‘AI native’ is everywhere. Dave Sellick is building his practice deterministically.

In accounting technology, “AI native” has become the label to have. Dave Sellick, Founder of Sidgrove, says software founders are told they can’t raise money without it. When he asks what the label actually means, his answer is short: “my sort of key point is no one really knows.”

In Dave’s reading, AI native usually means agentic AI at the core of everything. For accounting, he doesn’t think that is a good fit. Large language models are probabilistic, and he sees that as a weakness for accounting work.

Dave uses GenAI heavily in his own practice. His approach is to use it to build tools, then let those tools do the work in a predictable way. That distinction is the centre of his argument.

The pitch behind AI native is that AI agents should sit at the heart of the product and handle the work. Dave’s concern is with the technology underneath. A probabilistic model can give a different answer to the same question, which is a weakness when the output is a set of accounts.

He applies the same thinking to the ledger itself. Dave doesn’t see the need for a ledger with AI built into it. In his words, the core ledger tool should be “deterministic and simple and boring”, with AI connecting to it from the outside.

Dave argues that the only way to judge these tools is to use them on real client work, under real pressure. He has been running his own month end through a tool he built for the last couple of months, finding problems and fixing them as he goes. That hands-on testing is where his view comes from.

A popular approach right now is to connect AI tools straight into a firm’s software and data, so an AI agent can read and act across systems. Dave has clients who ask to connect AI tools to his app this way. His response: “I don’t really think we should be putting a roulette wheel on an API.”

His reasoning is practical. Dave says he has tried working this way and seen it make mistakes multiple times. In his view, a mistake is mathematically certain sooner or later. Once a firm sees the risk built into the approach, he expects it to stop working that way, and the time spent learning it is lost.

He also questions the usual promise that automation frees accountants up for advisory work. If the underlying data can’t be relied on, the advisory work built on it can’t be either. Dave compares it to telling someone to buy “ready meals so they can spend more time in the gym so they can get healthier.” The quick fix undermines the goal.

Over 10 years, Dave built his practice’s know-how into spreadsheets. GenAI has let him turn that work into a proprietary app that sits on top of his clients’ ledger data. The app acts as an orchestration layer, running workflows from the month end through to personal tax and deadline monitoring.

The app works from one set of logic. When Dave makes an improvement in one place, every client gets the benefit, without him updating a separate spreadsheet for each one. Clients review work through a branded interface, and notification workflows feed into his team messaging tool.

Some of the most useful pieces are simple. Clients get dashboards drawn from their own sales and financial data. A cash flow tool pulls in bank statement data automatically, giving every client a view of their daily cash position. For some workflows, the original spreadsheets still run in parallel as a backup.

GenAI plays a role in how the app develops. Dave has it suggest improvements, which he works through as he goes. He is also sharing the code and prompts behind the app’s features, so other firms can build similar tools into their own systems.

He expects many firms, though not all, to end up with a proprietary app of their own over time. For firms that find all of this too much, his advice is simple: “The key thing I want you to know is build deterministically.” If that still feels like a stretch, he suggests using chat-based AI tools securely and carefully as a copilot.

Dave’s argument rests on work he has actually done, which is why he prefers to show it. For readers, the session turns the argument into something they can use.

He starts by explaining what “AI native” actually means and why probabilistic technology is a weak foundation for accounting. He then shows the stack and tools he used to build his orchestration layer, and runs through each part of it on his demo company.

Firms that want to build will leave with more than ideas. Dave plans to share a page where viewers can preview the features he has built and take the code and a prompt to add any of them to their own tools.

If the technical detail goes over a viewer’s head, Dave is clear that it is there for inspiration. His advice for those firms is to keep using AI chat tools securely and carefully as a copilot.

‘AI native’ is a lie: here’s how I’m actually transforming my accounting practice with GenAI (live demo)

Dave Sellick, Founder of Sidgrove, explains what “AI native” usually means and why probabilistic AI at the core of accounting concerns him. He then walks through the orchestration layer he has built over his practice data, with a demo using his demo company. His session is a chance to see deterministic GenAI tools running the workflows he uses in his own practice.

Thursday 12 November 2026, 2:00pm AEDT

Register for the session

Part of The AI in Practice Summit, 11 and 12 November 2026. Free, online, every session recorded.