Only 10% of my clients have auto-reconciliation switched on. That's the wrong number.
I had a look through my stable of Xero files recently. Just over 10% have Automatic Bank Reconciliation turned on.
Ten percent. For a feature that has already auto-reconciled more than 100 million transactions globally and that Xero says is saving time for close to three in four of the customers using it.
I know why the other 90% are off. I've heard the reasons, and I've said some of them myself. It coded a transaction wrong. It got the same supplier right one week and wrong the next. Fixing its mistakes took longer than doing it myself. So it got switched off, and that was that.
Here's my view: switching it off after two weeks is the wrong call. And here's the number that makes me say it — Xero's own figures, shared with me through my Xero Partner support contact, say that around 95% of users who persist with the feature for at least a month, decide it’s worth having.
That's the whole argument in one statistic. Almost nobody who gives it time goes back. The people switching it off are, overwhelmingly, the people who tried it for a fortnight.
It's a new tool, not a finished one
Automatic Bank Reconciliation is still in beta. Xero says so themselves — they're openly still working on the harder cases, like one payment covering multiple invoices.
Judging a tool that is openly still learning by whether it is perfect today misses what it actually is. This isn't a bank rule that either fires correctly or doesn't. It's a system that gets better as it sees more of your data and more of your corrections.
Think about how you onboard a new bookkeeper. Week one, they get things wrong. They code the power bill to the wrong account, they don't know that the payment from that one customer always covers three invoices, they miss that the “Smith” in the bank feed is a different Smith. You don't give up in week one. You correct them, they learn the business, and by month three things have settled.
Auto-reconciliation is the same deal. It just needs teaching. The difference is that most of us have never thought of software as something you train.
What teaching it actually looks like
This isn't hand-waving. The feedback loop is real and it's specific:
Correct it in place. You can now edit the account code or tax rate directly on the Reconciled page without pulling the whole reconciliation apart. Every one of those edits tells JAX something about how your organisation actually works, and it feeds into future suggestions.
Check how it decided. The Reconciled page shows you the method behind each transaction — whether it came from a match, a bank rule, a memory, a prediction, or from you doing it manually. Filter for the auto-reconciled ones and you can review them as a batch rather than trusting blind.
Give it good foundations. It works alongside the bank rules you've already built. A file with tidy rules, consistent contacts and a sensible chart of accounts will get better results than a messy one. Rubbish in, rubbish out — that hasn't changed.
Start with one account. It's enabled per bank account, so you can turn it on for your simplest, highest-volume account, watch it for a month, and go from there.
One thing worth knowing: it only acts when it's confident. When it isn't, it leaves the transaction sitting there for you. It isn't guessing wildly at your file.
Where I'll be blunt
Switching it on does not mean reviewing less. Anyone who tells you it's set-and-forget is selling you something.
For our clients, the work shifts from reconciling to reviewing — and for a GST-registered business heading into a return, the review is not optional. A wrongly coded transaction is your problem, not Xero's, and your name is on the return either way. If a client is going to switch it on and never look at the bank rec again, they are worse off, not better.
There are also files where it genuinely isn't ready yet. High volume, lots of one-off suppliers, part-payments across multiple invoices, a document workflow where receipts land days after the payment — the correction time can outweigh the saving right now. Test, measure, and if it isn't working on that file, turn it off on that file. That's a data-driven call, not a philosophical one.
The bigger point
There is no version of this where we wait quietly and the perfect tool arrives. It doesn't work like that. These tools get good because people use them, correct them and complain loudly about the gaps.
If you're sitting it out until it's flawless, you'll be starting from zero the day everyone else has a trained system and two hours a week back.
And remember what the people who pushed through actually did next: they kept it on. That tells you the problem was never the tool. It was the two-week trial.
I'd rather be in the group learning alongside the tool than the group waiting for a finished product that never ships.
Try this: pick one bank account this month. Turn it on. Review everything it does for four weeks and correct it properly rather than undoing it. Then decide.
If you want a hand deciding which client files are good candidates, that's exactly the conversation we like having.