An AI law firm won. A human still spoke in court.

On 14 May 2026, after a three-hour trial at Wandsworth County Court, a judge found in favour of freelance HR consultant Tamires Camal Taquidir, awarded the full £7,000 she was owed by a hospitality business, and dismissed the defendant's counterclaim. The work leading up to that hearing - the pre-action correspondence, the court filings, the witness statements, the trial bundles - was produced by Garfield AI, the first law firm in England and Wales authorised by the Solicitors Regulation Authority to deliver regulated legal services primarily through artificial intelligence.

It is worth being precise about what happened, because the social-media version is not quite the real one. A robot did not argue the case. Shortly before trial, Garfield instructed a human barrister to make the oral submissions and handle cross-examination. The AI took the matter, in the founders' own framing, up to the front door of the court. A person walked it through the door. The claimant paid around £400 in fees; the losing side arrived with both a solicitor and a barrister.

That distinction is the whole point. The expensive, repetitive, document-heavy preparation collapsed in cost. The irreducibly human part - advocacy in front of a judge - stayed human. Garfield says it has now handled more than 600 claims and recovered around £500,000 for users, on disputes ranging from about £30 to £10,000. This is not a gimmick. It is a working demonstration that regulated legal AI can move real money through real courts.

For most of Europe, enforcing a contract is uneconomic

Strip away the novelty and the Garfield case is about a very ordinary European problem: someone did the work, was not paid, and faced a legal system too slow and too costly to make pursuing the debt worthwhile. The numbers behind that problem are not small.

The European Commission's EU Payment Observatory has repeatedly found that roughly one in four corporate bankruptcies in the Union can be traced to late or unpaid invoices. In 2024, more than half of EU companies reported difficulties caused by late payment, with average payment periods running beyond 60 days for both business-to-business and government-to-business transactions, and public authorities paying later still. A widely cited European payment study put the cost of the time businesses spend chasing money they are already owed at around €275 billion a year, more than the entire GDP of Finland, with the average SME losing close to €9,200 annually to the chase.

The deterrent is rational. When the legal cost of recovery approaches or exceeds the sum in dispute, the economically sensible move is to write the debt off. Research on small businesses bears this out: only around one in ten small business owners regard lawyers as cost-effective, and roughly half handle legal problems alone. The right to enforce a contract exists on paper. The ability to afford it does not.

So the genuine achievement of a tool like Garfield is not that it is clever. It is that it changes the arithmetic. When preparation costs £400 instead of several thousand, a meritorious claimant who would otherwise have walked away can be heard. That is real access to justice, and it deserves the credit it is getting.

Enforcement is the symptom. The contract is the cause.

Here is the observation that built LegalHQ. Every recovery action is the consequence of an earlier moment that nobody priced: the moment the agreement was signed. The dispute that reached Wandsworth existed because terms were unclear, obligations were poorly understood, or the protections that would have prevented escalation were never in place. By the time a claim is being drafted, the cheapest opportunity to avoid the cost has already passed.

The EU's own policy machinery reflects this. The Observatory notes that most measures against late payment are aimed at prevention rather than enforcement, and the Late Payment Directive itself is being tightened toward shorter, clearer terms. Prevention is acknowledged to be where the leverage is. Yet the tooling available to a European freelancer or small business at the moment a contract lands in their inbox is still, overwhelmingly, nothing. They sign what they are sent.

The cheapest case to win is the one you never have to bring. Garfield made enforcement affordable. The larger prize is making the contract intelligible before anyone needs a courtroom.

That is the gap LegalHQ is built for. It reads a contract the way an experienced commercial lawyer would, in plain language, and surfaces what matters before signature: where the payment terms are weak, where liability is lopsided, which obligations carry real exposure, what is missing that should be there. Not to replace counsel on a complex matter, but to give the people who currently have no one a clear-eyed view of what they are agreeing to. The same logic that let Garfield win at the front door of the court, applied one step earlier, at the desk where the deal is actually made.

Regulated, EU-built, and aware of where the data lives

The path that authorised Garfield is specific to England and Wales. The SRA was able to approve a fundamentally AI-driven firm because it sits within a single regulator's remit. The EU has no equivalent single door. Legal services here are regulated nation by nation, largely through national bar associations, and the rules on who may provide regulated legal work, and how, vary across 27 member states. A UK or US tool cannot simply be exported into that environment and assumed compliant.

The regulatory weather is also changing fast. From August 2026, the core obligations of the EU AI Act for high-risk systems begin to bite, and AI used in the administration of justice sits explicitly in the Act's high-risk category, with the Digital Omnibus package possibly shifting some high-risk deadlines later. Add the GDPR overlap, fines reaching €35 million or 7 percent of global turnover, and a Commission tech-sovereignty package tabled in June 2026, and a single question starts to dominate procurement conversations across European firms: where, exactly, is my client data being processed?

The adoption gap makes this urgent rather than theoretical. A 2026 white paper from Spain's national bar found that around 60 percent of Spanish lawyers already use AI tools, while only 8 percent claimed deep knowledge of what those tools actually do. Europe is adopting legal AI faster than it is governing its own use of it. Tools built outside the EU, on infrastructure outside the EU, widen that exposure.

So "EU-built" has to mean something specific here. Model-agnostic, so no single foreign provider becomes a dependency or a single point of regulatory failure. Data processed under European rules, so the answer to "where does this run" is never a liability. And designed for the realities of European legal practice rather than retrofitted from another jurisdiction's assumptions. Sovereignty is not a slogan in this market. It is a procurement requirement.

Why we built LegalHQ

LegalHQ is EU contract intelligence for the businesses that have always been priced out of it. It takes a contract and tells you what you are signing: the risks, the gaps, the terms that will cost you later, in language a non-lawyer can act on. It is built for the freelancer, the founder, and the small business that experiences a handful of legal issues every year and cannot keep counsel on retainer for any of them.

It belongs to a single thesis that runs through everything yellow3 and naffe.ai do: give European SMBs capabilities that used to be reserved for those who could afford them. Garfield proved that regulated legal AI can recover real money in a real court. We took the lesson and moved it upstream, to the moment before the money is ever at risk. Because the borrowed phrase from that London courtroom is the right one. AI did not replace the judge, the barrister, or the legal system. It made the process accessible. LegalHQ aims to make the moment before the dispute accessible too.

Being right should not be a luxury good. For most of Europe's small businesses, today, it still is. That is the problem we are working on.

Sources

  1. "AI-powered law firm Garfield claims first county court victory," The Law Society Gazette, June 2026 - lawgazette.co.uk
  2. "First court case win for AI-powered law firm," Legal Futures, June 2026 - legalfutures.co.uk
  3. EU Payment Observatory, annual report and analysis, European Commission - single-market-economy.ec.europa.eu
  4. "Late payments cost European SMEs €275bn a year," IBS Intelligence - ibsintelligence.com
  5. "UK legal services access" (citing the Legal Services Board), Lawyer Monthly, 2025 - lawyer-monthly.com
  6. "Regulatory framework for AI" (AI Act timeline and high-risk scope), European Commission - digital-strategy.ec.europa.eu
  7. "A Lawyer's Guide to the EU AI Act," Bloomberg Law, 2025 - pro.bloomberglaw.com
  8. "EU AI Act 2026: What Your Law Firm Actually Needs to Do" (CGAE white paper figures), Gradion, January 2026 - gradion.io