AI for accountants and bookkeepers

Artificial intelligence is here to stay.

But while the tech world moves fast, many accountants and bookkeepers are still figuring out where AI fits into their practices, services, and day-to-day work.

At Accountex London 2025, the mood was clear: you want to use AI but you might not have the confidence or know how to use it.

Whether you’re just starting or experimenting with AI tools, this guide is for you.

Step 1: Understand the gap between awareness and adoption

At Accountex London, Chris Downing, director of accountants and bookkeepers at Sage, laid out where the profession stands and what needs to happen next.

He asked, “Who’s heard of AI tools like ChatGPT, Microsoft Copilot, or Sage Copilot?”

Nearly every hand in the room went up.

“Who uses these tools weekly? Daily?”

The number of hands dropped sharply.

This sums up the state of AI in accounting: high awareness, low adoption.

It’s not a lack of interest; it’s a lack of clarity.

You may have tried AI tools but if the results were mixed and outcomes hazy, the practical next steps may seem uncertain.

Yet the opportunity, clearly set out in Sage report The Practice Of Now, shows that there is recognition of the power of AI, machine learning, and automation to transform insight.

According to the report, 92% of accountants and bookkeepers said they spend too much time on manual admin and compliance tasks, which could be drastically reduced using AI.

But while many have started their tech advancement journeys, others haven’t. By building capability today, you will be ahead of the curve.

Don’t think about the technical side so much. Rethink how you approach AI. Chris said, “AI is your assistant. It won’t automate you—it will amplify you.”

AI isn’t here to replace accountants—it’s here to empower you. Instead of treating AI as a novelty or a threat, treat it as a long-term strategic partner.

Here are some mindset shifts you can take forward:

Know your ‘why’

Consider why you want to use AI in the first place. Is it about saving time? Improving client communication? Growing your practice?

Chris says, “You need to be laser-focused on where you will play first. Don’t try to do everything—focus on one area and learn from it.”

Being clear on your goals helps avoid AI being bolted on with no real benefit.

Involve the right people

AI isn’t just for tech enthusiasts or senior partners—it should be part of everyone’s toolkit.

Consider your internal stakeholders and how to bring the whole team on the journey. Chris says you should ask: “Who are the stakeholders? Who’s involved? Who needs to be kept informed?”

Some clients may have already asked you about AI, and if your team isn’t trained or aware, you risk being caught off guard.

Start small

Rather than overhauling everything, pick one use case and test AI in a safe, low-risk environment. That could be admin, marketing, communications, or internal reporting.

Chris explains, “You get so excited about the opportunity that you start doing too many things… But it’s about learning by application, then scaling up.”

Choose your tools wisely

Some accountancy practices are jumping straight into general-use Large Language Models (LLMs) such as Claude, Gemini, or ChatGPT.

Others are exploring integrated tools like Microsoft Copilot or Sage Copilot, which can sit inside their existing systems.

“Standalone tools are great, but you must consider your data, team, and structure. Pick the right tool for the right problem.”

Chris Downing, Director of accountants and bookkeepers, Sage

Step 3: Understand what AI can do for you today

At Accountex London, Chris Downing gave live demos of how AI can make everyday accounting work faster and easier.

Here are some of the examples he shared:

• Bank reconciliation: review your bank statement with an AI tool and ask it to reconcile. It can quickly spot problems such as transposition errors or missing transactions.
• Writing client letters: want to explain Making Tax Digital (MTD) for Income Tax to a group of clients? Ask AI to draft a letter in your voice, including key deadlines and actions. You can even ask it to make the tone more formal or more casual.
• Summarising technical content: take HMRC guidance, feed it into an AI tool, and ask for a plain-English summary. This is great for client comms, internal training, or marketing.
• Creating LinkedIn posts: feed in a blog, prompt, or thought-leadership idea and ask the tool to create five LinkedIn posts for different audience types.
• Spotting spreadsheet errors: AI can look at your spreadsheets and help identify unusual patterns, highlight anomalies, or suggest formula fixes.

Prompt engineering 101

The better your prompt, the better the result. Chris suggested that if you’re using a general-use LLM, you could try structuring prompts like this:

“I am an accountant at a UK-based firm. A client has asked how Making Tax Digital for Income Tax will affect them. Draft a short, friendly email explaining what Making Tax Digital is, what deadlines apply, and what actions they should take.”

You can also say things like:

• “make it more casual”
• “turn this into a LinkedIn post”
• “summarise this in bullet points”

The AI will respond accordingly.

Step 4: Prioritise skills, training, and new habits

Chris clarified that the biggest barrier to using AI effectively isn’t technology, it’s people.

He said, “AI won’t take your job. But someone using AI might.”

And that shift starts with skill building. You may have tried AI tools and been underwhelmed, not because they don’t work but because you don’t know how to use them well enough.

There are three common problems with AI:

• Not knowing what to ask: AI responds best to specific prompts, not vague questions.
• Giving up after a bad result: one weak response doesn’t mean the tool is useless—just that your prompt or context needs refining.
• Failing to review or edit outputs: AI gives you a draft, not a final answer. Human judgment is still essential.

That’s why you need to consistently carve out space for learning. Make AI training part of your routine. Do you dedicate time every week to upskilling?

Treat AI like any other business tool—something that requires continuous investment. Even just 30 minutes a week can transform how confidently your team uses AI.

Try these training methods:

• “AI Fridays”: a regular, low-pressure session where team members try new prompts or share tips
• Internal champions: nominate one or two AI-savvy team members to lead learning and gather best practices
• Peer coaching: create a Slack thread or shared doc where people post prompts and outputs that worked for them.

“Just get stuck into something that isn’t client-facing, that isn’t sensitive client data… Forget drafting a P&L or sending an email directly—just ask it to draft a reply. See how it feels.”

Jordan Smith, Operations Director, AG Smith & Co

Embrace informal learning

Chris shares his own learning habit. He said, “For my own training, I go to TikTok. There are people dropping videos every day on ChatGPT, Copilot, and other tools.”

Don’t underestimate platforms such as TikTok, YouTube, or Reddit.

Bite-sized, visual examples can be more effective than long webinars. Combined with product-specific training, they can form a complete toolkit for upskilling.

The key is habit formation. To succeed with AI, you must normalise experimentation, curiosity, and internal sharing—not just buy licences and hope for the best.

Step 5: Set guardrails: Policies and responsible use

AI is powerful, but with power comes risk. Don’t shy away from that.

Chris says, “You need to put policies in place. People use these tools without permission, training, or understanding the risks.”

The biggest concern is that you—and your clients—may be feeding personal or financial data into open LLM tools without safeguards.

Chris warned about scenarios in which users might paste sensitive data, such as a bank statement, into an open AI tool without realising where that information could go.

“That’s the challenge,” he said. “These tools are incredibly powerful, but that means we have a responsibility.”

While tools like Sage Copilot are built with data security in mind, it’s still critical to educate your clients about what data should and shouldn’t be used in any AI environment.

Start by creating basic AI guidelines

Your practice doesn’t need a 20-page policy to start. A simple one-pager can go a long way:

• What tools can we use?
• What data is off-limits? (for example, client names, bank details, ID numbers)
• How should AI-generated outputs be reviewed? (must be checked by a human before being sent externally, for example)
• When do we disclose AI usage to clients? (such as in reports, emails, and marketing comms)

Be proactive—not just about how you use AI, but how you talk to your clients about it; Chris says: “You’re being inquisitive about AI—but so are your clients. And who are they going to ask? You.”

Sage Copilot: Built-in guardrails

Note the advantage of integrated AI tools such as Sage Copilot, which operate securely within your accounting software.

This reduces risk because:

• Data stays within the system
• Access is permissioned and traceable
• Outputs are context-aware and tailored for finance workflows.

Even if you’re starting with free or general-purpose tools, the long-term goal should be clear: move toward secure, embedded solutions that support your financial workflows without adding exposure.

Don’t stifle usage—guide it

Don’t be scared off—but be responsible as well as curious.

Chris warns, “We can’t lock AI down. If you do, people will still use the tools without training, support, or policy.”

The better approach is to set boundaries, offer training, and create a culture that supports safe experimentation rather than punishes it.

Step 6: Use AI for support with MTD, growth, and resilience

AI isn’t just a productivity tool; it’s a strategic enabler.

Whether you’re navigating regulatory change or trying to scale without burnout, AI can support your practice in these key financial areas:

Making Tax Digital (MTD) for Income Tax: Simplify compliance, deepen relationships

MTD for Income Tax will require practice to manage quarterly submissions, segment clients by digital readiness, and deliver consistent communication. AI can help streamline each of those tasks.

“You can’t treat MTD like a background task anymore,” Chris warned. “With AI, you can test messaging, generate emails, and prepare comms in minutes. That’s time you get back for client conversations.”

Using AI to support MTD not only lightens admin, it gives your team time to be proactive with clients instead of reactive to deadlines.

Growth: More clients, better service, same headcount

Practices that embrace AI early see tangible time savings on low-value work such as email drafting, reconciliation, and meeting prep.

That saved time translates into real business benefits:

• You can handle more clients with the same staff
• You can spend more time on advisory or strategy
• You can reduce turnaround times on deliverables.

“Just imagine how many more clients you could manage if the basics were automated,” Chris said. “That’s where the real value starts to show up.”

Resilience: Adapt faster, flex easier

When a crisis hits—or even if it’s just a deadline pile-up—AI can act as a shock absorber to:

• Generate responses quickly
• Catch spreadsheet errors in seconds
• Summarise data, policies, or client histories on demand.

“Coronavirus showed us how quickly things can change,” Chris reminded the audience.

“AI won’t predict the next crisis, but it will help you adapt when it comes.”

“If AI frees up time spent inputting data and being doers, we can spend more time being controllers, reviewers, and advising.”

“We can be more strategic with what value we’re adding and what services we offer—adding more value to clients without having to hike the prices through the roof.”

Jordan Smith, Operations Director, AG Smith & Co

Step 7: Get started on your AI journey today

Don’t wait for the perfect tool or perfect time.

Start small, learn fast, and scale smart. This isn’t about launching an AI “initiative” but building confidence through low-risk experimentation.

Here’s a simple plan to begin with:

Choose one admin-heavy task

Pick something repetitive, time-consuming, and easy to test.

Examples:

• Drafting monthly client updates
• Writing MTD for Income Tax reminder emails
• Creating internal meeting summaries
• Reconciling small spreadsheets
• Preparing short LinkedIn posts or blog intros.

The task should be frequent enough that time savings matter but not so critical that failure causes issues.

Start small, stay safe, build trust.

Try solving it with AI

Run the task through a tool of your choice:

• ChatGPT, Claude, or Gemini for creative generation, summaries, or spreadsheet logic
• Microsoft Copilot for Excel/Word/Outlook productivity
• Sage Copilot for accounting-specific workflows in your existing Sage software.

Write a short prompt. Feed in sample data or context. Review the result.

Don’t expect perfection—look for potential.

Reflect on the outcome

Ask yourself:

• Did the AI save you time?
• Was the output accurate or close?
• What would you do differently next time?

Chris encourages a “learning by doing” approach. Prompt tuning is a skill; like any skill, it improves with feedback.

Top tip:

Keep a simple document with successful prompts or use-cases. It can grow into your team’s AI playbook.

Share, repeat, and scale

Don’t keep wins (or failures) to yourself.

Share examples with your team, especially across roles. Encourage others to try similar experiments.

Once you’ve tested a few tasks successfully:

• Formalise them into workflows
• Document your own AI best practices
• Consider using secure, integrated tools to scale usage across client work.

This isn’t a one-off project—it’s the beginning of a habit.

Final thoughts: AI won’t replace you, but it will change your role

Now you understand the seven steps needed to successfully use AI in your practice.

By adopting AI thoughtfully, you’ll be more efficient and valuable.

You’ll have more time for advisory.

You can move your clients off spreadsheets and manual records more quickly, getting them onto cloud accounting systems and digital workflows. You’ll be ready for whatever comes next.

Chris ends with a reminder that applies to all accountants and bookkeepers: “AI is happening. So is MTD for Income Tax. You need a strategy for both.”

Start now, stay curious, and treat AI like a teammate.

Because in the future of accounting, the real competitive edge won’t come from knowing about AI—it’ll come from using it well.