Insights, guides, and updates
Thomson Reuters spent $40 million putting its archive inside its own AI model, and the final training run cost under $450,000 because it built on a model anyone can download. Part 5 of Buy vs. build closes the series with what that means for a 12-person practice: the knowledge Thomson Reuters put inside a model, your firm puts on top of one, in a page of written-down rules. Two moves, one engagement, this month.
A Xero-affiliated Instagram ad told 77,000 followers that Claude had replaced an £800-a-month accountant, and the profession decided Xero was going the way of Intuit. The vendors going direct is arithmetic, and the profession has been outraged about that arithmetic since QuickBooks Live in 2019. Seven years later, the invoice still describes the work as a report.
An Instagram ad with a Xero discount code told 77,000 followers that Claude had replaced an £800-a-month accountant, and Xero apologised twice without saying it couldn't be done. The same week the AICPA said a CPA's fee has to price in the liability, Intuit's CFO called human accountability the thing AI can't take from tax and said Intuit sold it for $150, and BILL's CEO said if his software saves you work, you should pay him for it. Nobody said hours.
Puzzle, the $60 AI-native ledger we priced against QuickBooks two weeks ago, is selling its firm-facing business to Accrual and will stop marketing software directly to CPA firms. Also this week: OpenAI launched GPT-6 with a financial-statement tie-out as its lead customer story, Docusign is opening its connector to every AI agent on September 30, and a services firm priced an agent-led assurance engagement 30 to 50% below its own rate card.
An AI agent was asked to attach a spreadsheet to an email. It couldn't reach the folder, so it found a file with the same name in an old email, attached that, and reported the job done. In a CAS practice the same failure looks like a flawless margin analysis built on the pre-adjustment trial balance, and none of the checks your team was trained to run will catch it.
BILL is repricing AI-agent work from seats to consumption, Xero says one in five ecosystem connections is now a custom-built app, and frontier model prices dropped again. Part 4 of Buy vs. build re-runs the series' five-layer framework against a month of on-the-record evidence, and the question has moved: not whether firms will build, but what's bounded enough to own.
