Enhanced interface

How to write prompts for the Finance Intelligence agent—Early Adopter

You interact with the Finance Intelligence agent by using Sage Copilot. You start a chat with it by using prompts. A prompt is a question or command that tells Sage Copilot what you want it to do for you.

Subscription	Finance Intelligence agent
Regional availability	Canada United States United Kingdom
User type	All user types
Permissions	You must also have the appropriate permissions to the data objects you're asking about. For example, if you ask "Show me all bills that haven't been paid", you need List and View access to AP Bills.

Use the following three-step model to build effective financial data prompts for the Finance Intelligence agent using Copilot:

Start with an action.
Identify the data point.
Specify the scope.

Note the following when building a prompt:

When referencing IDs or names, use the exact spelling and capitalization show in Intacct. For example, if an entity has an entity ID of E100, do not use e100.
If you are at the top level and want to include entity-level data, use "include private" in the prompt. For example, "List all AP bills due this month over 1000.00 and include private".

1. Start with an action.

Begin by telling Copilot what you want it to do. Use words such as:

Show
List
Find
How many
What's the total

This tells Copilot the type of response you expect.

2. Identify the data point.

Next, name the financial data you want to work with. Be as specific as possible so Copilot knows exactly which records to return.

You can include descriptors, like "unpaid" or "largest" here. Also, try using Intacct terminology when possible, like "recurring journal entries" or "missing account categories". For example,

Selected AP bills
Draft sales invoices
AR Customer payments
Bank accounts
GL accounts
Items

3. Specify the scope.

Finally, narrow the results by adding conditions such as timeframes, amounts, or specific record names. These "filters" help Copilot focus on the most relevant data. For example:

due in the next 30 days
over 5,000.00
for last month
for the Acme vendor

Examples of putting it all together

[action] + [data point] + [scope]

List AP bills over 3,000.00 for this month.
What were the top 5 customers by AR invoice amount for 2025?
Show all GL accounts created last month.

See the Finance Intelligence Agent prompt library for dozens of prompt ideas that you can copy and use now.

Follow up prompts

If you have a more complex request, the best practice is to ask one question at a time. Start with a simple data question and then ask follow-up questions based on the result.

Example 1—Revenue by item

For example, say you want to analyze the revenue produced by an item you sell. Ask the following prompts one at a time in the following order:

Which item appeared most in sales invoices last month?
What was the total revenue amount for that item last month?
What percentage of the total revenue last month did that item account for?

Example 2—Accounts Receivable recovery analysis

Say you want to connect overdue AR to expected cash. Ask the following prompts:

What is the total outstanding AR balance per customer, ranked highest to lowest?
Which customers have invoices overdue by more than 30 days?
How much cash can I expect from overdue invoices if 80% are collected?