
Five questions that AI can answer correctly only over a data warehouse with a semantic layer:
- What was the margin in category X last month after subtracting returns?
- Which marketing channel brought us the most margin, not just revenue, this year?
- Why did our margin fall in September compared to last year?
- Which products do we have in stock for more than 90 days?
- How do we stack up against the first half plan?
How do you know the answers fit?
The correct answer will tell not only the number, but also how it was created: what definition, what sources, what period.
Before we hand over the warehouse to you, we test the AI on a set of real questions with known answers. You will see the results.
And when the definition changes, the answer changes for everyone — once, in the semantic layer.
Frequently asked questions
ChatGPT tells me it doesn't know our sales. He's not making anything up, so where's the problem?
You are right, an honest passerby will say 'I don't know'. The problem starts when you give him a box of invoices — upload an export to the chat. Then it answers confidently, but according to the definition it has come up with. And this happens every day in companies.
We have our own analyst. What is it to us?
The semantic layer is your analyst's knowledge, written once. Everyone gets answers according to his definitions — and it stops answering the same question for the fifth time.
Which AI tool should we choose?
The one your team is already using. With the finished context, the choice of tool is reversible.
How do we know the AI is answering correctly?
Before handover, we test the AI on a set of real business questions with known answers and tune the semantic layer until it passes the test. You will see the results.
More on the topic
Make AI chat your analyst
Free and non-binding online consultation. We'll go through your systems and show you where to start.
Free consultation