Most growing accounting firms now run a stack of SaaS subscriptions where half the tools are write-only. Data goes in. Nothing comes back out into the workflow that matters. The meeting transcription tool nobody opens. The workflow app that doesn't talk to practice management. The "AI assistant" that emails a summary to a folder no one searches.
Each one looked like productivity at the point of sale. Each one is now costing the firm money and attention.
A recent Xero case study on UK firm Shapes is being shared as a productivity success story, 140% revenue growth, 30 hours a week saved, a six-person team punching above its weight. Look past the tools Shapes bought and the real story is the question they asked before buying them. Most firms skip that question. Their app stacks turn into silos as a result.
The firms moving fastest aren't the ones with the most tools. They're the ones who learned to ask one question before buying any of them.
The question
Does this land where the work already happens, or does it create another place my team has to remember to check?
It sounds obvious written down, and almost nobody applies it before the demo call. The vendor shows you the feature. The feature is impressive. You buy. Six months later the tool is still in use, the subscription renews, but the information it produces never reaches the client record where decisions get made.
Shapes applied the test out loud. The Xero case study shows what happens when a firm gets it right, and on a closer read, where the limits of that story are.
What Shapes did
Shapes is a six-person UK firm founded in 2023 by Jim Leeves (the full Xero case study is here). Service mix: accounting, systems implementation, and marketing under one roof. They run 15 to 20 client meetings a week, heavy for a team that size, and were losing four to five hours a day across the team to meeting admin: writing up notes, chasing actions, reconstructing conversations after the fact.
Jim's line in the case study lands the real problem:
"Not only can I not be across everything, but I got to the point where it's unhelpful to be across everything as well."
The bottleneck was him, not capacity. The firm couldn't scale while the institutional knowledge lived in his head and his notebook. Most growing firms hit this wall and hire. Shapes changed where the knowledge lived.
They built around Xero, moved practice management to Karbon, and added Vinyl, an AI meeting tool that records, transcribes, and pushes summaries directly onto the Karbon client record via native integration. Not exported. Not emailed. Onto the record, where the rest of the work happens.
Jim is direct about why Vinyl beat the alternatives:
"If we're going to have all of our client information in Karbon anyway, and we've got an integration that the others definitely wouldn't have provided, that's going to work really well for us."
The silo test, applied in real time before anyone called it that. Other tools may have been better at transcription. They lost on the only question that mattered.
What the numbers don't tell you
Shapes reports turnover up 140% year-on-year with a similar-sized team. Jim estimates Vinyl saves each team member an hour a day, about 30 hours a week across the team.
The 140% is unaudited, self-reported, and published without a baseline. For a firm founded in 2023, it could mean any number of absolute revenue figures. Jim himself credits "the right clients and the right tools", the stack didn't generate the growth, it removed a ceiling.
The hour-a-day saving is an estimate. Plausible, but not a measurement.
The story holds up as one founder's account, sitting on top of an unusual setup: a firm doing implementation and marketing alongside compliance, which generates far more discussion-heavy meetings than a traditional firm. A compliance shop running quarterly BAS conversations won't see the same gains.
What the case study skips
Two claims the original piece reports without scrutiny.
"Not a single client pushed back on meetings being recorded." In the UK, that sits inside GDPR. In Australia, it sits inside state-based surveillance and listening device laws that vary by jurisdiction, plus Privacy Act obligations on storage. Universal client acceptance is possible. It's also the kind of claim that holds up until the one client where it doesn't. Any firm copying this needs a documented consent flow.
"Adoption was immediate." Vendor language. Real adoption is never immediate. Someone forgets to turn it on. Someone hates the first summary. Someone has a meeting they didn't want transcribed. Shapes may have had a smooth run. The absence of friction in the telling is the part to be suspicious of.
What to do on Monday
None of those caveats break the logic. They mean the test matters more. The Shapes stack is UK, but Xero, Karbon, and Vinyl all operate in Australia. The stack translates. The test translates better.
Three questions to ask any vendor before signing:
Where does the output land? If the answer is "we email a summary" or "we generate a report you can download," that's a silo. If it's "it writes directly into the client record in your practice management system," that's an integration.
Who has to remember to check it? If the answer involves your team building a new habit, the tool fails in month three when the habit lapses. If the answer is "nobody, it appears where they already look," the tool has a chance.
What happens to the data when a staff member leaves? If it lives in their inbox, their notebook, or their head, it walks out the door with them. If it lives on the client record, it stays. This question exposes most "AI assistant" tools immediately.
Apply those three to whatever's already in your firm's app stack. The tools that fail aren't necessarily bad. They're doing work in the wrong place, which costs more, over time, than the subscription line on the invoice suggests.
Jim Leeves' closing line in the case study: "You won't be disappointed. You'll regret that you've waited this long." The honest version is shorter. The firms pulling ahead are buying tools that talk to each other, and walking away from the ones that don't.
You can read the full Xero case study on Shapes here.
The test is free. Skipping it is the expensive part.
