← All use cases

Use case #23 · CPA

Firm owner builds a Karbon connector so Claude can reassign work and turn a client email into work items

Beau Gaudron, Growthwise (The Firm) ·

The short version

What they did

Gaudron, who runs a firm in Newcastle, used Claude on Karbon exports to check contacts for missing mobile numbers and build an hours dashboard from timesheets. He then had Claude generate an MCP connector from Karbon’s published API spec. With it, he asked Claude to move a staff member’s jobs to a colleague while she was on leave, and to turn a client email into work items. He notes that Karbon has announced its own MCP server and advises checking its status before building your own.

Result (as claimed)

“Across 157 contacts it ran the check and flagged the gaps in seconds.” On the connector: “It found her four work items and reassigned them in one request.” And from one client email, Claude “turned that one message into nine work items and assigned them to the right person.”

Prompt

Start with a contact export:

I run an accounting firm and want to check our Karbon contact records. I’ll attach [contact export] and [notes explaining its columns]. Help me find missing mobile numbers before I update anything.

1. Count the contacts you read and identify any rows you couldn’t process.
2. List every contact without a mobile number, grouped by client. Show the row reference so I can find it.
3. Put email addresses or phone numbers that appear malformed in a separate table, with a short reason.

Use plain language and finish with the total checked and the counts flagged. Don’t invent missing details, treat an unfamiliar phone format as certainly wrong, or change any records. Ask me if you can’t identify the client or mobile-number columns.

For the connector, give Claude Code this request:

I run an accounting firm and want Claude to work with Karbon through a connector. I’ll provide [Karbon’s published API specification], [our existing connector code, if any] and [the work actions we need]. First help me check whether an available Karbon connector covers those actions.

1. If a custom connector is needed, explain the plan and access required in plain language, then wait for my review.
2. After I approve, build a test version that can list one person’s open work, propose reassignments during leave, and turn [a sample client email] into proposed work items.
3. Show the exact changes for review before making them in [our test workspace].

Return the code, short setup instructions and tests for each action. Don’t assume the API supports a feature without checking the specification. Don’t change live jobs or create live work items during testing.

Do it yourself

These steps are my suggested way in, not Beau Gaudron’s setup.

Level: Advanced for a connector that changes work in your practice system. Checking an export in a chat is Beginner.

What you need

  • A chat assistant that reads spreadsheets, such as Claude.
  • An export of contacts or timesheets from your practice management system.
  • For the full version: a connector (MCP server) to that system. Check whether your vendor offers one before building your own.

Start simple

  1. Export your contacts and paste the contact-export prompt with the file.
  2. Confirm the count it checked matches the number of rows.
  3. Export timesheets and ask for hours by person and client in a simple table.
  4. Fix the gaps it found in your system yourself.

Take it further

  • Connect the assistant to your practice system with an official or custom connector.
  • Start with read-only questions, such as listing one person’s open work.
  • Then try small changes, like reassigning work for someone on leave, and approve each one.
  • Paste a client email and ask it to propose work items before it creates any.

Check your work

  • Check the count and a few flagged contacts against the system.
  • Review every reassignment or new work item before you accept it.
  • A connector can change your records. Give it the least access it needs and keep its keys private.

Before you rely on it

  • This is general information about how accountants use AI, not professional guidance for your engagements.
  • You stay responsible for the work under your professional standards, including review and sign-off. An AI tool’s output is a draft, not evidence.
  • Before you send client data to an AI tool, read the tool’s data and training terms, and get any client consent you need. For tax return information, that includes IRC section 7216.

Source checked. The link, author, date and quote were confirmed against the source. The result itself has not been reproduced. Prompt adapted for this library, inspired by the case. Business size: Not stated. Results are quoted as the source claimed them and never improved. Spot an error or a broken link? Email charlie@cbarmorecpa.com.

More in Practice admin

New cases go out every Saturday in Inside AI.