By Ariel Harmoko, CEO - Artifact AI

The prevailing narrative in accounting AI is this: AI will handle the routine work, accountants will become reviewers and advisors, the profession will evolve. It's a clean story. It sounds like progress. And there is some truth in it.

But follow it to its conclusion and it describes something closer to a gradual collapse. 

Our thesis is different. Accountants should not merely become reviewers. They should become doers of fundamentally different work.

The Apprenticeship Problem

Judgment in accounting is not abstract intelligence. It is accumulated experience. You learn to reconcile correctly by reconciling incorrectly and being corrected. You learn to spot the anomaly in a client's payroll because you've worked through enough payrolls to know what normal looks like for this client, in this industry, with this ownership structure. Remove the doing, and you remove the mechanism through which the profession reproduces expertise.

That is exactly what shifting accountants to mere “reviewers” leads to. A junior reviewing a completed reconciliation sees the match, not the signals that informed it. Reviewing correct outputs does not build the same judgment that making – and correcting – errors does.

The honest fact is that the traditional apprenticeship was already failing before AI arrived. Between 2019 and 2022, 300,000 accountants left the profession – not just because AI disrupted their learning path, but because the learning path itself was gruelling. Manual reconciliations, data entry across disconnected systems, close-season hours. The work that builds expertise also drives people out.

So the profession faces a double bind. The "doers to reviewers" path breaks the learning model without replacing it. The traditional "doers" path is not sustainable. And with 75% of CPAs eligible to retire within the decade, the profession isn't just losing capacity. It is potentially losing the accumulated expertise that capacity carries.

If there is a third path, it needs to be one where the work is different enough to sustain the learning, but compelling enough to retain the people. It needs a layer of technology that not only automates, but fundamentally transforms the work.

Capturing Wet Code

With the traditional route to accounting expertise under threat, the instinct is to capture what it produced before it disappears.

The most valuable thing a firm possesses isn't the data in its systems of record. That data is structured, exportable, and increasingly accessible. What's genuinely valuable – and genuinely hard to replicate – is the human decision layer on top of it. The context of why the data looks the way it does. What precedents apply to which clients. The accumulated experience and intuition of senior accountants, built over years, across many clients, in specific industries.

When a client pays for a senior accountant's time, this is what they're paying for.

Ultimately, a firm’s accumulated experience – built across decades and hundreds of clients – exists as what we call 'wet code': unwritten rules, client-specific logic, and contextual knowledge that sits between systems and determines how accounting actually gets done.

How much revenue do you recognise on a half-finished job with an unsigned change order? The answer isn't in the ERP. It's in a WhatsApp conversation eighteen months ago between the partner and the client's CFO. A client's consolidation logic lives in a spreadsheet only one person understands. Every new client a firm onboards compounds this: another tech stack, another body of wet code that must be carried in someone's head, translated across the firm, and passed down to someone who wasn't there when the decisions were made.

The industry’s response to capturing wet code has addressed the symptoms rather than the underlying problem. Building internal engineering teams takes years and scales headcount rather than building leverage. Forcing clients onto a single platform simplifies the firm's stack but alienates clients and ignores the reality that every new engagement brings its own tools. And point solutions — AI tools that automate inside individual systems — make the fragmentation actively worse. Without an orchestration layer, each one adds its own integration surface, its own data model, its own exception handling. The overhead at the seams between tools often exceeds the labour they saved. All of these approaches work inside systems. The wet code that lives between them remains un-captured.

What the profession needs is an orchestration layer that connects the work across a firm's and its clients' tech stacks, workflows and systems. And a way of working where wet code captures itself as a natural byproduct of doing the work.  

What Accountants Will Become With Omni

This is the shift we've been building toward. Not automating the old work faster, but transforming the work itself — and with it, the nature of accounting expertise. In this world, the new form of expertise is orchestrating agentic workflows across a firm's entire client base.

Today we launch Omni to make that possible. Omni enables accountants to build, run, and produce cross-platform, multi-step workflows across any tech stack — all using natural language. The workflow you build for one construction client on Sage can be templatized and deployed across every construction client, configurable for specific needs in minutes. Workflows are reusable and reconfigurable: per client, per industry, per department.

This is how wet code gets encoded as a natural byproduct of doing the work. The senior accountant in this world becomes what we call the "Accounting Engineer". Their value is not in executing workflows, but in knowing which ones to build — and that judgment comes from something no model can replicate: the accumulated context of working deeply inside many clients’ business, across every department, over many years. Every workflow they design through Omni, every dashboard they build, every exception they handle feeds their understanding of how the client's business actually works – where the margin pressure comes from, which revenue lines are fragile, what the cash flow patterns mean for hiring decisions.

A firm that spends the next five years moving juniors into review roles and nothing else will have multiplied throughput – but it will not have captured any new expertise that will survive the future of the profession, and will face a cliff when its seniors retire. The fastest path is not always the best path.

The long term vision is not a profession that depends on its AI vendor's implementation team. It is a profession that can build for itself. And that is why Omni does more than just execute. It becomes the training ground for a new way of working.

The next decade belongs to the firms that stop optimising for what accounting was and start building for what it can be.

Join the Waitlist