Sarah is the only accountant at a 20-person firm, and collections ate the first two hours of every morning. This is how we found the time she was losing, what we proposed, and what the report said it was worth.
One person, one workflow, six tools. This is the shape the diagnostic starts from, before it surfaces anything.
A 20-person professional services firm. The owner knew the mornings felt slow but couldn't say why, and nobody inside the work could see the shape of it either.
Sarah is the firm's only full-time accountant, and the kind of person who quietly holds a company together. Collections is hers alone. She runs it every morning before anything else, she built her own tracking spreadsheet because the standard reports never showed what she needed, and when the exports got messy she found her own AI tool to clean them up.
Six surfaces, every workday. QuickBooks for the books of record, the firm's billing system, her hand-built AR spreadsheet, two email inboxes for client replies, and the project management tool she cross-references. Plus ChatGPT on a personal account she pays for herself, where she cleans up messy transaction exports before pasting them anywhere else.
None of this came from an audit at a distance. We pulled the invoice history, interviewed the team, then sat with Sarah through two of her mornings and watched the work happen. Three findings came out of it. Each one is observation, not proposal. The proposal comes next.
We timed the mornings. Sarah spends two hours a day across QuickBooks, the AR spreadsheet, and the project management tool, finding overdue accounts, drafting reminders, and logging what happened. About 90 percent of it follows a pattern. The other 10 percent is judgment, and that stays with Sarah.
The spreadsheet showed up in the inventory pass. Sarah maintains it by hand because QuickBooks reports do not show what she needs in the format she needs. When the two diverge, QuickBooks is technically correct. But the spreadsheet is what the owner sees.
This one surfaced in her interview, when we asked how she cleans the exports. Sarah pays for ChatGPT herself, expenses nothing, and has no formal approval. The data she pastes in includes client names, amounts, and account references, and there is no record of what gets sent. Leadership had no idea.
The Phase 5 report proposes a single Path A build, scoped to AR collections. The 90 percent mechanical part of Sarah's mornings becomes an automated draft-and-review workflow. The 10 percent judgment portion stays with Sarah. The spreadsheet folds in. The shadow ChatGPT exposure closes as part of the same build.
Pulling balances, drafting reminders, logging outcomes. The mechanical morning, end-to-end.
Review and exception handling. The softer-touch clients. Sign-off on every send.
The AR tracking spreadsheet disappears. The build reads from QuickBooks directly and the firm stops running against a separate document.
Sarah moves from a personal ChatGPT account to a sanctioned, managed AI workspace. The exposure becomes a policy decision instead of an unknown.
The Phase 3 measurement attaches numbers to the proposal. Two carry dollars, one is calendar.
Two of those are dollars and one is hours, but the dollars are not all the same kind of dollars. Tool spend reduction is cash, money the firm stops sending out. The labor capacity recovered is hours. What those hours buy depends on what the firm reallocates them to. The report keeps the two separate.
And Sarah's mornings move to client work she finds more rewarding.
Thirty minutes. We can tell you whether the diagnostic fits your team and what it would likely surface, the way the report surfaced things for Sarah.