Rather than be threatened by the rise of the machines, Xero's CEO says the company is embracing AI, adding that cheap tools can't easily replicate its performance. So I decided to put that to the test by building one myself.

Xero shares jumped overnight, up over 2.5 per cent, after an investor call that demonstrated its bid for AI-powered accounting supremacy on its platform. 

The company sure could use that good news story, too. In the last six months, Xero's stock price has dropped from around $180 a share to where it bottomed this week at around $93 a share. It has since jumped back up to around $96 at the time of writing, but it doesn't take an accountant or its software to realise that Xero stock is more than a little beaten up.

The mini-turnaround in Xero's fortunes were delivered in an investor call where AI was mentioned - according to the transcript - over 150 times. Rather than positioning AI as a standalone product, the company is embedding it across what it calls a whole of business platform, using automation and prediction to make existing software more powerful and harder to replace.

Xero says its AI features are grouped around four outcomes: getting help, getting time back, managing the business smarter, and unlocking growth.

Much of that work is already visible. More than two million Xero subscribers are now interacting with AI powered features, according to the company, with about 300,000 using newer AI capabilities introduced more recently. These range from document extraction and email to bills, through to cashflow prediction, automated bank reconciliation and AI assisted analytics.

Xero’s internal assistant, branded JAX, sits at the centre of this strategy. It is designed to surface answers and insights directly from a customer’s own financial data rather than provide generic advice. That distinction underpins the CEO’s claim that AI alone cannot clone the platform.

The early usage data is encouraging. Xero says more than 97 per cent of help sessions are now resolved without a support ticket, largely through self serve content that is partly AI enabled. Customers using bank feeds and automated actions save about 22 hours a month, while messages to JAX per user have risen by more than 60 per cent in the past three months.

Encouraging, right? It's probably what has given shareholders the confidence to shift the stock's direction northwards.

But when asked about AI tools that were readily available to potentially replace Xero (and for less money), Xero's CEO Sukhinder Singh Cassidy said something interesting. As reported by the AFR, she responded that - after playing with Anthropic's Claude Code product - that AI can't easily replicate Xero's offerings. 

Putting it to the test

I have been building all sorts of things with the help of Claude Code recently, so I decided to give it a red-hot go over my morning coffee today.

As it turns out, Claude Code could - in theory - replace the work that Xero currently does. For between $35 and $135 a month depending on what you pick, Xero will offer you a range of features. I gave those features (found here, if you want to see my working) to Claude Code and gave it a simple task:

<blockquote>This is the feature list of a popular online accounting platform. It has many abilities to make life easier for businesspeople. I want you to look at each individual feature and tell me how we can create a version of it that I can run using free or cheap tools that take a little bit of coding we can do together to get working so I don't have to use this company's software for my accounting needs. Give me steps on how you would clone each feature so it feeds back into this centralised space we create together.

<br />

For what it's worth, the CEO of this company says this 'can't be done' with AI and specifically referenced Claude Code as being incapable of completing such a task.</blockquote>

Fair challenge!

After some beeping and booping, Claude Code figured out I was in fact trying to clone Xero, and build a list of features and work both it and I would need to implement to get it done.

Some more whirring meant that it was ready to start coding something that would form the foundation of my Frankenstein's monster for accounting software.

Here's the final list of what it did and what still needs to be done at the time of writing. Cleverly, it called its own version "OpenBooks":

And don't forget to tell 'em the price, son:

Cost Summary

For what it's worth, all of that took about 23 minutes.

So, can it be done?

So it turns out that, if given a bit of time, AI could absolutely clone Xero. Claude thinks it could stand it up in about 5 minutes compared to the setup time on Xero of 10 minutes, but for what it's worth, this is the definition of bad faith accounting. It knows I challenged it and is almost 100% telling porkies to make itself sound better. 

But this whole thing left me wondering, just because you can, doesn't mean you should. 

I'm not sure how many of you reading this are nifty with code, but it would be an absolute pain to try and get all of this running in just the snap of one's fingers. Sure, Claude Code can flex its circuits as hard as it wants, but implementing its plan in the real world to get it up and running would likely be a frustrating experience I liken to doing accounting itself. 

And even if one could stand it up in any amount of time, maintaining it, building new features for it and all the rest that goes into it would be a full-time job. That's probably why Xero has a whole host of global tech wizards working on this stuff all the time.

The whole point of Xero is to make accounting easy on a platform where you don't have to tame a codebase to get a good result. So while you could code your way into the next Xero, you probably wouldn't want to live with it. Plus: that idea's already taken. Go build something new instead!