Artificial intelligence has the potential to save accountancy firms time, automate routine work and help teams work more efficiently.
But according to Alistair Barlow, founder and former CEO of Flinder, and AI developer Sebastian Triff, many of the biggest risks stem from how firms choose to use it.
Talking to TaxCalc TV, they argued that poor adoption decisions can create just as many problems as the technology solves.
As firms continue to balance AI adoption alongside the realities of Making Tax Digital, understanding where AI can help – and where it can't – is becoming increasingly important.
Watch the full webinar replay here.
Using AI because it's there
One of Alistair's biggest concerns is what he describes as a "scattergun approach" to AI adoption.
While some firms are being "very purposeful and deliberate", Alistair says others are using AI for "anything and everything" without first deciding what business problem they are trying to solve.
"What is the strategy I have in my firm? What is the vision? What am I trying to achieve?"
He compares it to having a hammer and then looking for nails.
"Just because we have a hammer, AI, we therefore go looking for nails to hit in."
For Alistair, technology should support a firm's strategy rather than define it.
"It comes back to the firm's vision. What are they trying to achieve, rather than using technology just for the sake of technology."
Why MTD is the test case
Making Tax Digital provides a useful example of where AI can add value – and where firms need to be careful.
Alistair recommends assessing activities based on two factors: how often they happen and the cost of getting them wrong.
For practices managing MTD obligations, that distinction matters.
Drafting client reminders, summarising information and answering routine questions are all potential candidates for AI assistance. Quarterly submissions, calculations and compliance obligations are different.
As Sebastian explains, AI models are probabilistic. Run the same request twice and you may get different answers.
"There's no such thing as a deterministic model," says Seb.
That doesn't mean AI is unreliable. It means firms need to think carefully about where they use it.
As Alistair puts it: "As accountants, we want something to be materially accurate."
For firms navigating MTD, AI may be extremely useful around the filing process, helping with communications, administration and workflow efficiency. But when it comes to submissions and compliance-critical outputs, accuracy and consistency become far more important than speed, Alastair said.
Watch the replay
Why AI-generated content often misses the mark
Seb says marketing is one area where firms frequently misuse AI.
Too many people, he argues, simply paste information into ChatGPT or Claude and ask it to create content.
"People just chuck a bunch of messages into ChatGPT or Claude and go, 'write me my LinkedIn post'."
The result?
"They sound flat and they sound boring."
For Seb, the risk is that firms lose the expertise, personality and insight that make their communications valuable.
"Can someone be bothered to actually write something that means something?"
The same lesson applies beyond marketing. AI can accelerate work, but that doesn't mean every output should be accepted without review.
Better inputs lead to better outputs
Many users focus on prompts. Seb believes context is often more important.
"The more context, the better."
AI can only work with the information it receives. The better the inputs, the better the chances of a useful output.
Poor-quality information can have the opposite effect.
As Seb puts it:
"Dirty data is dirty context."
For accountants, that means carefully selecting relevant information rather than simply feeding large volumes of data into a model and hoping for the best.
An assistant, not an autopilot
Seb is enthusiastic about AI agents and automation, but he warns firms against handing over too much control too soon.
"It's not very reliable yet."
Problems tend to arise when users expect AI to operate without oversight.
"People try to let it go full autonomous. Do everything for me. And the quality is not there."
For firms dealing with compliance obligations, client data and MTD workflows, both speakers believe the same principle applies: use AI where it adds value, but understand its limits.
The technology can accelerate progress. The firms most likely to benefit, however, are those that adopt it deliberately rather than simply because it's available.
.png)
