Thought Leadership

Most of what you’re being sold is a wrapper

Many of the AI tools targeting accountants and tax advisors are built on top of the same underlying models (GPT, Claude, Gemini) with a layer of prompting and sector-specific context added. Some of these are genuinely good products. The developers know their domain, they’ve thought carefully about security, and there’s real value in the customization.

But the barrier to entry has fallen so far that plenty of others are not. In reality, you can now build a functioning tax tool in an afternoon with no coding experience. That’s worth remembering when a company presents itself as a serious AI platform.

The due diligence questions have changed as a result. Features matter less than they used to. What matters more is:

1. Where exactly is client data going when it leaves your system?
2. What’s actually running underneath?
3. Are there security professionals involved or is this essentially a well-prompted chatbot with a logo on it

Some vendors will have good answers. Many won’t, and the ones who go quiet when you ask are probably telling you something.

The billing model problem nobody wants to talk about

One firm at the session raised something that got an uncomfortable laugh. They had been using AI for due diligence work, the kind of job that used to take a senior person the better part of a day. With the right tools it now takes under an hour. The output is comparable. So, what do you charge?

Most firms are quietly absorbing the efficiency gains while the billing model stays the same. That works for a while. It stops working when clients start doing their own arithmetic, or when a competitor decides to make the price difference visible.

The shift that’s coming is away from billing time and towards billing for judgment.

What you’ll be charging for is the expertise required to know what to do with the output, catch what the AI missed, and take professional responsibility for the advice. Firms have been talking about value-based pricing for years without much urgency. AI is about to give them some.

Governance is where the advisory opportunity actually lives

The tendency is to treat the governance side of AI, such as data protection, client consent, EU AI literacy requirements and ensuring AI-assisted work is auditable, as the boring part you deal with before you get to the “interesting” stuff. The discussion in Warsaw suggested that’s the wrong way round.

Clients are already asking about this, and not because firms have been raising it. They’re arriving with their own questions about data privacy, about whether they can rely on AI-assisted work, about what their advisors are actually doing with their information. One participant put it well:

“We didn’t push for it. We just showed what we were doing and clients came.”

There’s a service line in this that most firms haven’t named yet.

Firms that sort out their own AI governance properly are in a much stronger position on the pricing question too. If the tools are available to everyone, the differentiator becomes being the firm whose judgment a client trusts enough to pay for.

What firms are actually doing

The Warsaw session had a useful mix of where people are.

• One firm has an accountant (not an IT person) who built an internal policy tool using AI in an afternoon; it’s now how the whole team looks up procedures.

• Another is experimenting with chaining small AI tasks together into review workflows, testing each step before connecting it to the next.

• A third spent around €5,000 on an on-premises server running an open-source model, which keeps everything inside the firm and sidesteps a lot of the data questions entirely.

While all these firms are still experimenting with AI, they’re all making deliberate choices rather than just adopting whatever arrives in their inbox.

The approach to avoid is moving without thinking - by signing up for tools you haven’t scrutinized, by not revisiting how you charge for work that now takes a fraction of the time, and treating governance as someone else’s problem until it isn’t.

The technology question will keep changing, but the business questions underneath it won’t: what are clients actually paying you for, how do you demonstrate it’s worth it, and who do they trust when the tools are the same everywhere?