On March 2, Xero's updated Developer Platform Terms took full effect. The new rules explicitly prohibit desktop agents, ban using API data to train AI models, and require API partners to commit — contractually — to not building desktop automation tools. Simultaneously, Xero shifted from free API access to tiered pricing starting at $99/month, with high-volume integrations running $895/month or more. One developer, RecHound, projected annual costs of $17,340 under the new model.
The timing isn't subtle. In the same period, Xero launched JAX — Just Ask Xero — a proprietary agentic AI built with OpenAI, designed to handle bank reconciliations, data entry, and payment processing. The message: we'll build the AI. You use ours.
Chad Davis, CPA put it bluntly on LinkedIn the same day: "I wonder sometimes what it looks like when the cost to access something is greater than the benefit, or worse, it's taken away completely and we're forced to use whatever AI tool they're pushing." He's talking about Xero, Intuit, project management tools, sales tools — the entire cloud stack that modern practices depend on.
He's right. And if you run a CAS practice, this is your problem now.
Next Friday, March 20 at 3pm Eastern, I'm going deep on Sequoia Capital's thesis that your practice is a $50–80 billion disruption target — and what to do about it. A full hour. Free. Register at theaiaccountant.ai/webinar.
The wall you're already hitting
Here's what this looks like in practice. In Xero, there's a report called Uncoded Statement Lines — it shows every bank transaction that hasn't been posted to the ledger. Basic operational necessity. QuickBooks Online has no equivalent report. The data exists inside QBO, but it's not exposed through any API endpoint and there's no native report to surface it.
We needed that report. So we built it ourselves using browser automation through Claude Cowork. It works. But we had to navigate multiple roadblocks — authentication barriers, missing API coverage, platform behavior that actively resists external tooling — just to get a list of transactions that the platform already knows about.
That's not an edge case. That's the daily reality of trying to use AI in a modern CAS practice. The platforms that hold your data aren't opening the door to external AI. They're closing it.
AI in the gaps
This is what I call AI in the gaps — deploying AI in the spaces between your core platforms, where human effort still dominates. Not replacing Xero or QBO. Not waiting for their proprietary AI to mature. Instead, identifying every workflow where your team is currently the integration layer — reading, interpreting, typing, copying, deciding, communicating — and asking whether AI can do that work faster and more consistently.
The gaps are enormous. Your team drafts client emails, writes report narratives, chases missing documents, prepares tax organizers, builds meeting briefs, reviews categorized transactions, creates SOPs, and answers client questions about their own financials. None of that happens inside your accounting platform. All of it happens between platforms — in email, in Word documents, in spreadsheets, in the heads of your experienced staff.
That's where the ROI lives. Not behind an API wall. Not inside a vendor's proprietary AI. In the work your people do every day that no platform controls.
Last week I made the case that AI can already replace your accounting software for a meaningful slice of compliance work. This is the other side of that argument. For the work that stays inside Xero or QBO, the platforms are actively restricting your ability to bring your own AI. Which makes the work that happens outside those platforms — the gaps — your highest-leverage investment.
The compounding effect
The transformation isn't in any single workflow. It's in the compound effect of applying AI across dozens of them simultaneously. When you automate the monthly report narrative and the client email and the meeting prep brief and the variance analysis and the review checklist — the cumulative savings change the economics of a client engagement. You're not saving 10 minutes. You're reclaiming hours per client per month.
And every client context document you build, every SOP you structure, every prompt you refine becomes machine-readable institutional knowledge. New staff onboard faster. Quality improves. The firm compounds its intelligence in a way no vendor can gatekeep.
AI-native accounting tools are in development — platforms built from the ground up around AI capabilities rather than bolting them on after the fact. They're worth watching. But they're not stable enough for production use in your practice today. The gaps are here now. So is the AI to fill them.
Your move
Your accounting platform has decided that AI access to your data is their opportunity, not yours. You can wait for them to deliver on that promise, or you can start building where they can't stop you.
I've mapped 100 specific workflows across 20 categories where AI works in the gaps of a CAS practice — scored by effort and impact, no API access required, no platform integration needed. Download the free AI in the Gaps Toolkit and start with three. Then pick three more.