--- title: "Claude for Bookkeepers: 1-on-1 AI Coaching Case Study | Clustr" description: "How Clustr coached a bookkeeping firm owner from a first Claude login to workflows in Excel, QuickBooks reporting and Outlook, on the firm's own files." url: https://getclustr.ai/case-studies/bookkeeping-firm-claude-workflows source: https://getclustr.ai/case-studies/bookkeeping-firm-claude-workflows --- # A bookkeeping and accounting firm: from Claude 101 to workflows, one-on-one with the owner The owner of a small bookkeeping practice had never logged into an AI tool. Over a series of live one-on-one sessions, we set up Claude safely for client data, then built it into the work the firm does every month: general ledger review, employee time analysis, client reporting and a crowded inbox. Updated October 2026 ## How did Clustr teach a bookkeeping firm owner to use Claude? Clustr Innovations, a human-first AI enablement consultancy, ran a one-on-one AI coaching engagement with the owner of a small bookkeeping and accounting firm in 2026. Each live session worked on the firm's own files, starting with a safe Claude setup for client data and moving into Excel analysis, QuickBooks reporting and Outlook triage, with a practice checklist between sessions. | Client | A bookkeeping and accounting firm (anonymized) | | --- | --- | | Firm size | The owner plus one employee, serving multiple small-business clients | | Program | 1-1 AI Coaching: ten live sessions with the firm's owner | | Format | Live virtual sessions, moving to a weekly cadence from October 2026 | | Starting point | Session 1 was the owner's first time logging into Claude or any AI tool | | Firm's tools | QuickBooks Online, Excel and Outlook | | Claude features used | Claude Pro, Projects, the Claude add-in for Excel, the QuickBooks and Microsoft 365 connectors, Claude in Chrome | | Status | In progress, with sessions scheduled through November 2026 | ## What did the coaching cover? Fundamentals first, then one real workflow at a time. Every exercise used the firm's own documents, and every session ended with a short checklist, because the owner told us that is the best way to learn. ### 1-1 AI Coaching Owner-led bookkeeping and accounting practice Small firms rarely get AI training built for them. The owner runs the books for several clients with one employee, closes a monthly reporting package for each client in QuickBooks, reviews general ledgers and time data in Excel, and worries about missing a client email in Outlook. The goal was to see what AI could take off the owner's plate without putting client data at risk. #### How It Ran ##### Claude 101, set up for client data Upgraded to Claude Pro and switched off model training on conversations, since the firm handles client financials. Added custom instructions, set up one Project per client, and covered when to use which model. A live demo turned a phone photo of a receipt into an Excel file of bookkeeping fields. ##### Claude inside Excel, on a real client file Installed the Claude add-in for Excel, had Claude build a general ledger review template with a time-tracking pivot sheet, then opened a live client time file. Claude ranked the top 12 projects by cost, said which overhead lines it left out, and surfaced staff logging hours to a client without naming a project. ##### Mapping the monthly reporting package Walked through the month-end routine of eight QuickBooks reports, exported to PDF or Excel and checked against the prior month. We tested what the QuickBooks connector covers and where Claude in Chrome can click through the rest, which runs on a manual trigger and needs supervised testing first. ##### Inbox triage and the weekly cadence After tax-extension season the owner asked to move to weekly sessions. The agenda: connect Outlook through the Microsoft 365 connector with rules that flag known clients, archive newsletters and never delete client email, and deepen the Excel and reporting work. Every exercise ran on the firm's own files, never on generic examples. ## What do the workflows look like? Two of the workflows from the sessions, drawn step by step. The owner stays the reviewer in both: Claude prepares the work, and the owner decides what goes to the client. 1. TriggerBiweekly time exportEmployee hours pulled from QuickBooks into Excel every two weeks. 2. InputClient time fileHours by project, rates on a second sheet, cost per line from a lookup. 3. AI stepClaude in ExcelRanks projects by cost, says what it excluded, and flags lines with no project. 4. Human reviewOwner reviewChecks the ranking and the flagged lines against what the owner knows of the client. 5. OutputCleaner dataA note to staff to log time to a specific project, not just the client name. 1. TriggerMonth-endData already reconciled and trusted; only the reports remain. 2. InputQuickBooks OnlineFinancial position, activity by month and class, budget vs. actuals, general ledger. 3. AI stepClaude pulls reportsThe QuickBooks connector where it reaches, Claude in Chrome for the rest. 4. Human reviewOwner checksPage breaks, formatting and a comparison with the prior month. 5. OutputClient packagePDF and Excel files saved to each client's folder. ## What has changed so far? So far the owner has gone from never using AI to a safe, configured Claude workspace, found a useful cost issue in a live client file, mapped what to automate next, and asked for weekly sessions. The engagement is still running. These are the changes on record from the first sessions, not projected results. ### From first login to a safe, configured workspace In one session the owner went from never having used AI to a paid Claude plan with model training switched off, custom instructions, and a Project for each client. ### A useful finding from real data in the first Excel session Claude's cost ranking on a live client file showed staff logging time to a client without a project, something the owner could act on that day. ### A clear list of what to automate next The monthly reporting package and Outlook triage were mapped step by step, with the parts Claude can do today separated from the parts that still need a person. ### A weekly habit the owner asked for Once tax-extension season ended, the owner asked to move from sessions every few weeks to a weekly slot for the rest of the engagement. Engagement in progress ยท outcomes from later sessions will be added once documented ## The Bookkeeping firm engagement, answered directly ### Can a bookkeeping firm use Claude safely with client data? Yes, with the right setup. Before uploading anything, we put the owner on a paid Claude plan, switched off the setting that lets conversations be used to improve the models, and kept each client's files in a separate Project so context never mixes between clients. ### What can Claude do in Excel for bookkeepers? With the Claude add-in for Excel, Claude works in a side panel on the open workbook. In this engagement it built a general ledger review template and a time-tracking pivot sheet, ranked a client's projects by cost, explained which lines it excluded, and flagged hours logged without a project. ### Can Claude automate monthly QuickBooks reports? Partly. The QuickBooks connector covers reports such as profit and loss, cash flow and benchmarking, but not everything QuickBooks offers. Claude in Chrome can click through the remaining reports, but it runs when you trigger it rather than on a schedule, so we test it under supervision before relying on it. ### Is one-on-one AI coaching worth it for a small firm? For an owner-led practice, it fits better than a group workshop. Every session works on the owner's own files and workflows, the pace follows the firm's busy seasons, and the checklist between sessions turns each demo into a habit. ### Who was the client? A small bookkeeping and accounting practice: the owner and one employee, serving multiple small-business clients from QuickBooks Online, Excel and Outlook. We do not name the firm or its clients publicly. ## Be the next case study Tell us where your team is today. We'll map the fastest path from licenses to daily use. We use cookies We use cookies and Google Analytics to understand how visitors use our site. [Privacy Policy](/privacy)