Why mandatory AI training is becoming a professional-services requirement

Sidley Austin has made a four-month AI rotation compulsory for every trainee in its London office. The move is more significant than another law-firm technology announcement. It signals that practical AI competence is becoming part of professional formation — something firms must teach, supervise and assess before people qualify.

Under Sidley’s revised six-seat training contract, trainees complete their first two rotations in traditional practice groups before becoming eligible for the AI Knowledge Lab. The firm describes the lab as a working seat rather than a classroom: trainees carry out genuine legal assignments, work with knowledge, research and data teams, test new approaches and learn to identify when AI adds value and when a lawyer must intervene.

That model offers a useful lesson beyond law. Accountancy, consulting, insurance and other UK professional-services firms face the same underlying problem: buying AI tools is easier than building professionals who can use them with sound judgement.

The shift from optional tool to professional competence

Most firms began their generative AI programmes with short awareness sessions: how to write a prompt, what not to upload and which approved tool to use. That was a sensible starting point, but it is not enough for work that affects clients, financial decisions or regulatory obligations.

The difficult skills emerge only in context. A professional must recognise when an apparently confident output is incomplete, when source material is too sensitive for a particular system, when an answer requires independent verification and when efficiency is being gained at the expense of quality.

A mandatory working rotation addresses that gap. It makes AI use observable. Supervisors can see how trainees frame tasks, test outputs, document assumptions and escalate uncertainty. The emphasis moves from access to competence.

This distinction matters because firms remain accountable for the work delivered under their name. AI can contribute to research, drafting and analysis, but it cannot absorb professional responsibility.

A labour-market signal, not just a legal-sector experiment

The wider UK employment data points in the same direction. PwC’s 2026 AI Jobs Barometer found that specialist AI job postings rose by 61% year on year, from 112,000 to 180,000. Growth was led by “AI user” roles — people who apply AI within an existing field — rather than roles focused solely on developing AI systems.

PwC also found that AI-exposed entry-level roles are increasingly asking for capabilities normally associated with more senior work, including judgement, leadership and team-building. Its analysis describes a two-track market: roles in which AI augments expertise are growing faster and creating more value than roles in which technology primarily simplifies tasks.

For professional-services leaders, the implication is clear. The competitive workforce will not consist only of technical specialists. It will include lawyers, accountants, consultants and risk professionals who understand their discipline and can apply AI critically inside it.

But there is a problem. If firms automate too much foundational work without redesigning training, junior staff may be expected to exercise senior judgement before they have accumulated the experience needed to do so.

The professional apprenticeship is under pressure

A UK government and LinkedIn snapshot published in June 2026 found that entry-level hiring was weak across the labour market, with the UK hiring rate down 14% year on year in April. The report found particularly sharp falls in several knowledge-sector roles, including accountants at 29%, legal assistants at 14% and data analysts at 15%.

The report is careful not to claim that AI caused those declines. Hiring was falling across seniority levels, and broader labour-market weakness remains an important factor. Even so, it identified a mismatch between the general analytical skills candidates often offer and the operational, production-ready capabilities employers increasingly seek.

That mismatch should concern partnership boards. Junior work has traditionally served two purposes: it produces an output today and develops the professional who will exercise judgement tomorrow. If AI removes the task but the firm does not replace its learning value, the organisation creates a capability gap that may become visible only years later.

Sidley’s model points to a more constructive response. Do not choose between adoption and development. Redesign development around adoption.

Regulation makes supervised practice essential

The Solicitors Regulation Authority’s August 2026 warning notice reinforces why practical training must include controls, not just productivity techniques. The regulator highlighted AI-generated inaccuracies in legal research and court submissions, as well as confidentiality risks when client information is entered into tools without appropriate safeguards.

The SRA received 42 reports concerning potential AI misuse between July 2025 and July 2026. Its position is unambiguous: individuals remain responsible for their work and advice whether or not AI was involved, while firms must maintain appropriate oversight and governance.

Although the warning applies directly to solicitors, the operating principle travels across regulated professional services. A firm needs named accountability, approved systems, human review, evidence trails and clear escalation points. Training should demonstrate that staff can operate those controls under realistic conditions.

A practical model for UK firms

Professional-services firms do not need to copy a four-month legal training seat exactly. They do need to move beyond one-off awareness sessions. A practical programme can be built around five elements.

1. Use real, bounded workflows

Select work that is frequent enough to practise but controlled enough to supervise: document review, research synthesis, first-draft reporting, meeting analysis or structured due-diligence support. Define what the AI may do, what requires review and what remains prohibited.

2. Assess judgement, not prompt fluency

A polished prompt is not evidence of professional competence. Assess whether participants verify sources, detect omissions, recognise uncertainty, protect confidential information and know when to stop using the tool.

3. Make review visible

Require a short record of the input, output, checks performed, corrections made and final human decision. This provides a learning mechanism for trainees and an audit trail for the firm.

4. Protect the apprenticeship pathway

For every task automated, identify the judgement or contextual knowledge that task used to develop. Replace lost learning deliberately through supervised simulations, client exposure, review exercises and rotations across functions.

5. Measure client and business outcomes

Track quality, turnaround time, rework, escalation rates and client impact. Time saved is useful, but it is not sufficient if error rates rise or senior reviewers inherit additional correction work.

The strategic lesson

The most important part of Sidley’s announcement is not the word “AI”. It is the decision to place AI inside the formal system through which professionals learn their craft.

That is where the market is heading. AI capability will increasingly be judged as applied professional competence: the ability to use systems effectively, challenge their outputs, protect clients and remain accountable for the result.

UK firms that build this capability now can improve productivity without hollowing out their future talent pipeline. Those that rely on licences and optional webinars may discover that they have deployed the technology but not developed the people required to operate it safely.

At Quantum & AI Technologies, we help professional-services firms turn AI adoption into governed, measurable operating capability — from workflow selection and training design to oversight and implementation. Book a 30-minute consultation.