Research conducted by HR data and intelligence provider Brightmine in July 2026 found that 87% of UK organisations have seen an increase in grievances drafted with AI assistance over the past year. The documents now arriving in HR inboxes are longer, more legally framed and slower to conclude than their predecessors, and 94% of those organisations have no guidelines in place on how employees should use AI when raising a complaint. For People directors and HR teams already navigating a period of regulatory change, that policy vacuum carries a measurable cost in management time, legal risk and the prospect of an Employment Tribunal compensation uplift.

What the Brightmine data shows

Brightmine surveyed 201 UK HR professionals representing over 530,000 employees in July 2026. Among organisations that have received AI-assisted grievances, 92% report overly complex or exaggerated submissions, 83% report longer complaints and 82% report that employees arrive with unrealistic expectations about outcomes. Those are not incidental drafting artefacts. Each one adds scoping and investigation work before a formal process has even properly begun.

Across the broader caseload, 88% of organisations have seen increased use of formal or legal language in grievances, 82% have seen more detail in individual complaints and 81% say more issues are now bundled into a single submission. Nearly three in four (73%) say the substance of complaints has become more complex, and 71% say grievances now take longer to resolve. Volume is also tracking upward: 71% of organisations received more grievances over the last two years. That increase in words has not produced more clarity. A third of HR professionals (33%) say the intelligibility of grievances has actually declined over the same period.

The Chartered Institute of Personnel and Development (CIPD), whose annual survey of UK HR practice provides a useful benchmark across organisations of comparable scale, has consistently flagged informal resolution capability as an underdeveloped discipline. The Brightmine figures add quantitative weight to that assessment.

Where the operational pressure concentrates

The most common grievance categories are relational rather than technical in nature: relationship with a manager (67%), bullying or harassment (52%) and relationship with a colleague (49%). These are precisely the situations where a well-handled early conversation tends to be more productive than an extended written exchange. Under the Equality Act 2010 s.26, harassment is defined by conduct that has the purpose or effect of violating an individual's dignity or creating an intimidating environment. An AI-drafted submission can readily frame a tense working relationship in those statutory terms, regardless of whether the underlying facts meet that threshold. The result is a grievance that reads like a protected-characteristic claim from its opening paragraph, and one that an HR manager must address with proportionate rigour.

Sheila Attwood, Senior Content Manager at Brightmine, characterises the employee impulse as entirely understandable: AI enables people to consolidate their experiences, reference statutory frameworks they believe have been breached and present their complaint in a register that feels more authoritative. The difficulty sits firmly on the employer side of the relationship.

"The challenge for employers is that AI can also introduce unnecessary legal complexity and create unrealistic expectations about outcomes and timeframes."

In practice, that mismatch tends to emerge at the outcome or appeal stage, once positions have hardened and the underlying working relationship is considerably harder to repair. Managing that expectation gap earlier in the process is more efficient and less costly than correcting it at the end.

The policy gap and why a blanket prohibition is a weak response

With 94% of organisations offering no guidance, most employees are effectively setting their own norms around AI use in complaints. HR teams show little confidence that this trend is improving outcomes: only 11% believe recent technological changes have had a positive overall impact on the grievance process.

The regulatory backdrop is shifting at the same time. ACAS (the Advisory, Conciliation and Arbitration Service) launched a consultation in July 2026 on a revised Code of Practice on disciplinary and grievance procedures, its first substantial update since 2009. The consultation, which closed on 23 September 2026, asked directly whether the Code or its accompanying guidance should address AI use in these processes, though the draft text remained silent on the point. Organisations monitoring the revision should note that the final Code, once issued, will form the benchmark against which Employment Tribunal panels assess procedural reasonableness.

The financial exposure is already defined under existing rules. Where an employer unreasonably fails to follow the ACAS Code, a tribunal may increase any relevant compensation award by up to 25%, under the Employment Act 2008 s.3(2). Receiving an AI-drafted grievance does not reduce the employer's procedural obligations: the complaint still requires a reasonable investigation, conducted with an open mind and a genuine opportunity for the employee to explain the underlying concern. Employment Rights Act 1996 ss. 94-98 governs the broader framework within which unfair dismissal claims connected to grievance mishandling are assessed, and case law consistently rewards proportionate process over technical defensibility.

An outright ban on AI-assisted submissions sits poorly within that framework. A blanket prohibition is unlikely to withstand scrutiny as a reasonable workplace rule, whereas a grievance policy that sets clear expectations about how AI is used in drafting is both proportionate and defensible.

The line-manager bottleneck

Two in three HR professionals (66%) say employees tend to bypass informal routes and raise formal grievances directly, and only 36% believe line managers have the capability to handle concerns informally before they escalate. These two figures, read together, describe a process with a missing first step. Where managers are not equipped for early resolution conversations, concerns travel straight into the formal channel, and AI-assisted drafting makes that channel heavier at every stage.

The combination is particularly acute because the manager relationship is itself the most common grievance type. The individual most likely to be the subject of a complaint is often the same person expected to create the conditions for informal resolution. ACAS's early conciliation service data, published annually, shows that a significant proportion of employment disputes referred to the service could have been addressed at workplace level had a capable informal conversation happened earlier. Building that capability within the line-management population is a more durable intervention than policy alone.

Verified employee sentiment data offers an operational lever here. Where review data surfaces recurring commentary about management style in a specific business unit, or where engagement signals point to concentrated friction in particular teams, HR has the basis to target manager development before a concern becomes a formal case. The CIPD's guidance on people analytics makes clear that this kind of leading-indicator monitoring is increasingly within reach for mid-sized organisations, not only large enterprises.

Practical steps HR teams can take ahead of the revised ACAS Code

None of the following requires waiting for the final text of the revised ACAS Code of Practice. Organisations that manage AI-assisted grievances effectively tend to focus on five operational areas:

• Add an AI-use clause to the grievance policy. Set out clearly that employees may use AI tools to organise their thoughts, but that the final submission should reflect their own account of relevant facts and must not involve uploading colleagues' personal data or confidential business information to external platforms. Data protection obligations under the UK GDPR and the Data Protection Act 2018 apply to employees as well as employers in this context.
• Hold a scoping meeting before investigation begins. Agree with the employee the core concern, the specific allegations and the outcome they are seeking. This converts a sprawling multi-issue document into a manageable scope without dismissing anything substantive that has been raised.
• Set written expectations on process and outcome. Explain in writing what the grievance procedure can and cannot deliver, and the realistic timescale for each stage. Where legal language has been imported into the submission, addressing that framing early prevents it becoming a false benchmark at the outcome stage.
• Invest in manager capability for informal resolution. Prioritise teams with high complaint volumes. The focus should be on early conversations, recording agreed outcomes and recognising the point at which an informal route is no longer appropriate and formal escalation is required.
• Apply the same standards to HR's own AI use. Where HR teams use AI to summarise, draft or respond to grievance documentation, approved tools, data protection checks and human judgement on every substantive decision should be the baseline, not an aspiration.

Measuring the effect matters as much as implementing these steps. Grievance volume, average time to resolution and the proportion of concerns resolved at the informal stage are straightforward indicators that show whether policy and manager development investment is working. They also give senior leadership a clear line of sight to what the current gap is costing the organisation.

Keeping the workplace concern at the centre

Attwood's reading of the Brightmine research is direct: employers should not seek to prohibit AI from the process, but they do need clearer guidance on its appropriate use, while keeping the underlying workplace concern at the centre of every grievance conversation. The longer that guidance is absent, the more complaints will arrive already cast as legal disputes rather than as workplace problems capable of resolution.

For HR Directors and People leads, the strategic frame is less about AI specifically and more about process architecture: a clear rule for employees, a capable first line of management and a procedure that consistently routes every complaint back to the substantive issue behind it. Organisations that establish those three elements before the revised ACAS Code of Practice is finalised will be in a stronger position to absorb whatever guidance it ultimately contains, whether or not it addresses AI by name. The Employment Rights Act 1996 framework and ACAS procedural benchmarks have not changed. What has changed is the character of the documents arriving under them.