The deciding question is where your ledger lives. Since the Xero integration went live in May 2026, Claude can work against real client financials: outstanding invoices, cash position, revenue questions, answered from the books rather than from an export someone remembered to attach. Copilot has no equivalent view into Xero; it improves the Office documents around the ledger, not the ledger work itself.
Copilot earns its seat in a different way. If the whole practice is standardised on Microsoft 365, having drafting and summarising built into Outlook and Excel is friction-free in a way a separate app never is. The accounting-specific catch is the pricing shape: Copilot is an add-on per user on top of the base licence, so a ten-seat firm is paying for in-app convenience across every seat, whether or not that seat does ledger work.
For a Xero practice, start with Claude: the ledger connection plus spreadsheet and document skills covers the month-end grind Copilot cannot see. Add Copilot later if the in-app polish justifies per-seat cost. Firms locked into Microsoft accounting stacks read this the other way around. And when the goal moves from a better assistant to an automated practice, that is a build question, not a subscription question.