TwinLadder Weekly

Issue #6 | April 2025

---

Editor's Note

A friend who runs a small plumbing business in London told me over dinner that he had written off roughly eight thousand pounds in unpaid invoices last year. Not because the debts were disputed. Not because the customers had disappeared. Simply because hiring a solicitor to chase a two-thousand-pound invoice would cost more than the invoice was worth.

This is the access-to-justice problem that legal professionals discuss at conferences while charging conference fees that illustrate the problem. The court system exists. The economics do not. Someone finally built something that addresses the actual barrier.

What interests me is not the technology — it is the pricing. Two pounds for a polite chaser. Seven pounds fifty for a letter before action. When the barrier to legal access drops by 99%, you are not improving the existing system. You are creating a different one. And the regulatory story is as significant as the technology story, because it happened in the UK — where the SRA authorised the first AI-only law firm — and not, as you might expect, in the United States. The implications for European legal markets are worth examining carefully.

---

The Seven-Pound-Fifty Legal Letter: What Garfield.Law Reveals About Our Profession

Garfield.Law was founded by Philip Young, a former Baker McKenzie partner, and Daniel Long, a quantum physicist turned CTO. Their model is brutally simple: AI-powered debt recovery for small claims up to ten thousand pounds, priced at what small businesses can actually pay.

Traditional Solicitor	Garfield.Law	Difference
~GBP 1,500 for debt recovery	GBP 2 polite chaser	99.9% cost reduction
GBP 200-500 letter before action	GBP 7.50 letter before action	97-98% cost reduction
GBP 1,000-3,000 full claim process	~GBP 50 through to hearing	95-97% cost reduction
Fees paid by claimant regardless of outcome	Most fees recoverable under CPR Part 45	Risk transferred to defendant

A polite chaser costs two pounds. A letter before action — the formal demand that satisfies Civil Procedure Rules requirements — costs seven pounds fifty. Claim forms, default judgment applications, even full trial support through to hearing: roughly fifty pounds total. Compare that to fifteen hundred pounds or more from a traditional solicitor.

The economics work because of a single insight: 80% of small claims end at the letter before action stage. Most debtors pay when they receive formal legal correspondence. They just could not receive one before because nobody could afford to send it. That is not a technology insight. It is a market insight that technology made actionable.

The critical technical decision — and this is what I find most interesting from a competence standpoint — is that Garfield's AI cannot propose case law. The system is built on domain-specific procedural knowledge following Civil Procedure Rules. But it deliberately excludes case law citation, the highest-risk area for LLM hallucination. As Lord Justice Birss put it, this approach is "absolutely at the core of what we can do for access to justice."

They did not solve the hallucination problem. They designed around it by building a tool that does not need to cite cases. Small claims court is procedural. You do not need Donoghue v Stevenson to draft a letter before action. You need the correct form, the right deadlines, and proper service. This is a lesson in AI deployment that extends far beyond debt recovery: the most successful AI legal tools may not be the ones that try to do everything. They may be the ones that deliberately constrain their scope to what AI can do reliably.

The SRA's eight-month approval process examined quality-checking processes, client confidentiality, conflict protections, and hallucination risk management. SRA Chief Executive Paul Philip called it "a landmark moment for legal services in this country," noting that "with so many people and small businesses struggling to access legal services, we cannot afford to pull up the drawbridge on innovations that could have big public benefits."

For European practitioners, this raises a question our regulators have not yet answered. The EU AI Act creates a regulatory framework for AI systems, but national bar associations and justice ministries still control who can provide legal services and how. Could a Garfield equivalent operate in Germany, where the Rechtsanwaltskammer controls legal service authorisation? In France, where the Conseil National des Barreaux regulates the profession? In Latvia, where the Advokatu padome would need to consider the implications? The UK moved first because the SRA's regulatory architecture — principles-based, outcomes-focused — was designed to accommodate innovation. Most European legal regulators were not.

The platform is not autonomous. Each action requires client approval. Named solicitors remain accountable under SRA rules. The AI handles drafting, procedure tracking, and document generation — the tasks that drive traditional legal costs. It integrates with accounting software (Xero, Sage, QuickBooks, FreeAgent), enabling automated triggers: invoice 60 days overdue, send polite chaser; no response in 14 days, send letter before action. One small plumbing company recovered twelve thousand pounds in previously written-off debts in its first year. Total Garfield fees: under one hundred pounds.

The defined scope is deliberate and essential. Garfield handles debt recovery where the debt is undisputed. Genuine disputes over service quality or contract interpretation fall outside its remit. A user reported: "The system sent a letter before action for twenty-five hundred pounds. Client responded with detailed complaints about our work. Garfield couldn't help me navigate the dispute — I needed actual legal advice." That is not a failure of the model. It is the model working as designed. AI handles the procedural. Humans handle the contested. The boundary is clear, and that clarity is what makes it work.

---

The Competence Question

Here is what Garfield's model exposes about traditional practice. A freelance web designer with a legitimate three-thousand-four-hundred-pound invoice was quoted eighteen hundred pounds by the cheapest solicitor she could find. She had written the debt off. A seven-pound-fifty letter before action got payment within two weeks.

The legal system did not fail her because of complexity. It failed her because we priced her out.

The Access Problem	The Numbers
66% of small claims result in default judgment	The debtor simply does not respond — no legal argument needed
80% of cases end at letter before action	Most debtors pay when they receive formal correspondence
GBP 8,000 average annual write-off for UK small businesses	Across the economy, billions in legitimate debts go unpursued
GBP 1,500 minimum traditional solicitor fee	Economically irrational for debts under GBP 3,000

The competence question is not whether AI can handle procedural debt recovery. It clearly can. The question is whether the profession can honestly justify fifteen-hundred-pound fees for work that AI delivers at seven pounds fifty. And whether European legal markets, with their own access-to-justice crises, will follow the UK's lead or protect incumbents.

This does not mean all legal work should cost seven pounds fifty. Disputed invoices, cross-border complications, strategic debtors who game the system — these require human judgment that Garfield deliberately excludes from its scope. The model works precisely because it is narrow. But that narrow scope covers a market that traditional firms cannot profitably serve and millions of small businesses across Europe cannot afford to access.

We need to be honest about which parts of our work command premium pricing because of genuine complexity, and which parts command premium pricing because nobody offered an alternative. The arrival of that alternative is not a threat to the profession. It is a mirror. And what it reflects should make us uncomfortable enough to change.

---

What To Do

1. If you advise SME clients, know Garfield exists. Your clients with unpaid invoice problems now have a seven-pound-fifty option. Telling them about it builds trust. Pretending it does not exist erodes it. The same principle applies to whatever access-to-justice tools emerge in your jurisdiction — your clients will find them regardless, so be the one who introduces them.

2. Evaluate your own small claims economics honestly. If you are charging fifteen hundred pounds for routine debt recovery, understand that clients now have a choice. Consider whether fixed-fee alternatives make sense for your practice. For European firms, the Garfield model suggests that similar services will emerge in your market — probably sooner than your bar association anticipates.

3. When recommending any AI legal service, verify regulatory authorisation. Check sra.org.uk/consumers/register/ for UK services. For European services, check the relevant national bar association or legal services regulator. Garfield is SRA-regulated with named solicitors accountable for output. Not every AI legal tool will be. The distinction matters for your clients and for your own professional obligations.

4. Study the scope-constraint model. Garfield's deliberate exclusion of case law citation is a masterclass in responsible AI deployment. When evaluating AI tools for your own practice, ask: what has this tool deliberately excluded? The tools that try to do everything will fail at the edges. The tools that constrain their scope to what AI can do reliably will deliver consistent value. Apply this thinking to your own AI deployments.

5. Monitor the European regulatory trajectory. The UK Artificial Intelligence (Regulation) Bill has been reintroduced to the House of Lords. The EU AI Act mandatory compliance begins August 2026. Different frameworks, different implications for AI legal services. The question for European regulators is whether they will enable Garfield-style innovation or whether incumbency protection will prevail. The answer will shape access to justice across the continent.

---

Quick Reads

• CaseCraft AI raises five hundred fifty thousand pounds for small claims platform expansion — a second entrant signalling market viability. Without SRA authorisation it operates as legal information rather than legal advice, a distinction that matters legally but may not matter practically to the small business owner trying to recover a debt.

• 96% of UK law firms now integrate AI into operations in some form. Adoption is essentially universal, though sophistication varies enormously. The gap between "we use AI" and "we deploy AI responsibly with documented governance" remains vast.

• UK vs EU regulatory comparison: the UK's sector-specific, principles-based approach through existing regulators enabled faster Garfield authorisation than the EU AI Act's cross-sectoral risk-based framework would likely permit. Prime Minister Starmer reaffirms "light-touch" approach. For European firms, the lesson is not that light-touch is better — it is that regulatory architecture determines innovation speed.

• Dechert's analysis of the SRA authorisation provides a useful overview for firms considering the implications for their own AI deployments. Read it alongside the EU AI Office's guidance for a cross-jurisdictional perspective.

---

One Question

If seven-pound-fifty legal letters recover debts as effectively as fifteen-hundred-pound solicitor letters for 80% of cases, what does that say about the value we have been charging for — the legal expertise, or the letterhead?

---

TwinLadder Weekly | Issue #6 | April 2025

Helping European professionals build AI competence through honest education.