At a glance
- AI adoption faces barriers, but simple tools can help shrink your workload.
- AI excels at language tasks like summarising, but always needs human supervision.
- Use AI to transcribe meetings, summarise contracts, and even generate Excel code.
- Experiment with different AI tools to discover what works best for your practice.
If you’re in an accounting practice, odds are that you’re already using artificial intelligence (AI) tools to get jobs done. But in its June 2025 final report, the UK government’s SME Digital Adoption Taskforce suggests three key barriers are slowing SMEs’ adoption of digital technology such as AI:
- Adoption often looks too hard and costly.
- SMEs lack expertise and execution support.
- Switching from one technology to another can feel too risky.
In one sense, AI is now inserting itself into our business processes all the time anyway. The “Accounting Agent” in QuickBooks automatically categorises transactions, reconciles books, and detects anomalies. Google’s Gmail service summarises long email threads.
But you don’t need to have someone build an app in order to move AI into new parts of your practice. One possibility: adopt relatively simple AI processes with publicly-available tools to help move staff up the AI curve. That may not only aid the business now, but also spur the firm’s people to come up with new ideas for using AI more creatively.
The following ideas have been informed by business interviews, media reports, online discussions and in some cases, Financial Accountant’s own experience, over a number of months. Many of the ideas have a common theme: AI can’t do the whole task, but it can shrink the size of the workload you face.
Before we start, three reminders:
- LLMs have produced a jump in AI results, but the letters stand for “large LANGUAGE model”. AI shines at summarising and simplifying long, complex texts and discussions, and at pulling together disparate ideas. Despite being a computing product, AI does poorly at maths.
- AI works like a big team of enthusiastic interns. It can get through a lot of work fast, and excels at pattern recognition, but it stumbles over nuance, ethics and context. It needs a human eye to run over its work, checking for errors.
- Those enthusiastic robot interns often seem like they’ve popped over to the pub for a few quick pints. Many LLMs still seem to hallucinate at times, littering their factual results with often genuine-sounding but inaccurate notions.
That said, here are nine ways accountants are leveraging AI in their daily operations.
AI your meetings
We can start with one AI use you’ve almost certainly discovered – transcription. A number of AI tools will transcribe your meetings as they happen, summarise key discussions, and list action items. ChatGPT, for instance, has a useful “record mode” for paid subscribers which will delete the recording after transcription.
Or you can use a dedicated transcription tool. Among the most highly rated such tools is Otter.ai, which integrates with most of the popular video-conferencing platforms. Just be sure to get recording permissions from all the meeting’s participants.
If you’re feeling guilty about not taking good meeting notes, this is your simple way to fix the problem.
AI your Excel code
AI may not be great at maths, but it’s fine at code. It won’t get Excel code right first time every time. But if you’re out of ideas on how to get a particular result, try quickly describing your task to ChatGPT and seeing what it comes up with.
AI your mail
Mail? No, not email – your email provider is already providing AI for that. We’re talking here about the stuff in paper envelopes that you may still be receiving. If for whatever reason you still get a pile of physical mail, you can automate that by scanning it and having AI summarise it for staff response.
AI your explainers
If you have a group of documents from which you create responses to particular circumstances, it’s simple to provide those documents to an AI engine and have it formulate responses to the questions you feed it.
AI your contracts
If you need to work your way through a complex legal agreement or contract, in particular, you can get a head start on the work by having AI summarise it and extract key points. You’ll still need to work, because the AI is likely to miss or misunderstand some points. But AI can make the exercise less daunting.
AI your complex documents (into conversations)
Most of us have to review complex documents – and it’s inevitably often difficult to get them into your head. Firms including Google (via its NotebookLM product) now offer to turn such documents into “conversations” or “podcasts” that you can review while walking the dog.
AI your thinking time
The “thought partners” phrase is overused, but you can use many of the most popular AI tools to bounce ideas off during research. You’ll probably need to experiment to find the tool and process that suit you best. This is one place where a conversation with an AI engine may work for you.
AI your customer problem-solving
Record your customer support conversations – getting customers’ permission first – and transcribe them. Then you can put AI to work analysing the transcripts for common themes and generating advice, while your people spend their time analysing the results.
AI your tricky conversations
You can use an AI to take a trial run at a difficult dialogue you expect to have with partners, customers, suppliers or colleagues. Not everyone likes this, but some people find it tremendously helpful. Many advanced users do it. But you don’t need to be an expert to talk with your AI.
Experiment with your AI
Whatever you end up doing with AI, it’s worth experimenting with ideas to see whether they work, and how you can make them work better.
And different AI engines produce different results, depending on the task and the data available. You can use the AI available to you – Gemini for Google users, CoPilot because it’s built into Office, ChatGPT because it’s the one your firm’s people have most experience with. But you can also experiment to find out which AI does well at which task. One alternative is the relatively affordable Perplexity, which gives you access to most of the major AI alternatives and lets you repeat the same query using different engines.
AI successes are unpredictable; whichever engines you use, you can get great results on one task and nothing useful on another. But AI is also cheap and easy – the perfect circumstance for running experiments.
Gain insights on implementing AI in your practice and the latest updates from industry specialists by revisiting the IFA AI and emerging technologies conference online available on demand here.
