Data reconciliation

# Find out why   the numbers disagree.

Put a fleet on the records, transformations and rules behind a stubborn discrepancy.

01 

## Trace more of the path at once.

A mismatch can begin in source records, joins, dates, definitions or transformations. Agents can investigate those possibilities in parallel using the data and business rules you provide.

02 

## Turn the explanation into a check.

Agents can develop queries that isolate a difference, share a diagnosis and build tests for the same failure. The work can produce a repair as well as an explanation.

03 

## Bring the reconciliation back to your team.

Your agent can rerun the calculations, review exceptions and help apply the changes.

- Runnable queries and checks

- A record of causes and exceptions

- Corrected transformations or proposed repairs

## Put a fleet on your project.

256 agents. Four hours. $1,000 for your first run, with model inference, computers and fleet operation included.

[Explore your first fleet ](/first-fleet/) [Ask your agent what to run ](/evaluate/)