Publication

Managing AI complaints: Part one – A practical guide to dealing with AI-generated complaints

ailaw.jpg ailaw.jpg

Artificial intelligence (AI) has made it easier than ever to produce formal written complaints. With only a small amount of effort, generative AI systems, often programmed to agree with users, can create complaints that at first glance appear persuasive, detailed and legally sophisticated.

As a result, organisations are now reporting a rise in complaints attributed to such use of AI. These complaints often present practical challenges for those responsible for investigating and responding to them, as they are usually lengthy, repetitive, unclear on the underlying facts or contain legal assertions that do not assist in identifying the real issues requiring investigation or resolution.

Adjacent to a rise in complaints, we have seen a rise in litigants in person who are using generative AI systems to file and pursue formal court claims against organisations, leading to an increase in the number and length of filings.

The issue is becoming so pronounced that many regulators, such as the Information Commissioner’s Office, have now issued guidance on how the organisations they regulate should deal with AI-generated complaints.

Importantly, as such guidance makes clear, the fact that a complaint or claim may have involved the use of AI does not mean it can be disregarded or treated differently. At the end of the day, an AI-generated complaint or claim is still a complaint or claim and should be handled in the same way as any other.

As such, the key underlying principles of effective complaint and claims handling remain largely unchanged.

Why are AI-generated complaints and claims becoming more common?

Freely available generative AI tools have significantly reduced the effort required to draft a formal written complaint or to complete the court documents required to pursue a formal court claim against an organisation. Complaints that might previously have required substantial time, research and effort to prepare from scratch – or even the assistance of professional advisers – can now be produced quickly, at minimal cost and often without proper review before the complainant presses “send” (albeit, in the case of court claims, the requirement for the claimant, unless exempt, to pay a court fee (ranging from £35 for claims worth up to £300, to £10,000 for claims over £200,000) before they can issue their claim form can be an effective filter).

As a result, many organisations are reporting:

  • An increase in the number of formal written complaints received

  • Longer and more detailed complaint submissions

  • Complaints containing extensive legal or technical language

  • Greater difficulty identifying the specific issues requiring investigation

  • Increased time and cost associated with complaint handling

  • Complaints being received for which they are not responsible because AI has incorrectly identified them as the responsible party

  • Formal court claims being filed against them that are unduly lengthy or legally flawed, usually as a result of the inclusion of hallucinated case law and/or reliance on incorrect legal principles

Importantly, the use of AI does not necessarily mean that a complaint or claim lacks merit. Equally, the mere appearance of legal sophistication does not necessarily mean that the underlying concerns are well founded, or that the law cited is relevant or even correct.

Why does this matter?

The challenge is not whether or not AI-generated complaints or claims are more likely to be justified. Rather, it is that they often require more time and effort to process, diverting resources away from dealing with the substance of the complaint or claim in question. As a result, many organisations are having to reassess their approach to complaints and claims handling to ensure they strike the right balance between efficient resourcing and complying with their legal and regulatory obligations to deal with complaints and claims appropriately.

How to spot a potentially AI-generated complaint or claim

It is often impossible to know with certainty whether AI has been used and, ultimately, the answer to that question may be irrelevant. That said, common indicators include:

  • Unusually formal language – Complaints may read more like legal correspondence than a typical complaint. The tone can appear overly formal, adversarial or technical in comparison to normal everyday correspondence.

  • Excessive length – AI-generated complaints or claims are often significantly longer than necessary and may include extensive background material, lengthy chronologies, repeated allegations or language that would not normally be expected in a complaint or claim written from scratch by a lay person.

  • Speed of response – Lengthy replies may be received in a matter of hours, quicker than they could have been produced if writing them from scratch.

  • Repetition – In an AI-generated complaint or claim, the same concern may be raised multiple times using slightly different language, creating the impression of numerous separate issues when there may only be one or two issues requiring investigation.

  • Extensive legal references – AI-generated complaints or claims may contain numerous references to legislation, regulations, legal principles or case law that it would be unusual to see in a complaint or claim drafted by someone unfamiliar with the law. Some references may be inaccurate, irrelevant, relate to the law in foreign jurisdictions or just be entirely hallucinated.

  • Lack of specificity – Paradoxically, despite their length, AI-generated complaints or claims frequently lack the detail needed to investigate properly. Broad assertions may be made without identifying specific events, dates, problems, issues or supporting evidence.

  • Formulaic wording – Many AI-generated complaints or claims contain generic phrases about seeking a “constructive resolution”, acting in “good faith” or wanting a “safe environment” without clearly explaining what outcome is actually being sought, which may be as simple as a refund or replacement.

Should you ask whether AI was used?

In most cases, this is unlikely to be a productive question and may lead to a defensive response.

As noted, even if AI has been used, knowing that fact rarely changes how a complaint or claim should be handled. Focusing on the quality of the complaint or claim, the issues raised and what further information is required to investigate and resolve it is usually more constructive than spending too much time investigating how it was originally drafted.

Practical tips for dealing with complaints and claims (whether AI-generated or not)

  • Confirm if it is a “normal” complaint or a formal court claim – It is important for those dealing with complaints to understand the difference between receiving a “normal” complaint and being served with notification of a formal court claim. In most cases, this will be easy to establish, as a formal court claim must be commenced by issuing a court claim form in the prescribed form, headed “Claim Form” and bearing the court logo. It should also be sealed by the court following payment by the claimant of the applicable court fee. This is an important first filter, as receiving a validly issued and sealed formal court claim will then trigger a different response process to that for dealing with a “normal” complaint.

  • If it is a court claim, confirm the claim form has been validly served – A claimant cannot simply complete and send an organisation a claim form in order to commence a formal court claim against them. The claim form must first have been sealed by the court (and the applicable court fee paid by the claimant) and then served on the defendant within four months (assuming they are located in England and Wales). There are prescribed rules as to how and where a formal claim should be served, though these are often overlooked by litigants in person. As such, even if a claim form is received, before responding it is important to quickly verify that it has been properly issued and served. For example, check that a claim number and issue date have been inserted into the “for court use only” section of the form, and that an official court seal has been applied below that.

  • Maintain privilege – If a matter results in formal court litigation, internal communications and records relating to the complaint or claim are at risk of being disclosed to the complainant unless they benefit from legal privilege (which protects communications between a lawyer and their client provided they are for the dominant purpose of seeking and receiving legal advice) or litigation privilege (which potentially also protects communications with nonlawyers, but only if the dominant purpose of obtaining the information or advice is in connection with the conduct of existing or reasonably contemplated litigation). Where a valid court claim is received or a complaint is at risk of escalating to formal litigation, involving external legal advisers or in-house counsel early in the process can help ensure privilege is maintained. An added advantage is that external legal advisers using enterprise-grade AI tools can harness the benefits of AI in support of complaints handling and litigation while maintaining legal privilege – something that is lost if an organisation itself inputs privileged information into a public AI system.

  • Do not miss court deadlines – If a valid formal court claim is received, an organisation then has 14 days to file an acknowledgement, admission or defence to that claim with the court. However, if an acknowledgement of service is filed within this 14 day period, the deadline to file a full defence is extended to 28 days. As such, it is very important to ensure that these deadlines are not missed, to avoid unintended consequences such as judgment being filed against you in default (which litigants may again look to AI for assistance with).

  • Verify the complainant, and that you are responsible – Before expending significant time and effort investigating a complaint or claim, carry out initial basic verification to confirm that it is from a genuine complainant, and that you are legally responsible for dealing with it.

    For example, if the complaint or claim relates to goods or services, do your records show that the complainant is, in fact, a customer? If the complaint relates to an event you are alleged to have held, did you, in fact, hold the event in question?

    Similarly, consider whether any complaint or claim is clearly out of time before progressing to a more detailed investigation. But even if a formal claim is clearly out of time, you should still respond raising that defence, rather than simply ignore it.

  • Focus on the substance, not the style – Avoid becoming distracted by the length, tone or apparent sophistication of a complaint or claim, or by questions around whether AI has been used.

    The key questions that organisations need to ask are usually simple:

    • What is the complainant actually saying has happened?

    • What do they want you to do about it?

  • One of the most effective techniques is to distinguish factual allegations from commentary, conclusions and opinion. Instead, for each issue, ask:

    • What is being complained about?

    • When did it allegedly occur?

    • What supporting information has been provided, and what more do you need to investigate and assess?

    • What outcome is being sought?

    • When did it allegedly occur?

    This process will often reduce a complex complaint or claim to a manageable list of issues that can either be verified quickly or investigated further where necessary.

    • Seek clarification where necessary – A lengthy complaint or claim does not always equal a clear complaint.

      If relevant issues are unclear, seek clarification at an early stage. Where appropriate, it may be helpful to ask the complainant to:

      • Identify their principal concerns

      • Provide any missing factual details

      • Explain which issues they consider most important

      • Clarify their desired outcome

      • Explain that without this further information you will be unable to progress with investigating their complaint further

    • It is good to talk – Many organisations now rely heavily on online channels for customer interaction. While this may be efficient, it can also encourage lengthy written exchanges where tone and context can be misinterpreted. In some circumstances, a short telephone conversation may achieve more than several rounds of correspondence, particularly if those are likely to involve further use of AI by the complainant.

      If a complaint or claim appears to have merit, consider whether direct engagement could help clarify the issues and identify an appropriate resolution more quickly. A conversation can allow you to assess how reasonable a complainant is likely to be and what their real concerns are, and often achieve in a few minutes what an extended period of correspondence cannot.

      However, it is important to make clear that any offers you make to settle the complaint or claim are expressed on a “without prejudice” basis, as doing so should help ensure that should the claim reach the courts, any offers made by you to settle cannot then be used as evidence of an admission of liability on your part.

    • Do not feel the need to respond to every claim or correct every error – The apparent complexity of some complaints or claims can lead organisations to over-investigate relatively straightforward issues.

      In particular, it can sometimes be difficult to resist the temptation to prepare lengthy point-by-point responses or correct every factual or legal error in the complaint or claim. However, particularly in the case of complaints, doing so rarely leads to a quicker or more effective resolution, and can often antagonise the complainant.

      Maintain a proportionate approach. The appropriate level of investigation should be determined by the seriousness and nature of the underlying concerns, not by the volume of material submitted.

      In most cases, it is more effective to:

      • Identify the key issues

      • Verify and investigate only those key issues

      • Make findings based on those key issues

      • Explain the reasons for the conclusions reached, making reference to applicable laws and regulations if relevant and helpful to do so

        A structured response focused on substantive matters is usually clearer and more defensible than attempting to rebut every individual point.

    • Encourage good complaints – Complaints are a fact of life when doing business.

      While it may seem counterintuitive to encourage complaints, providing guidance on how to make an effective complaint can help organisations deal with them more efficiently. In many sectors, organisations may, in any event, be required to publish information about their complaints procedures.

      Clearly explaining the information required to investigate a complaint, or encouraging the use of a complaint form, can help reduce the time and effort associated with assessing poor-quality complaints. If an AI tool is used by a complainant, there is a reasonable chance that this guidance will also influence the content that the AI then produces.

      Some organisations have gone a step further and issued guidance on the appropriate use of AI when making complaints, although it remains unclear at this early stage whether this is resulting in better-quality submissions.

    • Ensure personnel are familiar with applicable laws and regulations – Ensure that personnel who deal with and oversee complaints or claims are kept up to date with applicable laws and regulations, as that will, in turn, ensure they are equipped to quickly spot where a complaint or claim cites irrelevant or incorrect law.

      Similarly, it is important to ensure that complaints, and claims in particular, are processed in accordance with any specific applicable legal or regulatory requirements and, importantly, that any relevant deadlines for acknowledging receipt of complaints, requesting further information, etc., are not inadvertently missed.

    • Use AI responsibly when handling complaints or claims – Despite concerns about the rise in poor-quality AI-generated complaints and claims, many organisations are themselves deploying AI to assist with complaint and claims handling.

      Any organisation doing so should proceed with caution and remember that it remains accountable for its own use of AI systems. As key regulators such as the Competition & Markets Authority have already made clear, AI cannot be used as an excuse for failing to handle complaints or claims in accordance with applicable laws or regulatory requirements.

      AI tools can, however, add value when assisting with administrative or drafting tasks as detailed above, such as:

      • Summarising lengthy complaints or claims

      • Identifying themes and issues

      • Organising documents

      • Improving the clarity of draft correspondence

      However, organisations should ensure that:

      • Appropriate human oversight is maintained throughout the process.

      • Decisions remain the responsibility of appropriately trained and authorised personnel who are familiar with the applicable law, regulatory requirements and an organisation’s own policies and procedures.

      • Findings are based on evidence and human judgement.

      • Appropriate data protection safeguards are maintained.

      • Confidential information or information subject to legal or litigation privilege is not entered into public AI systems.

      • Particular care is taken when using AI tools in connection with documentation that will be submitted to a court. As the courts have now made clear on numerous occasions, errors in documents placed before them such as case law hallucinations will not be excused on the grounds of AI use and can lead to serious consequences, including the possibility of being held in contempt of court.

      An important distinction should always be maintained between using AI as a support tool and allowing AI to make unchecked decisions.

Key takeaways

AI-generated complaints and claims have quickly established themselves as a common feature of complaint handling, and that is unlikely to change in the near future. Organisations must adapt to reflect this trend, focusing on the substance rather than style of complaints and claims, and seek to avoid getting dragged into lengthy discussions over AI usage or errors that are unlikely to result in resolution.

The most effective response to an AI-generated complaint or claim is often the same as the most effective response to any other complaint or claim:

  • Verify that the complaint or claim is valid, and that you are responsible for it

  • Identify the key issues

  • Request any further information needed to investigate the key issues

  • Investigate and assess fairly in accordance with applicable laws and regulations

  • Communicate clearly using a means of communication most likely to result in a resolution

  • Ensure decisions are subject to appropriate human oversight and can be justified if challenged

  • Avoid creating further complaints or liability through your own inappropriate use of AI

On the horizon

As organisations begin to get to grips with dealing with AI-generated complaints, it is becoming apparent that another challenge is on the horizon and likely to be quite different again: the rise in complainants – and, in particular, individual consumers – utilising AI agents to make and resolve complaints on their behalf. In our next post, we consider the legal status of AI agents when used to make complaints on behalf of an individual, as well as providing practical tips for dealing specifically with complaints made using an AI agent.

How can we help?

Our International Dispute Resolution and Intellectual Property & Technology teams have extensive practical experience of both dealing with and defending complaints and claims, and of working with clients to design and implement complaint- and claims-handling processes that take a proportionate risk-based approach. For further information, please contact Carlton Daniel, Deborah Polden or Miles Robinson.