The general ledger (GL) platforms might not have much of a defensible competitive position anymore.
If you've built your firm's workflow around Xero or similar, this might be uncomfortable. But the evidence is stacking up and markets are pricing it in.
A couple of announcements landed a bit too close and caught my attention. The reaction from people I respect in accounting SaaS to my early thoughts made me want to dive a bit deeper into moats.[1]
A tale of two cities
A couple of weeks ago Xero announced a multi-year partnership with Anthropic (the company behind the wildly successful Claude.ai). It’s a very odd move to me:
- Xero switched from OpenAI to Claude in six months, so now JAX (the Xero copilot) runs on Anthropic’s models; there’ll be a Xero connector to Claude’s platform so that customers using eg. Claude Cowork can do planning and forecasting in there and not in Xero. Family dinners at Syft will be a bit awkward.
- Xero’s announcement also feels reactive: Intuit announced its own partnership with Anthropic in February. The scope differs and Intuit has a more diverse portfolio with QBO, TurboTax, CreditKarma and Mailchimp. Still, it looks like Xero’s busdev team grabbed whatever was in the pipeline to keep pace.
And this came in contrast with Intercom (a customer support platform) announcing the Fin Apex model: a custom, built in-house LLM, specifically trained for customer service on 15 years of their own data. Their model outperforms everything else available from the model labs: nearly 3% better resolution rates, faster responses and 65% fewer hallucinations. A vertical model, purpose-built for one thing, outperforming generic alternatives, for a fraction of the cost.
This is the kind of thing that gets the flywheel going: customers flock to Intercom for the ability to automate customer support efficiency, Intercom gets more data that gets fed into the model converting it into a proprietary advantage that benefits all Intercom customers - existing and new. Whomever comes after Intercom will have some real catching up to do now.
So while Xero announced an AI vendor switch for the second time in 6 months, Intercom showed what it looks like to build a software vendor’s own intelligence. The contrast is hard to ignore, at least for me.
To be fair, Claude is a genuinely outstanding product but partnering with someone else's intelligence is not the same as building your own.
Xero’s watching the sequel to a film they starred in
Not too long ago MYOB completely dominated the ANZ accounting market. Deep accountant channel relationships, tax compliance, switching costs that felt permanent. And then Xero launched…
Open APIs to foster an ecosystem of 3rd-party point solutions and an aggressive accountant channel strategy. Fast forward a few years, market shares inverted and Xero is now the leading player. KKR took MYOB private in 2019.
And a bit of an uncomfortable signal in all this: Xero’s cofounder Craig Walked joined Digits, an AI-native accounting startup. Craig called it “the pioneer of AI accounting” in the same paragraph as calling Xero “the pioneer of cloud accounting”.
Even if it’s not a verdict, it’s a signal worth paying attention to.
Does the GL have a data moat at all?
My thesis is that no, general ledger data, per se, is neither unique nor proprietary. Chart of accounts, transaction categories, reporting formats, all are fairly standardised across the industry and any startup can use the same models on the same type of data.
GL data isn’t as rich as Intercom’s customer service data. SMB income and expenses tend to fall into well-understood categories so a competitor should be able to achieve high levels of categorisation accuracy. And indeed they have.
However this doesn’t mean there are no switching costs. Xero still holds critical operational details that customers can’t export: bank rules, report configurations, activity logs, bank feed history, connected app configurations. There is no equivalent in Xero to a Google Takeout. Plus the changes to API pricing further raise the walls around the data.
But switching friction is not a growth flywheel. Craig’s example from inside Digits is “what took me a week to set up in Xero of Quickbooks, took me just minutes”. AI is compressing the switching friction that exists.
And this may be an important point: bank rules, categorisation patterns, reports, are things that AI can replicate easily.
Competitive advantage then moves into their large global base of accountant users, thousands of app integrations, bank feed agreements with tens of thousands of banks.
There’s another point I want to touch on: Xero’s been sitting on top of years of (non unique) financial data from millions of subscribers and… what have they done with it? I see no proprietary financial insights that only Xero offers. No benchmarking or predictive features that help a business owner achieve their goals.
Paraphrasing Satya Nadella, CEO of Microsoft, “a CRUD database with some business logic” and a nice UI.
The dumb pipes risk
Back in 2007, I was at a smallish mobile conference in Barcelona, and I remember that telcos were terrified of becoming “dumb pipes”. Their executives could see value shifting to the application layer. Well not even two years later Nokia and RIM/Blackberry market shares imploded, mobile portals died, telcos became invisible and the value moved to Apple, Google and app developers. Dumb pipes.
I think there’s a parallel story here. If an AI agent app like eg. Claude Cowork becomes the primary way small business owners or their accountants interact with their financial data, the GL becomes infrastructure. Users can ask Claude (just like I do now) to chase invoices, reconcile bank statements, forecast cash flow. Sure this all still goes through eg. Xero’s API, but its UI, brand relationship with the user and pricing power all surely erode.
So yeah, switching costs are real and layered, but MYOB had all these moats too.
And Xero’s share price reflects this: from ~A$196 to ~A$72 in 9 months. Down 63%. Some of it is the broader SaaS correction. Claude Cowork’s launch triggered what’s called the “SaaSpocalypse”, wiping billions from SaaS valuations. But the market’s pricing Xero at a lower EVR multiple than Intuit despite faster growth and higher margins. This is not a small thing.
Investors are not pricing in the death of accounting. It’s not an extinction event for the profession. They’re pricing in competitive pressure from hundreds of new entrants and the commoditisation of the GL platform. A structural repricing.
A thousand specialised Xeros
Another thesis of mine is that if the GL commoditises, the value won’t necessarily disappear, it’ll migrate.
The dominant, horizontal GL served an era of standardised workflows. However industrious have specific operational needs that horizontal GLs can’t meet at the ideal depth, eg. an e-commerce business needs multi-channel inventory reconciliation, whereas a hospitality business will need shift-based labour costing. Each needs distinct COA patterns, compliance requirements and operational workflows.
Vertical SaaS platforms like Toast or Shopify could end up deploying their own accounting via API-first ledgers, bypassing the standalone GL entirely. Vendors like Fragment, TigerBeetle and Open Ledger are building that plumbing.
One of the stronger drivers for the horizontal GL has been harmonisation, as accountants consolidate all (or as many of) their clients on a single platform for workflow efficiencies. But this depends on the accountant UI being the GL itself.
As accountants adopt agentic tools (again Cowork or similar), which can plug into the client’s system (eg. Toast or Shopify) via an MCP/API, then does the accountant really need every client to be on the same platform?
The agentic layer becomes the harmonisation layer. Accountants won't need every client on the same platform. Each client uses whatever operating system suits their industry and the accountant plugs into all these distinct systems via their AI agent.
If this does indeed happen, then a significant horizontal GL advantage disappears.
Roadmap for disruptors
Greg Sheehan, who cofounded one of Xero’s largest early accounting partners, shared a roadmap: agentic AI-native GL plugged into open banking, agentic AP/AR and agentic practice management. Then making it simple for firms to cut and switch.
His reframing of the accounting buy-decision is a sharp note: clients indeed don’t engage accountants to “do the numbers” but rather to transfer the anxiety of dealing with the tax authority. The number crunching will be gone, but the trust, interpretation and reassurance layer is what remains.
If that’s where the value sits, the moat can’t be in the ledger. It’s in the context layer: firm specific workflows, industry-specific logic, client relationships, regulatory nuance. This is where vertical AI earns its moat. And native AI codebases will be better placed to build this than GL codebases carrying decades of technical debt.
The demand is there even if the products are not - or not yet. LinkedIn is buzzing with accountants with budget wanting to buy agentic tools, and hundreds of millions in VC funding have poured into AI-native accounting startups: Basis, Digits, Pilot, Accrual and many others.
So the direction of travel is clear, even if the timeline isn’t. And this has a real impact on the platform decisions that firms are making today.
This isn’t just a Xero story though. Figma and Atlassian are under the same pressure. Xero just happens to be the example that’s closest to home, and the future[2] looks rough: become a dumb data storage for AI agents eg. Claude Cowork; get replaced by AI-native startups; slowly get replaced by more fit for purpose niche solutions; figure out how to generate value out of commodity data and protect its market share.
What this might mean for your firm
If you're a partner managing a few hundred clients on top of Xero or QBO, none of this means you need to do anything drastic tomorrow.
But the platform decision you made three years ago might be worth revisiting and reassessing where your firm's real value sits.
In my opinion, that value sits entirely in the context layer: industry knowledge, regulatory expertise, the judgement that no bank feed reconciliation can automate. AI amplifies this value, it doesn’t replace it.
The practical question for partners like you isn't "should I leave Xero?" It's "am I building capabilities that survive a platform transition?"
That's a question worth sitting with, even if the answer isn't urgent yet.
---
1- In software, a moat is a competitive advantage (switching costs, proprietary data, regulatory lock-in) that blocks competitors from taking away their customers even if they match every feature. It results in years/decades of market leadership, not just a few months.
2- when talking about "the future", it's helpful to keep a sense of time scale for these things. Everything is likely to take a few years.