Nontechnical teams can use AI-built scripts to automate repetitive work, reclaim valuable time, and turn an accounting firm’s operational problem into a marketable product.
This is a preview. The complete video episode, with commentary and transcript, is first available exclusively to PRO Members | Go PRO here
Sponsored by Radical Pricing: Optimize Profits, Delight Clients, and Build a Top-Value Firm
Subscribe to CPA Trendlines podcasts anywhere: Apple, Google/YouTube, Spotify, iHeart, Deezer, Amazon Music, Audible, Player FM, Audacy, RSS.
It’s Not Just the Numbers
With Penny Breslin and Damien Greathead
For CPA Trendlines
Most accounting firms talk about AI automation in the future tense. They’re waiting for the right tool, the right vendor, or the right moment to start using it seriously.
Trent Yesberg didn’t wait. As director of Regional Business Services in Townsville, Australia, he started asking Claude and ChatGPT to write scripts that run his firm’s internal workflows directly, with no ongoing subscription or per-task cost attached.
MORE Penny Breslin and Damien Greathead | MORE Advisory & Consulting
JOIN From Scorekeeper to Strategist CAS Workshop | BUY “It’s Not Just the Numbers”
Exclusively for PRO Members. Log in here or upgrade to PRO today.
What began as a way to save his own team a few hours a week turned into Twine Biller, a tool now used by other firms to fix one of accounting’s quietest profit leaks: subscription rebilling that slips through the cracks month after month.
His approach cuts against a common assumption, that AI automation in a firm has to mean either building an entire internal platform from scratch or renting one from a big software vendor. There is a middle path, and Yesberg has been living in it for the better part of a year.
Scripts That Run Without the Meter Running
At Regional Business Services, nobody on the team writes code from scratch. Yesberg describes the firm as non-technical, guided at every step by Claude. But the pattern he’s settled on is specific: instead of running a task through an AI model every time it needs doing, the firm asks Claude to write a script once, then runs that script locally on its own computers and servers, connected directly to its QuickBooks and Xero files.
Sponsored by Radical Pricing: Optimize Profits, Delight Clients, and Build a Top-Value Firm
The distinction matters more than it sounds. A workflow that lives inside a chat window costs money and time every time someone runs it. A script that Claude wrote once, sitting on a firm’s own machine, costs nothing extra to run again next month. It’s the difference between renting a tool and owning one, even when an AI model is what built it.
“We’re not just using Claude to do the work, we’re actually getting it to write our own scripts so that we can have it set up on our desktop or on our servers,” says Yesberg.
The Pharmacy Workflow That Proved the Idea
The clearest example is a group of pharmacy clients RBS has served for years. Their point-of-sale systems now email structured reports straight into a monitored inbox. From there, an automated process sorts the attachments into folders, scans the data inside each CSV, builds a per-store spreadsheet, and drafts a pre-populated email with a rendered PDF ready to send back to the client.
The only step a human still touches is the review, and even that is temporary. The workflow has existed for about a month, and Yesberg expects his team to keep narrowing the filters until review disappears, too. What used to be a manual, deadline-driven scramble is becoming a system that mostly runs itself.
The Team Didn’t Lose Hours, They Gained a Question
Automation freed up real time, but it also raised a harder question than anyone expected. Staff who used to spend a meaningful chunk of their week downloading, collating, and chasing client files suddenly found that work gone. Yesberg says the reaction was closer to shock than relief, the work had always felt valuable, even though it was really just process.
That leaves RBS deciding what to do with the time it got back. Bring on more clients at the same staffing level? Lean harder into advisory conversations that some clients want and others have never asked for? Yesberg hasn’t settled on an answer, and he’s treating it as a genuinely open question rather than assuming the freed-up hours automatically become billable advisory work.
The Billing Problem That Became a Product
Long before automation reached client workflows, Yesberg had already identified the most tedious process in his own firm: rebilling clients for the QuickBooks, Xero, and add-on subscriptions RBS resells at a margin. Tracking price changes, added seats, and expiring discounts across dozens of clients used to consume a full day, and firms routinely missed revenue simply because nobody caught every change in time.
That frustration became Twine Biller, the first product under Yesberg’s software company, Twine Apps. It maps every subscription to the right client in the ledger, applies a custom margin per client, and consolidates multiple charges into a single invoice.
“It took over a day,” Yesberg says. “We’ve now got it down to around about half an hour.”
Which Software Vendors Survive the Next Shift
Yesberg’s view of where the accounting tech stack is headed is blunt. He believes software providers that open their platforms to large language models will keep their customers, and the ones that stay closed will struggle. It’s less about any single AI feature and more about whether other software, scripts, and agents can talk to a given system at all.
He draws a direct line to the desktop-to-cloud shift he lived through earlier in his career, at the first QuickBooks Connect events. The tools that adapted survived, and the ones that resisted didn’t. He expects the same sorting process to play out again, just faster, as more of the accounting stack becomes something agents interact with directly rather than something a person clicks through by hand.