Skip to content
← All work

Own product / AI data assistant

Professor Smartypants

Product
Chat over connected data
Datasets
6 connected
Languages
Answers in yours

Ask the data. Get a straight answer.

A chatbot connected to real datasets through a semantic catalogue. Ask in plain words and get a checked answer with a chart, or an honest no with the closest thing it can do.

Abstract warm light waves on a dark background, the Professor Smartypants cover
Professor Smartypants / PhD in Company LorePublic demo of conversational reporting

The product

A straight answer, or an honest no.

Most data chatbots will answer anyway, even when the data cannot. Professor Smartypants is built the other way round: every figure in the answer comes from data that was actually read, the plan is checked before it runs, and each chart says what was counted, for which period and how it was sorted.

Six public datasets are connected today, so anyone can try it: Lithuanian Seimas · Chinook music store · Pokémon · The Simpsons · Netflix titles · Video game sales.

Professor Smartypants answering "Revenue by genre" with a reviewed bar chart
A checked answer / Chinook music storeLive demo, revenue by genre
How an answer is made / Select a step to see what it does

It writes a plan, not a query.

Each dataset comes with a semantic catalogue, written and tested in advance: what can be measured and how it can be broken down. The language model may only choose from it.

Asking "what can I ask?" returns a menu built from the same catalogue, so visitors always see what each dataset can actually answer.

The "How we got here" panel listing the six steps behind an answer
How we got here / Every answer shows its stepsLive demo

From many datasets to one honest answer.

Every connected dataset goes through the same fixed loop. Small talk takes a shortcut after the first step; data questions walk all five.

Connected dataLithuanian SeimasChinook music storePokémonThe SimpsonsNetflix titlesVideo game sales
Professor SmartypantsUnderstandPlanCheckRunReview
AnswerChart from the resultPlain-language summaryHonest no + closest option

A wrong number is worse than no number.

Ask the music store for last year’s profit margin and it will not guess. Profit margin is not tracked there: there is revenue, but no costs. The loop stops, says why, and offers the closest question it can answer, such as revenue by year.

The suggestion is always a real question from the catalogue, so a visitor is never stuck.

Professor Smartypants declining a profit margin question and offering revenue by year instead
An honest no / With the closest question it can answerLive demo

Fast where it can be. Careful where it must be.

Each job goes to the cheapest thing that can do it well. That keeps answers quick, cheap to run and easy to check.

Quick decisions
Sorting questions, reviewing results
Language model
Only to plan and to write the answer
Plain code
Everything that touches the data
Data access
Read-only, time-limited

Professor Smartypants is a kodeo product. Want it on your own data? Write to hello@kodeo.lt.

Want answers like this from your own data?

Professor Smartypants can be connected to your database, CRM or spreadsheets, with a catalogue built around your business. Tell me what your team keeps asking, and I will show you what it can answer.

Get in touchTry the demo

I reply within one working day.