Skip to content

← All products

Fennibi Chat Preview

Chat: your Power BI figures on demand

You want a number, not a search. Fennibi Chat is the one place where you ask. Fenni calculates the answer directly from your Power BI models, shows it as a chart and explains how it was reached.

The problem

  • The report zoo: The number you need is somewhere in one of twenty reports. Which one, only the person who built it knows.
  • The detour via specialists: Every question that no report answers becomes a request to the BI team. The answer costs time nobody has, neither you nor the specialist.
  • Contradictions: Two reports, two numbers for the same measure. The meeting discusses numbers instead of actions.

What you get

  • One interface to all numbers: You ask in your own words, Fenni finds the right model and the right measure.
  • Consistent answers: Calculations use the measures of your model, not a copy. The same question gives the same number, for everyone.
  • New analyses instead of searching what exists: Fenni also calculates multi-step questions directly in the semantic model.
  • A simple analysis tool: forecasts (“Where will we land at year end?”), simulations (“What if the price rises by 5%?”) and classifications (“Which customers are A, B, C?”), without Excel and without a data scientist.
  • Chart and conclusion: consistent charts and a conclusion that shows the numbers behind every statement.
  • Saved analyses: the same question at any time with fresh numbers.
  • Your permissions apply: On request, Fenni asks under the identity of each individual user. Then everyone sees only what they are allowed to see in Power BI.

Getting started: the analyst onboarding

An AI analyst is only as good as the model it works on. That is why Chat starts with an onboarding:

  1. 1You write down the questions your business really asks. That is your knowledge of your business, and that is exactly what Chat has to be measured against.
  2. 2Measurement before: how many of these questions does Fenni answer correctly?
  3. 3Corrections to the model: descriptions, units, slow measures.
  4. 4Measurement after: the hit rate, in black and white.

Your questions can be measured again at any time, for example after a change to the model. This makes sure that Chat keeps delivering good answers. The quality of the answers is measured automatically along the way.

Why it pays off

  • No more report zoo: one place instead of twenty reports.
  • Fewer requests to specialists: your BI team builds models instead of working through ad-hoc analyses.
  • Measured quality on your own models, not promised quality.