AI Engagement Team case / Discovery Brief
Practice case. Prairie Ledger CPA is a made-up firm. All names, records and numbers are fictional.
Discovery Brief: Prairie Ledger CPA
PRACTICE CASE. Prairie Ledger CPA is a made-up firm. Every name and number is fictional.
Client: Prairie Ledger CPA, Red Deer, Alberta Type and size: Accounting firm, 14 staff (2 partners, 4 accountants, 5 bookkeepers, 2 admin, 1 office manager) Services: Bookkeeping for about 90 small business clients, year-end financial statements, personal and corporate tax Lead source: Claude 101 workshop, then replied to the follow-up email Call date: Tuesday, November 10, 2026 (30-minute video call) Decision makers: Dana and her partner Greg. Greg is more cautious about spending and "will want to know what we get for the money." Software in use: Karbon, QuickBooks Online, Microsoft 365, a client portal for document uploads Rates (cost and billing, per role): - Bookkeeper: about $40/hr with everything included ($32 in wages). Bills at $75/hr. (stated) - Partner: bills at $250/hr. Cost not given. - Accountant: not given
What they asked for
"I want AI to write our client emails. We send hundreds of them. Reminders, questions, explanations. My staff spend forever on them."
What is actually going on
Most of those emails go to the same clients every month, chasing bank statements and receipts. Late documents push the month-end close past the 15th target. That leaves partners reviewing in a rush, and coding errors slip through until year-end. The rest of the email load is accountants writing fresh answers to the same client questions, and it gets much worse in tax season.
Top problems
Problem 1: Chasing clients for documents every month
- In their words: "Every month it's the same people. She sends the first reminder, then a second, then she phones the worst ones."
- Who and how often: All 5 bookkeepers, every month. Last month one bookkeeper had 30 of her 45 clients missing something.
- Time or cost: About 40 hrs/month across the team (estimated: "I've never actually counted")
- Knock-on effect: About half of clients close on the 20th or later, against a target of the 15th (stated)
- What they have tried: Manual email reminders, then phone calls
- Score (size / frequency / AI fit / data / risk): 4 / 5 / 5 / 4 / 5 = 23
Problem 2: Same client questions answered from scratch
- In their words: "Clients ask the same questions over and over. My accountants write a fresh answer every time."
- Examples given: "Why is my GST different this quarter?" "What's this entry on my statements?" "Can I deduct my truck?"
- Who and how often: Accountants, all year. "Ten times worse" from February to April.
- Time or cost: Hours not known. Asked in the follow-up email.
- What they have tried: Nothing firm-wide. Some accountants use personal ChatGPT accounts.
- Score: 3 / 5 / 5 / 3 / 3 = 19
Problem 3: Coding errors found at year-end instead of monthly
- In their words: "Something goes to the wrong account, and we find it at year-end. Fixing it then takes way longer."
- Who and how often: Errors happen monthly during rushed reviews. The cleanup happens once a year. Who did the cleanup is not known.
- Time or cost: 60 to 70 hrs last year-end (estimated)
- What they have tried: Partner review at month-end
- Score: 3 / 4 / 4 / 4 / 3 = 18
Build readiness
- Software features available:
- Problem 1: Karbon can send automatic client reminders. Not known whether they are turned on.
- Problem 2: Microsoft 365 (Outlook). Not known whether they use email templates.
- Problem 3: QuickBooks Online. Not asked which review features they use.
- Firm AI accounts: No. Staff use personal ChatGPT accounts, possibly with client information.
- Real examples for testing: Not yet asked. Requested in the follow-up email.
- Rates or hours still missing: accountant cost and billing rate, accountant hours on client questions, who did the year-end cleanup, the partner's review hours.
Concerns to address
- Data privacy: "I'm not sure what they're pasting into it. That worries me." Firm accounts and usage rules come before any testing on real records.
- Accuracy: ChatGPT gave "a confident wrong answer about a deduction. That spooked people." Every client-facing answer needs an accountant's review.
- Staff fear of job loss: One accountant asked if people are being replaced. Dana: "I'm not. I want them doing better work, not more admin."
- Skeptical long-time staff: The office manager (20 years at the firm) "thinks it's a fad." Two younger bookkeepers are keen.
- Cost: Greg wants clear numbers.
Not now
- Partner review time: Hours not known, and it should drop once documents arrive on time. Track it as a result.
- Tax season capacity: About 15 clients turned away, about $6,750 (estimated). Driven by Problems 1 and 2.
- Late month-end close: A result of Problem 1. Use the share of clients closed by the 15th to measure progress.