Glossary
How Is AI Being Used in Accounting?
AI is being used in accounting today to enter invoices, reconcile accounts, pull audit samples, build working papers, and flag fraud risk on payment runs, all with a human reviewing exceptions rather than doing the data entry. It is not replacing judgment. It is replacing the repetitive steps that used to consume most of a bookkeeper’s or junior accountant’s day. At Treewalk, we build these tools ourselves rather than waiting for software vendors to catch up, because the accounting logic, not the interface, is the hard part.
What AI is actually doing right now
The gap between “AI in accounting” as a headline and AI in accounting as a working system is bigger than most vendor pitches suggest. Here is what is actually shipped and running, not theoretical:
Accounts payable automation.
We run an in-house AP agent that reads invoices, applies coding logic across roughly a dozen decision points, and only stops for a human when it hits a new vendor or an error. It runs below human error rates and is deployed across our client base.
Audit-sample and CRA-request pulls.
A tool that automatically pulls the specific invoices and documents an auditor or the CRA requests directly out of QuickBooks, instead of a staff member digging through the file.
Continuity and working-paper schedules.
One of our senior team members built a tool that rolls forward a share-capital continuity schedule from a single sentence of instruction in about 25 minutes, and a related skill that flags discrepancies in a client file without a human touching it first.
Payment-run fraud checks.
A tool that scans an outgoing payment batch against verified bank details and flags anything changed or missing before the file goes to the bank, cutting a task that used to take up to 45 minutes down to nearly instant.
Why accounting is a different AI story than marketing
Most AI hype comes out of marketing, where the pitch is “generate more content, reach more people.” Accounting does not work that way. We see this play out constantly: you always want more marketing, but nobody wants more accounting. Nobody wants a bigger accounting department. They want the same accuracy for less effort. That means AI in accounting is fundamentally a cost and speed problem, not a demand problem, and it gets judged by a much less forgiving standard: people tolerate a human making a mistake far more readily than they tolerate a computer making one.
What AI still cannot do
A single invoice carries somewhere between 80 and 100 small human decisions: which GL account, which tax treatment, whether the vendor name matches a prior entry despite a typo, whether the amount is reasonable given history. That is why you cannot paste a chart of accounts into a general chatbot and expect a usable answer. Building AI that actually works in accounting means encoding that judgment, not just automating a form.
The hard part about building these tools is the accounting logic, not the software. There is no getting eighty percent of the way there and calling it done.
There is also a security layer that did not exist when a person did the work by hand: prompt injection hidden in invoice text, and increasingly convincing fake-vendor phishing. That is a direct reason human sign-off still applies to new vendors and amount anomalies, even in a heavily automated workflow.
The role is shifting, not disappearing
The accountant of the next several years looks less like someone keying entries and more like someone supervising digital workers: reviewing exceptions, catching fraud patterns, and applying the judgment a model does not have. That shift rewards people who understand both the accounting logic and how to direct a tool, and it puts a premium on the trust layer a CPA designation represents. Adoption, not the technology itself, tends to be the limiting factor. We have seen internally that a tool can be faster and more accurate than a person and still sit unused until the workflow around it changes, not just the software.
Frequently asked questions
Will AI replace bookkeepers?
It is replacing the most repetitive parts of bookkeeping, invoice entry and basic reconciliation, faster than it is replacing judgment-heavy review work. Bookkeepers who move into exception handling, client communication, and reviewing AI output tend to stay relevant. Pure data entry is the part actually at risk.
Will CPAs be replaced by AI?
Unlikely to be replaced outright, but the day-to-day work is changing quickly. Compliance-only, low-judgment tasks are the most automatable. The value of the CPA designation shifts toward oversight, advisory judgment, and being the trusted human who signs off when a model cannot.
What is the 30 percent rule for AI in accounting?
There is no single official standard here. It shows up informally as a rule of thumb: automate the roughly 30 percent of a workflow that is repetitive and low-risk first, and keep human review on the remaining high-judgment work. Treat it as a starting heuristic, not a fixed benchmark.
Can accountants still earn well as AI takes over routine work?
Yes, but the highest earners tend to be the ones moving up the value chain, into advisory, fractional CFO work, or transaction advisory, rather than staying in pure compliance. AI accelerates that shift by making routine work cheaper to deliver, which pushes value toward judgment-based services.
Is this the same as buying an off-the-shelf AI bookkeeping app?
No. Off-the-shelf tools solve narrow slices of the problem and often do not fit Canadian tax rules like input tax credits. We build and adapt our own tools around the actual accounting logic, which is why they hold up on messy, real files.
Where to next
If you are trying to figure out what AI could realistically take off your plate this year, whether that is accounts payable, month-end close, or reporting for a fractional CFO relationship, our team can walk through what is actually working versus what is still marketing. Reach out to Avnit Sekhon at avnit.sekhon@treewalk.com to talk it through.
