What If £4.5m Could Actually Transform Legal AI? The Training Question Nobody Is Asking

Imagine £4.5m landed in your firm's bank account tomorrow, ring-fenced for legal technology. What would you actually do with it?

That thought experiment is worth sitting with for a moment, because the Lord Chancellor has just announced exactly that sum for LawtechUK, the Ministry of Justice-backed initiative that has been promoting the legal technology sector since 2019. £1.5m per year for three years. The announcement lands alongside broader MoJ ambitions around AI in courts, including a £12m Justice AI Unit, intelligent listing pilots, and video hearing infrastructure.

It is not nothing. But it is worth being precise about what £4.5m actually buys you in the AI world, and what the profession genuinely needs from this moment.

What £4.5m Does and Doesn't Buy

OpenAI's annualised compute costs run to billions of dollars. DeepMind's research budget dwarfs the entire LawtechUK programme many times over. Even a serious mid-market legal AI startup will burn through £4.5m in runway before it has built anything courts-ready.

So this is not a research and development budget. Nobody should pretend otherwise.

LawtechUK has always been a promotional and convening body, not a technology builder. Its original £7.5m mandate produced regulatory sandboxes, industry reports, and introductions between startups and established firms. That has genuine value. Knowing which tools exist, which startups are credible, and where the regulatory boundaries sit matters. The refreshed programme will likely do more of the same: convening, promoting, signalling government seriousness about the sector.

The signal matters. A Lord Chancellor prepared to talk about AI in the same breath as court reform is a different posture from the institutional caution that characterised much of the last decade. And for UK lawtech startups eyeing procurement opportunities or partnerships with the likes of Microsoft, which has already been in conversations with the MoJ, that signal unlocks conversations that might otherwise stall.

But signals are not skills. And skills are where the profession has a serious gap.

The Training Problem That Funding Alone Won't Solve

Here is a question worth asking of every law firm, chambers, and in-house legal team in England and Wales: what structured AI training have your fee earners received in the last twelve months?

For most, the honest answer is very little, or none at all.

The tools already exist. Firms are already paying for them. Co-pilot integrations, contract review platforms, legal research assistants, document drafting aids. Licences have been purchased. The software is installed. And in many cases it sits largely unused, or used poorly, because nobody has been trained properly in how to apply it or where its limits lie.

This is not a technology shortage. It is a capability shortage.

The Solicitors Regulation Authority's position on AI has been consistent: competence obligations under the SRA Standards and Regulations apply regardless of whether the work is done by a human or assisted by a machine. Practitioners remain responsible for the output. That means understanding the tool well enough to catch its errors, which requires training, not just access.

The Bar Standards Board has said similar things. The Ayinde judgment, in which a barrister was sanctioned in part for conduct related to AI-generated content that had not been properly checked, is a marker of where regulatory attention is heading. The professional obligations are not waiting for government funding cycles to catch up.

What a Well-Spent £4.5m Would Actually Look Like

Indulge the thought experiment properly. If the purpose of this investment is to make the legal profession more capable with AI, where does the money actually move the needle?

Not in building new tools. The tools exist. Not in further reports on what AI can theoretically do for access to justice. Those reports exist too.

The answer is structured, profession-wide training. Not webinars. Not vendor demos dressed up as CPD. Proper curriculum development: how large language models work, where they hallucinate and why, how to construct prompts for legal tasks, how to review AI output critically, how to document AI-assisted work for audit purposes.

The Judicial College has already begun developing guidance for judges on AI in litigation. If there is appetite to build something equivalent for practitioners, whether through LawtechUK, the Law Society, or a body yet to be constituted, that is where sustained investment would compound.

The MoJ's Justice AI Unit, with its £12m budget, is reportedly focused on court-facing applications: listing optimisation, case triage, administrative efficiency. That is a different population from the practising profession. Both matter, but they need different interventions. Efficiency gains in the listing system do not automatically translate into a more AI-capable solicitors' profession.

The Monday Morning Test

What can a practitioner actually do with this news by Monday?

First, do not wait for government-funded training. The tools available now, used well, will differentiate your practice from competitors who are waiting for someone else to teach them. If your firm has an AI tool subscription, read the documentation. Run test tasks. Find the edges of what it can reliably do.

Second, if you supervise others, consider whether you have a clear policy on AI-assisted work. Not a prohibition, and not a blank cheque. A considered position on what AI-assisted output requires before it leaves the building. Review. Citation checking. Source verification. That is a supervision question, not a technology question.

Third, watch what LawtechUK actually does with this money. The programme has historically produced useful, if underread, market reports and sandbox outputs. If the refreshed iteration produces practical training resources or procurement frameworks for smaller firms, that is genuinely useful. If it produces more promotional material aimed at attracting investment into the sector, the value flows to startups rather than practitioners.

Those are different outcomes. It is worth knowing which one you are getting.

The Real Investment Case

Government investment in lawtech at this level is best understood as infrastructure spending on visibility: making the UK an attractive jurisdiction for legal AI development, keeping domestic startups in conversation with the courts and regulators, and maintaining institutional knowledge about what is being built.

That has worth. The alternative, leaving the field entirely to US platforms without a domestic counterweight, has real costs for how UK law develops in an AI-mediated world.

But the profession's own investment calculus is different. Firms, chambers, and in-house teams are not choosing between government grants. They are choosing how to spend their own training budgets and CPD hours.

The tools are ready. The regulatory obligations are live. The gap between what AI can do for a well-trained practitioner and what it currently does for most practitioners is very wide.

£4.5m of public money will not close that gap. A serious commitment to training might.