---
title: "From \"What Happened?\" to \"What Will Happen?\""
description: "Databricks Genie makes descriptive analytics accessible in natural language, but predictive questions still require specialized data science workflows and carefully prepared datasets. This post presents a multi-agent supervisor deployed as a Databricks App, combining Genie, TabPFN, and Agent Bricks to turn business questions into predictions. The orchestrator asks Genie to use governed Lakehouse data, schemas, relationships, and semantics to generate labeled training data through SQL, then sends it to TabPFN, which predicts in a single forward pass without feature preprocessing, model selection, or hyperparameter tuning. The resulting conversational experience supports descriptive and predictive analytics with Unity Catalog lineage and access control, while an MLflow GenAI evaluation harness monitors reliability and regressions. Its central limitation is that predictions depend on Genie producing a meaningful dataset with a clear label, so missing signals, joins, outcomes, or agent omissions can make results unreliable."
---

# From "What Happened?" to "What Will Happen?"

[Databricks](https://yomu.fyi/company/databricks) · Ryuta Yoshimatsu, Javier Poveda Panter, Dominik Safaric, Philipp Singer, Diana Kriuchkova, Sauraj Gambhir, Dael Williamson, Bryan Smith · May 21, 2026

**Type:** Problem & solution

## Summary

Databricks Genie makes descriptive analytics accessible in natural language, but predictive questions still require specialized data science workflows and carefully prepared datasets. This post presents a multi-agent supervisor deployed as a Databricks App, combining Genie, TabPFN, and Agent Bricks to turn business questions into predictions. The orchestrator asks Genie to use governed Lakehouse data, schemas, relationships, and semantics to generate labeled training data through SQL, then sends it to TabPFN, which predicts in a single forward pass without feature preprocessing, model selection, or hyperparameter tuning. The resulting conversational experience supports descriptive and predictive analytics with Unity Catalog lineage and access control, while an MLflow GenAI evaluation harness monitors reliability and regressions. Its central limitation is that predictions depend on Genie producing a meaningful dataset with a clear label, so missing signals, joins, outcomes, or agent omissions can make results unreliable.

## Context

Business intelligence commonly addresses descriptive and diagnostic questions, while predictive questions have required specialized teams to explore data, engineer features, train models, and maintain them. Although TabPFN makes prediction faster, someone still needs to translate a business question into a suitable dataset. The post addresses this boundary between conversational BI and predictive analytics.

## Approach / What changed

A multi-agent supervisor deployed as a Databricks App coordinates Genie, TabPFN, and Agent Bricks. It interprets a question, asks Genie to generate labeled data from governed Lakehouse sources using SQL, passes the formatted data to TabPFN for a single-forward-pass prediction, and returns a recommendation. Unity Catalog provides lineage and access control, while an MLflow GenAI evaluation harness evaluates the live agent and logs results.

## Takeaways

- Genie dynamically translates predictive questions into labeled datasets by using organizational schemas, relationships, and business semantics to generate SQL.
- TabPFN produces predictions in a single forward pass without feature preprocessing, model selection, or hyperparameter tuning.
- Reliability depends on whether Genie can construct a meaningful dataset with the required signals, joins, and outcome label; MLflow evaluation is used to identify reliable question classes and regressions.

**Tags:** [AI Agents](https://yomu.fyi/topic/ai-agents), [Genie](https://yomu.fyi/topic/genie), [Lakehouse](https://yomu.fyi/topic/lakehouse), [MLflow](https://yomu.fyi/topic/mlflow), [Unity Catalog](https://yomu.fyi/topic/unity-catalog)

- Source: [Databricks](https://www.databricks.com/blog/what-happened-what-will-happen)
- Source URL: https://www.databricks.com/blog/what-happened-what-will-happen
- Ingested by Yomu: 2026-08-31T03:33:39.237Z

[Read original post](https://www.databricks.com/blog/what-happened-what-will-happen)