How to Recommend an AI Bookkeeping Tool Without Looking Like a Reseller

15 October 2026·5 min read

Quick answer: Disclose any referral or commission relationship with an AI bookkeeping tool before you recommend it, in plain language, every single time. Then be honest about what the tool is actually good at versus where it still needs a human checking its work — categorisation errors and GST coding nuances don't disappear just because a dashboard looks clever. 📱

AI bookkeeping tools have genuinely changed what a small business owner can do themselves before anything reaches your desk, and plenty of accounting firms have quietly struck referral deals with the tools they recommend most. None of that is a problem on its own. 💖 What is a problem is when a client finds out later — usually by accident — that their accountant gets a commission every time they sign up, and nobody mentioned it. That one discovery can undo years of trust faster than any invoice dispute ever could.

What most accounting firms get wrong

There are two opposite failure modes here, and both are common. The first is over-hyping the tool — presenting it as 'set and forget' when in reality it still misclassifies transactions, struggles with multi-entity structures, and needs someone checking the GST treatment on anything unusual. The second is staying quiet about the commercial relationship entirely, either because it feels awkward to mention or because the firm assumes the client won't ask. Both erode the thing clients are actually paying for, which is judgement — not a software subscription with your name attached to it.

The three-line disclosure script

Say this (or your own version of it) every time you recommend a tool you have any commercial relationship with, before the client signs up — not after:

1. 'Just so you know, we have a referral relationship with [Tool] — if you sign up through our link, we may receive a commission.'
2. 'That's not why we're recommending it. We're recommending it because, for a business like yours, we think it genuinely saves time on the basics.'
3. 'It's not a replacement for us reviewing your accounts — it's good at the first-pass sorting, and we'll still be checking the bits that actually affect your tax position.'

Put the same three points in writing somewhere the client can revisit — an onboarding email or a short tools page — so the disclosure isn't only ever spoken once and forgotten.

Sole trader cafe owner: the AI tool was genuinely a good fit for daily receipt capture and bank feed matching, cutting a few hours a week of admin. The firm still flagged, in writing, that end-of-quarter BAS review would remain a human job because the tool had previously miscoded supplier rebates as income.

Multi-site trades business: the tool handled receipt categorisation reasonably well for one job site but started blending costs across sites once a second ute and a second crew were added — exactly the kind of limit that needs naming before the client relies on it for job-costing decisions.

NDIS service provider: a genuinely complex chart of accounts meant the AI tool's default categorisation was wrong often enough to be a liability rather than a time-saver, until someone manually retrained the rules. The firm recommended it with that caveat attached from day one.

How to build a disclosure-safe recommendation process

Write the disclosure into a short internal policy so every team member says the same thing, rather than leaving it to individual judgement on a busy day. Put the commission arrangement in writing somewhere a client could reasonably find it again — a tools page, an onboarding pack, or an email — not just a verbal mention in a meeting. Keep a short, honest 'what this tool is good at / what it still needs a human for' note per tool you recommend, and review it every six to twelve months as the software changes, because AI bookkeeping tools are genuinely improving and today's limitation might not be next year's. Finally, build a regular review cadence into the engagement so clients understand the tool reduces admin, it doesn't remove the need for your eyes on the numbers.

💡 Never let the tool make the judgement call alone. GST coding, related-party transactions and anything touching Division 7A territory should always get a human look, no matter how confident the AI's suggestion looks on screen.

Please note: general information, not tax advice — check current official guidance before relying on it.

Mistakes to avoid

Don't hide a commission or referral fee — disclose it before the client signs up, not if they happen to ask later.
Don't present any AI bookkeeping tool as a full replacement for a bookkeeper or for your own review.
Don't recommend the same tool to every client regardless of entity structure or complexity — some setups genuinely aren't a good fit yet.
Don't skip onboarding support — a tool recommended and then abandoned creates more mess than it saves.
Don't assume the tool's categorisation is correct just because it looks confident — check it, especially in the first few months.

Frequently asked questions

Do we have to disclose a referral commission even if it's small?

Yes — the size of the commission isn't really the point, trust is. A small, disclosed commission rarely bothers a client. An undisclosed one, discovered later, damages the relationship regardless of the dollar amount involved.

Are AI bookkeeping tools actually reliable?

Honestly, it depends on the business. They're genuinely good at repetitive, pattern-based tasks like bank feed matching and receipt capture for straightforward entities. They're noticeably less reliable with unusual transactions, multi-entity structures, or anything needing judgement about GST treatment — which is exactly the territory where human review still matters most.

Should we recommend a tool we don't personally use in-house?

Be cautious here. It's harder to speak honestly about a tool's real limits if your own team hasn't used it day to day, and clients can usually tell the difference between a genuine recommendation and a scripted one.

What if a client specifically asks whether we get paid to recommend it?

Answer plainly and immediately. A hesitant or vague answer to a direct question reads as evasive even when the underlying arrangement is perfectly reasonable.

Keep reading 🤍

Written by

Kate, founder of Chronically Online

I help Gold Coast and Brisbane businesses grow with branding, websites and marketing that actually works.

Work with me ✦