Broken Ladder: Are Lawyers Sleepwalking into a Competence Crisis?

By James Tuke, CEO, AI Futures Forum.

It is late January 2026. For the past two years, the legal sector has been in a race to deploy GenAI. We have moved from the initial hype phase to the deployment of tools like Harvey, Spellbook, and Microsoft Copilot. The efficiency gains are no longer theoretical; they are in the bottom line.

But while the balance sheets look healthier, the foundations of the profession are showing cracks.

In our latest Green Paper, ‘The Broken Ladder: Human Capital Risk in UK Professional Services’, we argue that the sector is witnessing the beginning of a ‘junior cull’ – a structural hollowing out of the pyramid model that has sustained law firms for decades. We are automating the ‘training wheels’ tasks – document review, bundling, basic drafting – that historically nurtured professional judgment.

The question facing law firm leaders in 2026 is no longer only ‘How do we adopt AI?’. It is also: ‘Where will the next generation of Partners come from if they never learn the law by doing the grunt work?’

The Data: The Junior Cull is Real

The professional services sector is often viewed as a monolith and the trends we are seeing in the Big Four are rapidly bleeding into the legal market. Our research highlights a stark contraction in graduate intake across the Golden Triangle of accounting and consulting – with firms like KPMG cutting entry-level roles by up to 29% in the last cycle.

Why does this matter for law firms? Because the economic drivers are identical.

Data from Thomson Reuters suggests that AI tools are now saving approximately 190 billable hours per lawyer annually. In a billable-hour model, that is a revenue hole of billions across the US and UK markets. Clients, armed with their own increasingly sophisticated legal operations teams and chargeback mechanisms, are reluctant to pay for Junior time spent on tasks they know can be automated.

The result is a decoupling of revenue from headcount. Firms are moving from the traditional Pyramid structure (leverage based on armies of associates) to a Diamond model – lean at the bottom, heavy with mid-level specialists and ‘AI Architects’, and exclusive at the top.

The Competence Trap: The Verification Gap

The most alarming finding in our Green Paper isn’t economic; it’s epistemic. We are creating a ‘verification gap’.

Regulators and insurers are increasingly demanding human-in-the-loop safeguards for AI output. But there is a paradox at the heart of this demand: to supervise an AI effectively, the human must be more competent than the machine.

However, a junior lawyer who has bypassed the ‘desirable difficulty’ of manual research and drafting – relying instead on an LLM to produce the first draft – may never develop the deep mental connections or legal intuition required to spot a hallucination or a subtle logic error.

We are seeing the emergence of what King’s College London’s recent study terms ‘cognitive deskilling’. We are raising a generation of ‘black box’ lawyers who can prompt an answer but may not be able to explain its derivation. This is not just a training issue; it is a liability time bomb. The global cost of AI hallucinations and reliability issues is estimated to have hit US$67.4 billion in 2025. When a junior associate misses a hallucinated clause because they lack the muscle memory of contract law, who pays the claim?

2026: The Age of the Digital Worker

Looking ahead to 2026, the Green Paper forecasts a shift from chatbots to Agentic AI.

We are about to see the widespread deployment of digital workers – autonomous agents capable of negotiating a clause, updating a case file, and billing the client without human intervention. This threatens to erode not just the trainee layer, but the mid-level associate role – the project managers of the legal world.

If an agent can orchestrate the workflow, the structural need for middle management diminishes. The ‘up or out’ model is morphing into ‘retrain or perish’.

The Solution: Simulated Training Environments

The industry cannot reverse the tide of automation. Clients will not subsidise inefficiency for the sake of in-house training. Therefore, firms must actively reconstruct the training models they have destroyed.

Our report argues that 2026 must be the year of human preservation, so we recommend firms invest in:

1. Simulated Training Environments: If the grunt work no longer exists in live client matters, we must create ‘flight simulators’ for lawyers – simulated transactions where juniors can fail safely and learn the mechanics of the law.
2. The Apprenticeship of Oversight: We need to overhaul the Trainee Solicitor/Associate curriculum to focus on auditing AI. We must teach juniors how to break the machine, creating a critical relationship with technology rather than a passive one.

Conclusion

The efficiency gains of 2025 may well have been purchased at the cost of future capability: we are profiting today by mortgaging the expertise of tomorrow’s partners.

The ladder is broken. It is time to build a bridge.

Download the Full Green Paper here

—

About the author: James Tuke is the founder of multiple tech-related ventures, such as Treat Digital, and is now also the CEO of the AI Futures Forum, a site dedicated to in-depth discussion about AI, and writer of ‘A Short Walk in AI’. He also advises firms and senior management about their AI strategy.

–

[ This is an educational guest post for Artificial Lawyer. ]