How to Get Your Accountancy Team on Board With AI (Without the Resistance)

Most practice owners who struggle with AI adoption are not struggling with the technology. They are struggling with the people. Choosing the right tool is the straightforward part. Getting a team of accountants to actually use it, consistently and correctly, is where most rollouts quietly stall.

This article is part of Runbook's complete guide to AI implementation for UK accountancy practices. What follows is a practical look at why staff resistance happens, what actually works to address it, and how to structure an AI rollout (including how to handle client-data rules) that your team can get behind rather than work around. If you are not yet sure whether your practice is ready to start, the free AI Readiness Scorecard gives you a clear picture in under five minutes.

Why resistance happens in accountancy practices

Staff resistance to AI in accountancy practices tends to cluster around three concerns, and understanding which concern is driving the resistance matters more than applying a generic change management framework.

The first concern is job security. Accountancy staff have watched automation replace roles in other industries and many assume, reasonably, that AI is another step in the same direction. This concern rarely gets voiced directly. It surfaces instead as scepticism about whether the tools are reliable, reluctance to engage with training, and a general preference for doing things the established way.

The second concern is competence. Using a new technology in a professional context, in front of colleagues and clients, carries the risk of looking incompetent. For experienced staff who take pride in their expertise, being asked to learn something unfamiliar in a work setting can feel exposing. The reluctance is about self-protection, not laziness.

The third concern is workload. Many accountancy teams are already stretched. When a new tool arrives without clear guidance, it does not feel like a time-saver. It feels like one more thing to learn on top of everything else. If the benefit is not immediately visible, the path of least resistance is to carry on as before.

Resistance that looks like stubbornness is usually one of these three things. Treating it as the right concern, rather than a general objection to change, is the starting point for addressing it effectively.

Involve staff before you introduce the tools

One of the most effective things a practice owner can do to reduce resistance is to involve one or two staff members in the selection process before anything is announced to the wider team.

This does not mean delegating the decision. It means identifying a staff member who is likely to be an engaged early adopter (someone curious about technology, relatively confident, and respected by colleagues) and giving them a role in evaluating the tools before a choice is made. Ask them to spend a week trying two or three options on real tasks and give you their honest view.

The practical effect of this is significant. When the rollout begins, you have an internal advocate rather than a top-down instruction. The rest of the team has a peer they can ask questions. And the person who did the evaluation has a stake in making it work, because they helped choose it.

For practices with more than ten staff, it is worth identifying one early adopter per team or department rather than relying on a single champion for the whole firm. The principle is the same at any size: involvement before introduction reduces resistance at introduction.

Start small and make the benefit visible

The biggest adoption mistake practices make is trying to introduce AI across multiple tasks and multiple staff simultaneously. The result is that no one develops genuine proficiency at anything, the time saving is diffuse and hard to see, and enthusiasm dissipates within a few weeks.

A more reliable approach is to identify one task, for one group of people, and run it for four weeks before expanding. The task should meet three criteria: it is high volume enough that staff encounter it regularly, it is low enough stakes that any errors are caught before they matter, and the time saving is visible enough that people can feel the difference within the first week or two.

Which tasks work best as starting points

Client email drafting meets all three criteria for most practices. A typical accountancy team sends dozens of routine client emails each week: appointment reminders, information request letters, standard responses to queries, covering letters for accounts or tax returns. These are formulaic enough for AI to produce a strong first draft, reviewed and sent in a fraction of the time it takes to write from scratch.

Meeting note summarisation is the other consistently strong starting point. Staff who attend client meetings can end the session with a structured summary ready to review rather than spending 20 to 30 minutes writing up notes by hand. The time saving is immediate and hard to argue with. Meeting notes should either be anonymised or processed only through an approved business tool covered by the firm's policy. Staff should not upload raw transcripts or client-identifiable notes into consumer AI tools.

What both tasks share is that the output is always reviewed before it reaches a client. That makes them safe places to build confidence, because mistakes are caught before they matter. Once staff have seen the tool produce genuinely useful output on a low-risk task, scepticism about whether it works tends to resolve itself.

Give people structure, not just access

Giving staff access to an AI tool and expecting them to figure it out is the second most common adoption mistake. It works for a small number of naturally curious, technically confident people. For everyone else, open-ended self-directed learning stalls within days.

What staff actually need is a short structured introduction, a defined starting task, and a clear answer to the question: what am I allowed to do with this?

What a useful introduction looks like

For most practices, a session of 60 to 90 minutes is sufficient to get a team started. It should cover three things: how the tool works in basic terms, the one task they will use it for first, and the firm's policy on what can and cannot be processed through it. The third element is not optional. Without it, staff default to caution and use the tool for almost nothing, or default to convenience and use it for things they should not.

The policy question matters particularly in accountancy because of the data protection implications. Staff need to know, clearly and in writing, whether they can use the tool with client information and under what conditions. This does not need to be a lengthy document. A one-page summary of approved tools, approved use cases, and the rule on client data is enough to give people confidence to use the tools without worrying they are doing something wrong. For a practical guide to writing one, including a template structure you can adapt, see how to write an AI policy for an accountancy practice.

For any use case involving personal data, the practice should record a basic data protection assessment and consider whether a Data Protection Impact Assessment (DPIA) is required. This should cover the tool used, the data entered, lawful basis, retention, access controls, human review and how client confidentiality will be protected.

Practices should also consider professional confidentiality obligations, engagement-letter commitments and client expectations before entering identifiable client information into any AI system. The issue is not only UK GDPR. It also touches on the professional duties your firm already owes to its clients.

Prompts are a skill, and staff need help developing them

One of the most common points of failure in team AI adoption is the gap between what staff expect the tool to produce and what it actually produces when given a vague instruction. A prompt like "write a client email about their tax return" will produce something generic and usually inadequate. A prompt that specifies the client's situation, the outcome to communicate, the tone required, and the length produces something genuinely usable.

Most staff do not know this at the start. The result is that they try the tool once, get a mediocre output, and conclude that it does not work. Providing a library of tested, specific prompts for the tasks your team actually does removes this barrier entirely. Staff get good outputs from the first session, which builds confidence and continued use.

The AI Implementation Checklist for UK Accountancy Practices includes a staff rollout section covering the policy, the introduction session structure, and the guidance your team will need to get started effectively. It is designed to give you everything in one place rather than building it from scratch.

Address the job security question directly

This is the concern that most practice owners handle least well, usually because they are uncomfortable with it or assume it will go away on its own. It does not.

The most productive approach is to address it early and directly, before the rollout begins. Not in a way that overpromises (telling staff that AI will never affect their roles is not credible, and being caught in an overpromise damages trust badly), but in a way that is honest about what you know, what you do not know, and what your intention is.

A straightforward message for most practices is this: the tasks AI handles well are the ones your experienced staff find least interesting and most time-consuming. Time saved on routine correspondence and meeting notes is time that can go into more complex work, better client relationships, and the kind of advisory work that is genuinely hard to automate. The practices that handle this transition well will have more capacity for the work that matters, not less need for qualified people.

That framing is honest. It does not pretend the landscape is not changing. But it positions your practice's AI adoption as something staff benefit from rather than something being done to them.

For practices where the concern is strongest, it is worth revisiting this conversation after the first four weeks of the pilot, when staff have first-hand experience of what the tool actually does. For many staff, abstract anxiety reduces once they see the tool being used for narrow, reviewed tasks rather than wholesale role replacement.

Sustaining adoption beyond the pilot

Getting a team through the first four weeks is the hardest part of the change management challenge. Sustaining and expanding adoption from there requires a different approach: less hand-holding, more structure.

Build AI use into existing workflows rather than alongside them

The practices that sustain adoption most successfully are the ones where AI use becomes the standard way of doing a task, not an optional extra. This means updating your standard operating procedures to reflect the new workflow. If the approved process for drafting a client appointment reminder now involves the AI tool, that is what the procedure says. Staff follow procedures; optional extras get used by some people sometimes.

Review and refine regularly

The first set of prompts your team uses will not be the best ones. Practices that build in a brief monthly review (what worked, what produced poor outputs, what tasks should we try next) improve faster than those that set something up and leave it running unchanged. This does not need to be a formal meeting. Ten minutes at the end of a team catch-up is enough.

Expand the scope gradually

Once a team has genuine proficiency on one task, the barrier to adding a second is much lower. The confidence is there, the policy framework is in place, and the scepticism has reduced. Add one new use case at a time, give it four weeks, and review before adding another. Practices that try to expand too quickly tend to see quality drop across all tasks rather than improving progressively.

For practices that want a complete framework for this, covering the pilot, the expansion, the policy, and the 90-day rollout plan, the AI Implementation Checklist for UK Accountancy Practices provides a step-by-step framework for taking a practice from first steps to firm-wide adoption, with templates and guidance at each stage.

The adoption challenges covered in this article apply wherever there is a team to manage, whether that is two people or twenty. Running a practice entirely on your own raises a different set of questions: the buy-in problem disappears, but the decisions about which tasks to start with, which tools to use, and how to handle the data protection groundwork are just as important. The guide to AI for sole practitioner accountants covers the practical starting point and the checks that matter most for a one-person practice.

Frequently asked questions