Employment tribunals in the UK are being pushed into a familiar but increasingly dangerous corner: more people are bringing claims, the system is not scaling fast enough, and the knock-on effects are starting to look less like a temporary backlog and more like a structural stress test. A new report points to a surge in workplace-related disputes that is clogging timetables, stretching caseworkers, and forcing parties to wait longer than anyone involved—workers, employers, or tribunal staff—would consider acceptable.
What makes this moment different is not only the volume of claims, but the way those claims are being prepared and submitted. Artificial intelligence tools are now part of the workflow for many claimants, helping them draft particulars, organise evidence, and navigate procedural steps that used to be daunting even for people with legal support. That shift is changing the speed at which cases enter the system. It may also be changing the shape of what arrives: more filings, more structured narratives, and potentially more claims that are “complete” on paper even when the underlying dispute is complex.
The result is a tribunal landscape that is struggling to keep up—not necessarily because employment law has become more contentious overnight, but because the pipeline from grievance to formal claim has been accelerated by technology, and because the capacity of the tribunal system has not been redesigned for that new reality.
A system designed for a certain pace meets a new throughput problem
Employment tribunals were never built to operate like an automated intake line. They are quasi-judicial bodies that must balance fairness, due process, and the practicalities of hearing evidence. That means there are limits to how quickly cases can be processed without compromising quality. When demand rises, the system absorbs it through delays: later hearings, longer waits for preliminary decisions, and slower resolution of disputes that could otherwise be settled earlier.
The report’s central concern is operational. Backlogs are not just inconvenient; they affect leverage and behaviour. When timelines stretch, employers may become more willing to settle simply to reduce uncertainty, while workers may feel pressure to accept outcomes without full clarity about their prospects. Delays can also increase the cost of litigation for both sides—legal fees, time off work, and the administrative burden of preparing for hearings that keep moving.
In other words, the tribunal system is not only adjudicating disputes; it is shaping the incentives around those disputes. When the system slows down, the incentives change. And when AI helps claimants move faster through the early stages, the system can end up receiving more cases than it can realistically schedule.
AI as a “claim accelerator,” not a magic wand
It is tempting to frame AI’s role as either a democratizing force or a destabilising one. The reality is more nuanced. AI tools do not decide cases. They do not interpret employment law in a way that replaces a judge. But they can reduce friction in the steps that come before a tribunal ever sees a dispute.
For many claimants, the hardest part is not understanding that something went wrong—it is translating a lived experience into the language of legal pleadings. Employment claims often require specific allegations, dates, comparators, and evidence links. Claimants must also navigate procedural requirements: deadlines, forms, and the structure of submissions. Even when people have a clear sense of injustice, turning that into a coherent claim can be overwhelming.
AI tools can help with exactly that translation. They can summarise events, suggest timelines, help draft statements, and organise documents. They can also help users identify what information is missing and prompt them to fill gaps. In effect, AI can act like a scaffolding layer between a claimant’s raw narrative and the formal structure the tribunal expects.
That scaffolding can be beneficial. It can improve clarity, reduce errors, and help claimants avoid procedural mistakes that would otherwise lead to rejection or delay. It can also make it easier for people without legal representation to participate meaningfully in a system that historically has been difficult for non-lawyers to navigate.
But there is a second-order effect: if more people can prepare claims quickly and correctly, the number of claims entering the system may rise. Not because more people are suddenly wronged, but because the barrier to filing is lower. When barriers fall, filing rates can increase—especially when people believe they have a credible path to a hearing.
The report does not claim that AI is causing fraudulent claims or that it guarantees success. Instead, it highlights a capacity issue: the system is being asked to absorb a higher volume of cases at a time when tribunal resources and scheduling constraints remain largely unchanged.
Backlogs as a policy signal, not a partisan verdict
One reason the report’s framing matters is that it avoids turning the story into a simple blame game. It does not present the backlog as proof that claimants are abusing the system or that employers are uniquely targeted. It treats the backlog as a signal about system design.
Employment tribunals are part of a wider justice ecosystem. If the intake pipeline becomes faster—through AI-assisted preparation, improved guidance, or changes in how people understand their rights—then the system needs corresponding adjustments. Those adjustments could include more tribunal capacity, better triage, improved settlement pathways, or reforms to how cases are managed.
Without such changes, backlogs become self-reinforcing. Parties learn that hearings take longer. Some claimants may file earlier or more readily, expecting that delays will be the norm. Employers may respond by adjusting settlement strategies, sometimes offering earlier resolutions to avoid prolonged uncertainty, sometimes contesting more aggressively because the cost of waiting is already baked into the timeline.
The backlog therefore becomes more than a queue. It becomes a market condition that shapes behaviour.
Why “more complete” claims can still overwhelm the system
A common misconception is that if AI helps claimants submit better-prepared cases, the system should handle them more efficiently. In some respects, that is true. Clearer pleadings can reduce confusion. Better evidence organisation can shorten time spent on basic clarifications.
However, tribunal capacity is not only about paperwork. It is about hearings, judicial time, and the practicalities of evidence. Even if AI improves the quality of submissions, each claim still requires a decision pathway: preliminary checks, case management, disclosure considerations, and ultimately a hearing or settlement.
If the number of claims rises faster than the number of hearings that can be scheduled, the system still backs up. Better preparation may reduce rework, but it cannot eliminate the fundamental throughput constraint.
There is also the possibility that AI-assisted drafting changes the distribution of case types. For example, if AI helps claimants articulate multiple allegations within a single filing, the complexity of each case may increase. A claim that might previously have been filed as a simpler complaint could become a more detailed set of allegations once the claimant has support in structuring the narrative. Complexity can slow down case management even when the submission is well written.
So the question becomes not only “Are claims being filed?” but “How many claims are being filed, and how complex are they when they arrive?”
The human cost of delay—and the strategic cost
Backlogs are often discussed in terms of administrative strain, but the impact on individuals is immediate. Workers who bring claims are frequently dealing with financial insecurity, ongoing job searching, and the emotional toll of conflict with former employers. Waiting months—or longer—for a hearing can prolong stress and uncertainty. It can also affect evidence quality: witnesses move on, documents get lost, and memories fade.
Employers face their own burdens. Even when they believe they have strong defences, delays extend the period of reputational risk and internal disruption. Managers and HR teams may spend time preparing for a dispute that keeps shifting. Legal costs can accumulate, and the longer the timeline, the harder it can be to maintain consistent documentation.
Delay also changes settlement dynamics. When hearings are far away, parties may settle based on risk calculations rather than on a clear sense of how the tribunal will view the merits. That can lead to settlements that are not necessarily optimal for either side—just the best available under uncertainty.
In that sense, the backlog is not neutral. It influences outcomes indirectly, even if judges remain impartial.
AI and the “paperwork gap”: narrowing access, widening intake
The report’s most interesting implication is that AI may be narrowing the access gap between represented and unrepresented claimants. Historically, legal representation has been uneven. Many workers cannot afford solicitors, and even those who can may face limitations in scope or funding. Without representation, claimants often struggle with the procedural and drafting aspects of bringing a claim.
AI tools can partially substitute for that support. They can help users produce coherent documents, understand what details matter, and structure their allegations in a way that aligns with tribunal expectations. That can be a genuine improvement in access to justice.
But access improvements can create system-level consequences. If more people can successfully navigate the early stages, the system receives more claims. That is not inherently bad—justice systems should be accessible. The problem arises when the system’s capacity does not expand alongside access.
This is where the “breaking the system” concern comes in. Not because AI is inherently harmful, but because the system may be operating at a fixed capacity while demand is becoming more elastic.
When demand becomes elastic, fixed capacity fails
Many public services are designed around predictable demand. Employment tribunals have long faced fluctuations, but the current situation suggests a more persistent mismatch between demand and capacity. If AI reduces the friction to filing, demand may become more responsive to perceived opportunity. People who might previously have decided not to pursue a claim may now file, especially if they believe AI can help them avoid mistakes.
Elastic demand is not a theoretical concept here. It shows up in the real world as more submissions, more case management work, and more scheduling pressure. If tribunal staffing and hearing availability do not scale accordingly, backlogs grow.
And once backlogs grow, the system’s ability to manage cases efficiently can degrade. Case management becomes more complex when everything is delayed. Administrative tasks pile up. Judges and tribunal staff must prioritise, triage, and allocate limited time across a growing queue.
The report’s warning is essentially that the system may