Use AI chat and canvas

Ask questions about financial data, inspect how an answer was formed, and review fetched artifacts in a structured canvas.

Use AI chat and canvas

Use AI chat to investigate financial questions with data from Control. The canvas keeps fetched artifacts organized while the conversation provides room for follow-up questions, interpretation, and next actions.

Your organization must explicitly enable AI and select a processing policy before AI chat can process requests.

Ask a focused financial question

Include the scope needed to interpret the answer:

• the entity or group;
• the reporting period and comparison period;
• the statement, account, or metric;
• relevant dimensions or currency;
• the decision or anomaly you are investigating.

For example:

Compare operating expenses for the group in June with May, identify the three largest account-level changes, and show the supporting data.

A focused question reduces ambiguity and creates a result that is easier to reconcile with a statement or transaction view.

Review the answer and its evidence

Do not stop at the written conclusion. Inspect the data sources, fetched artifacts, and reasoning available with the answer.

Check that:

1. the requested entity and period were used;
2. the relevant artifact contains the expected rows and totals;
3. the comparison basis matches the question;
4. material claims are supported by the retrieved financial data;
5. any uncertainty or missing scope is visible before you act on the result.

Use follow-up questions to narrow the analysis, test an alternative explanation, or request transaction-level support.

Work with artifacts in canvas

Canvas tabs organize fetched data sets and other artifacts separately from the message stream. Use them to compare results side by side without losing the conversation that produced them.

Keep artifact identity and metadata in view when several data sets are involved. Similar-looking tables can represent different entities, periods, dimensions, or stages of the investigation.

Continue an investigation

Longer conversations can move through several steps:

Ask about a movement
  → review the fetched statement data
  → request an account breakdown
  → inspect supporting transactions
  → summarize the likely business cause
  → verify in the product workflow

If the connection is interrupted, confirm the active scope and artifacts after resuming before relying on a later answer.

Important considerations

• AI output should support financial review, not replace it.
• Reconcile material figures with the relevant Control statement and transactions.
• Do not include credentials or unnecessary sensitive information in a prompt.
• Treat an interpretation as a proposal until its underlying data and scope are verified.

Related content

• Enable AI and choose a processing policy
• Investigate transactions
• Build and review financial statements
• Connect external tools through MCP