Accountants and bookkeepers know that bank reconciliation can be one of those fiddly jobs that eats up time.
You’re clicking through transactions, second-guessing whether to accept Xero’s suggested matches, and checking bank rules. What if there was a faster way?
That’s the idea that sparked Beau Gaudron from Growth Wise to experiment with building a Chrome extension.
After spotting a post from Enzo on LinkedIn, Beau decided to dig deeper, learn how Chrome extensions work, and see if AI could lend a hand.
The result was a working demo: an extension that plugs into Xero’s bank feed screen, sends transaction data to a language model, and returns a reasoned response about whether to reconcile or not. Let’s walk through how it came about.
Finding inspiration on LinkedIn
The story started with a screenshot. Beau explained, “Nat Lennon had sent me a screenshot of a LinkedIn post that was talking about reconciling Xero using a Chrome extension.” Intrigued, he looked up the author, Enzo, and read more about what had been built.
That sparked a question from Nat:
how hard is it to build a Chrome extension?
Beau admitted, “I’ve never done one, I have no idea, but let’s go and find out.”
Later that day, Beau discovered that Automation Town had already featured Enzo in a video explaining his approach. Even better, Enzo generously shared insights directly when Beau reached out. That encouragement gave Beau the push to try it himself.
Building the extension
Beau wasted no time setting up a test inside the Xero demo file. On the bank feed screen, a new ninja icon appeared in the top-right corner, a personal touch for the extension.
Clicking the icon revealed a playful note:
the time has come to push the buttons
. A bit of fun, but also a signal that something practical was about to happen.
With one click on
press to check matches
, the extension connected to a large language model. Transaction data from the Xero page was passed through, and within moments the model returned a screen showing potential matches, reasonings, and confidence weightings.
What the AI looks for
Rather than trying to reconcile every single transaction blindly, the AI focused on cases where Xero had already provided some guidance.
Beau explained, “the OK button will only show if there’s already a suggested match, so that’s what we’re targeting with this.” The aim wasn’t to replace human judgement entirely but to make reviewing suggested matches quicker and clearer.
The extension displayed different reasoning types:
• Xero suggested coding - matches pre-filled by Xero’s own prediction models.
• Bank rule detected - cases where an existing bank rule, like a recurring 7-Eleven transaction, provided confidence.
• Other suggestions - based on patterns seen in the data.
This gave the user a better sense of why a transaction might be correct, rather than just blindly accepting suggestions.
Reconciling at the click of a button
Once the matches had been reviewed, the user could click a single button to reconcile them all at once.
As Beau showed, “it’s gonna say execute 6 selected transactions, click OK, and now you can see in the background, they’re all reconciled.”
For a demo, it was simple but effective. In practice, this could save accountants from repetitive clicks, while still letting them stay in control of the reconciliation process.
Why share experiments like this
Beau was quick to point out that this was just a basic demonstration. The extension was built as a learning exercise, inspired by someone else’s idea. But that’s exactly why he wanted to share it.
“I thought it was fun to share how things that are posted by other people can affect other people,” he said. By seeing Enzo’s original post, Beau was motivated to try something new, and now others can take inspiration too.
It’s a reminder that sharing experiments, even small ones, can spark creativity across the accounting community.
Final thoughts
Reconciliation in Xero might never be glamorous, but tools like this show how AI can reduce the friction. By plugging directly into the bank feed, pulling data into a language model, and reasoning through suggestions, accountants can move faster and with more confidence.
This experiment proves you don’t need to wait for official features. With curiosity, collaboration, and a bit of tinkering, it’s possible to build your own shortcuts.
So here’s the question: what’s stopping your firm from building something clever to make the day-to-day grind easier?
