Which AI tools should a bookkeeping firm ignore?

Which categories of tools can a firm ignore?

The AI tools worth ignoring are recognizable by category, not by brand. The first category: tools that duplicate capability already included in the accounting software the firm pays for every month; QuickBooks Online and Xero already ship receipt recognition, bank rules and basic reminders. The second: tools that assume a desktop ledger installed locally; a firm running Acomba or Sage 50 on the desktop should hear an honest vendor say “we would not be the right choice for you today” rather than be sold a fragile workaround. The third: anything that requires the firm to become its own IT department to install, connect and troubleshoot the tool. The fourth: generators that impress in a demo but connect to nothing in the real flow of client files; brilliant text that never enters an engagement gives no one an hour back. The pattern across all four categories is the same: judge the tool by its connection to the work, never by its demo.

What actually deserves a bookkeeping firm’s attention?

A bookkeeping firm’s attention belongs in the reverse order of novelty. First, the capability already sitting in the accounting software it pays for monthly: bank rules, document recognition, built-in reminders; turning those on costs nothing and often solves part of the problem outright; many firms discover they were about to buy features they already owned. Second, one AI assistant genuinely mastered rather than several half-tried: Claude, for example, covers drafting, summarizing and explaining for the whole team, and one assistant properly installed, properly taught and actually adopted beats a collection of sleeping subscriptions. Third, automations wired into the firm’s real workflows: document follow-ups, sorting and filing, data entry, deadline tracking. An automation that lives inside the flow of client files returns non-billable hours; a tool that lives in a browser tab returns none, however good its demo looked.

What makes a verdict on tools neutral?

The filtering described above has a name: curation.

Curation, in the context of business software and AI, means filtering a crowded market down to a verdict for one specific organization: what to use, what to ignore, and in what order. The valuable half is the negative advice: the tools not to buy, and the cases where capability a business already pays for is sufficient. It is vendor-neutral only if the person giving it is willing to recommend against their own products, which is what separates curation from a sales pitch with a diagnostic attached.

The test is easy to run: a firm can ask any vendor what it should not buy, including from that vendor. The most useful answer sometimes fits in one sentence: “the capability already included in your accounting software does this very well; use it, and pay no one.” A vendor who cannot say that sentence is delivering a pitch, not curation.

In what order should a firm adopt what remains?

The order of adoption matters as much as the choice of tools. First: turn on what the accounting software in place already does, because it is paid for and immediate. Second: install one AI assistant for the team and take the time to master it on real client files, not on the surface. Third: automate one workflow at a time, starting with whatever eats the most non-billable hours, prove it, then extend. A firm that follows this order spends little, learns fast and accumulates no dead subscriptions; it also keeps the team on side, because each step proves itself before the next one asks for trust. In Québec, Agentica delivers this verdict on a free 30-minute video call, with no system access: what to use, what to ignore, in what order, including what not to buy. The firm leaves with a written plan it keeps, whether or not it ever hires us.