One area where those risks are beginning to surface is complaints. There are two related issues for firms to consider. The first is complaints caused, or partly caused, by a firm’s own use of AI. The second is the growing use of AI by clients when drafting complaints. Both raise practical challenges, but neither changes the core principle: a complaint remains a complaint and should be handled through the firm’s usual process, in line with professional obligations and the terms set out in the engagement letter.

When AI use leads to a complaint

AI does not shift responsibility away from the firm. Whether advice has been supported by a traditional research process, specialist software or a generative AI tool, the firm remains responsible for the work it issues to the client. That means firms need to be clear about where AI is being used, how outputs are reviewed, and what level of reliance is appropriate in each context.

An obvious complaint risk is incorrect advice. AI-generated outputs can appear authoritative, but still be wrong, incomplete or based on sources that are not appropriate for the client’s circumstances. The safeguard is not to avoid AI altogether, but to ensure that staff are trained to use it properly and that outputs are scrutinised before they are relied on. This includes sense checking the answer, considering the sources or data being used, and applying professional judgement to whether the conclusion is reasonable. It is worth remembering that any member of staff working in an ICAEW member firm has free access to Bloomsbury Accounting and Tax Service which provides a trusted and credible source of information to support technical work, as well as suites of helpsheets and factsheets produced by ICAEW teams.

In addition, the FRC has published Generative and Agentic AI Guidance specifically aimed at audit testing, but may be of interest for non-audit engagements.

Another risk is over-reliance. AI can be very effective at summarising large volumes of information, such as board minutes or client records, but it may miss points that a trained professional would identify as significant. The question for firms is not simply whether the tool is useful, but whether it is being used in a way that is proportionate to the task and the risk. Human oversight remains essential, particularly where the output informs advice, assurance work or client decisions.

A final risk is inappropriate use by team members. Staff may use public AI tools, or AI features within software, in ways that do not align with firm policy. This can create confidentiality, data protection and quality issues, particularly if identifiable client information is entered into tools that have not been approved by the firm. Firms cannot control every individual action, but they can demonstrate that they have taken reasonable steps: clear policies, practical training, appropriate supervision, and documented review processes.

AI-generated complaints can feel different

The other side of the issue is the use of AI by clients when making complaints. Anecdotally, this appears to be becoming more common, and it reflects a wider societal shift. People now use AI to draft everyday communications, appeal decisions, write reviews and structure arguments. There is nothing inherently wrong with a client using AI to help explain why they are unhappy.

However, AI-generated complaints can feel more difficult to deal with. The tone may be more formal or legalistic than previous correspondence from the client. The complaint may include a long list of issues, some of which are marginal or repetitive. In some cases, the real concern may be obscured by the volume and style of the wording. For a practitioner who has worked with the client for many years, that change in tone can be unsettling.

The important point is not to become distracted by how the complaint may have been produced. A complaint should not be discounted or treated as less valid because AI may have been used to draft it. The client is still expressing dissatisfaction, and the firm still needs to understand what has happened, what the client is seeking, and how the matter should be investigated.

Focus on the substance, not the tool

Where a complaint appears to have been drafted using AI, it may be helpful to take the process back to basics. Ask yourself, what is the client actually unhappy about? Is the issue about the scope of work, the timing of advice, fees, communication, an error, or an unmet expectation? Is the complaint about one defined matter, or several linked issues? Separating the substance from the style can help the firm respond proportionately and avoid being drawn into unnecessary complexity.

In some situations, the simplest and most effective response may be to speak to the client. A phone call or meeting can help clarify the root cause of the complaint and may reduce the risk of escalation. This does not replace the need to keep proper records or to follow the firm’s complaints procedure, but it can help ensure that the written process is focused on the right issues. If the client prefers to continue in writing, firms should aim to respond clearly and calmly, avoiding an unnecessarily defensive or overly legalistic tone unless the circumstances require it.

It can be tempting to respond to an AI-generated complaint with the same level of formality, but that may not always help. Clear, plain English can be more effective: acknowledge the complaint, identify the issues the firm understands to have been raised, ask for clarification where necessary, and set out the next steps in the complaints process. This approach helps preserve professionalism while keeping the focus on resolution.

Practical steps for firms

Firms should review their approach to AI through the same professional lens they use for other risk areas. This includes considering whether engagement terms explain the use of AI or software tools in service delivery, whether limitations and client responsibilities are sufficiently clear, and whether confidentiality and data protection provisions remain appropriate. Engagement letters should be tailored to the specific services provided and reviewed when there is a material change in how work is delivered.

Internal governance is equally important. Firms should define what AI tools may be used, for what purposes, and with what safeguards. Policies should be practical enough for staff to follow, rather than simply listing restrictions. Training should help people understand why review, confidentiality and professional judgement matter. Review processes should be documented where AI outputs contribute to client advice or other significant work.

Firms may also want to consider how complaints teams or partners will identify and manage AI-related issues. That does not mean trying to detect whether a complaint has been written by AI. In most cases, that will not be the right question. The better question is whether the complaint is clear enough to investigate. If it is not, the firm should seek clarification, record the steps taken, and continue to apply its normal process.

The complaints process still matters

Ultimately, AI should not change the fundamentals of complaints handling. The firm should follow the process set out in its engagement letter, investigate the matter appropriately, communicate clearly with the client and keep adequate records. Where AI has contributed to the underlying issue, the firm should also consider whether there are wider lessons for its policies, training, supervision or quality controls.

The growth of AI may make some complaints look and feel different, but it does not remove the need for a human, professional response. For members, the priority is to avoid being distracted by the technology, focus on the substance of the concern, and demonstrate that they have acted reasonably and in line with their obligations. Where firms are unsure how to respond, they should seek support early. Sometimes the most useful step is simply to talk through the facts, test the proposed response, and make sure the process remains fair, proportionate and focused on resolution.

Members can seek assistance on how to deal with a complaint from the Technical Advisory Service team.