AI and Access to Justice: Measuring the Gap Between Promise and Reality
A seven-pound-fifty letter before action is a genuine breakthrough. But let us be honest about what AI can and cannot fix in a system that was broken long before large language models existed.
I am a technologist. I build things. I believe in technology's capacity to solve problems. I also believe that honest assessment is more valuable than hype, and the access-to-justice conversation around AI needs a dose of honesty.
The headline numbers are compelling. Garfield, the first SRA-authorized AI law firm, delivers a letter before action for seven pounds fifty. Traditional services quoted 1,080 pounds for the same work in the Channel 4 Dispatches blind comparison. A senior solicitor judged Garfield's output "acceptable in a court of law."
That is a 99.3% cost reduction. In any other industry, that would be revolutionary. In legal services, it addresses one specific barrier in one specific context while leaving the structural problems largely untouched.
Let me explain both sides.
What Garfield Actually Demonstrates
The pricing structure is real and meaningful:
- Polite chaser letter: two pounds
- Letter before action: seven pounds fifty
- Full small claims representation: one hundred pounds plus VAT
A business owed five thousand pounds previously had no economically rational path to recovery through legal channels. Spending a thousand or more on professional fees to recover five thousand does not make sense. Spending a hundred does.
The platform handles limitation period checks, claim validity assessment, pre-action letter generation, claim form submission, defense responses, counterclaim management, document production, trial bundle preparation, and skeleton arguments. The full workflow from demand letter to trial.
The documented results include a seven-thousand-pound debt recovery for seven pounds fifty in professional fees. At that ratio, the economics of small claims change completely.
The Institutional Signal
Lord Justice Birss, deputy head of civil justice, described Garfield as "absolutely at the core of what we can do for access to justice." The Justice Select Committee called it "ground-breaking." SRA Chief Executive Paul Philip said: "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."
When the deputy head of civil justice publicly endorses your technology, that is not just marketing. It signals that courts will receive AI-assisted filings without additional scrutiny — at least within the small claims scope.
The Numbers That Matter
Here is the context that makes Garfield significant. The Ministry of Justice reported over 1.73 million county court claims in 2024. Small claims take nearly 50 weeks to reach trial on average. The vast majority of these claims proceed without professional assistance.
Not because the claimants do not want lawyers. Because they cannot afford them.
This is the definition of a market failure. Demand exists. Supply exists. The pricing mechanism prevents the two from meeting. AI changes the pricing mechanism. That is genuinely valuable.
CaseCraft AI has secured 550,000 pounds in funding to address similar gaps. The ABA Task Force Year 2 Report documented over 100 AI use cases in legal aid settings across the US. The pattern is clear: investment is flowing toward underserved legal markets that traditional firms cannot profitably address.
The Honest Limitations
Now here is where I think the conversation needs to be more careful.
Scope. Garfield handles debt recovery under ten thousand pounds and is exploring housing disrepair claims. That is a narrow slice of the access-to-justice problem. Complex matters, contested facts, novel legal questions, multi-party disputes — all remain outside scope. The cases that most desperately need professional assistance are often the ones that AI systems are least equipped to handle.
The digital divide. Users need digital literacy and internet access. The populations most in need of legal assistance often face exactly these barriers. An elderly person being defrauded by a telemarketing scheme, a recent immigrant navigating an employment dispute in a second language, a person with disabilities dealing with a housing issue — these individuals may not be able to interact with an AI legal platform at all.
Technology tools serve digitally capable populations. Access-to-justice gaps are widest among digitally marginal populations. There is a mismatch that technology alone does not solve.
Procedural complexity. When matters become complicated — counterclaims, evidentiary disputes, procedural motions — AI systems may reach their operational limits. The transition from automated to human assistance is not always smooth, and it may come at precisely the moment when the litigant is most vulnerable.
Appeals and escalation. If a small claims decision is appealed, the AI platform's scope likely ends. The claimant needs to find human representation for a process they may not understand, at a point where the stakes have just increased.
What AI Cannot Fix
The access-to-justice gap has structural causes that no amount of AI will address:
Litigation costs beyond fees. Court filing fees, time off work, childcare during court appearances, travel to court locations — these costs persist regardless of whether legal representation is human or AI.
Power imbalances. A tenant facing eviction by a corporate landlord with a legal department faces a power asymmetry that a seven-pound-fifty letter does not solve. AI can make the tenant's case more efficiently, but it does not equalize the resources available to each side.
Substantive complexity. Some legal problems resist simplification. Family law disputes involving child custody, employment discrimination claims requiring witness testimony, personal injury cases needing medical evidence — these require the kind of judgment, advocacy, and human interaction that AI systems do not provide.
Psychological barriers. Many people who could benefit from legal action do not pursue it because of anxiety, distrust of the legal system, or fear of retaliation. An AI platform does not address these barriers any more than a traditional lawyer does.
Measuring Impact Honestly
If we are going to claim that AI improves access to justice, we should measure it properly. The metrics that matter:
Volume of matters handled that would otherwise go unrepresented. Not total matters handled — specifically the increment that would not have been pursued without the AI option.
Outcomes compared to traditional representation. Are AI-assisted claimants winning at the same rate? Recovering the same amounts? Reaching resolution in comparable timeframes?
User satisfaction and comprehension. Do users understand what the AI is doing on their behalf? Do they feel represented? Can they make informed decisions at the approval gates?
Error rates and correction processes. When the AI gets something wrong — and it will — how is that detected and corrected? What happens to the user during the correction process?
Cost per resolved dispute. The true cost, including any human intervention required, not just the sticker price.
Garfield and similar services are beginning to generate these data points. But the evidence base remains thin. Early results are promising. Comprehensive validation will take years.
The Realistic Assessment
AI-powered legal services are not hype when applied to appropriate use cases. That seven-pound-fifty letter before action delivers genuine value to people who previously had no economically viable option.
But AI is not a comprehensive solution to the access-to-justice crisis. It is a tool that addresses specific cost barriers in specific contexts where the legal problem fits a standardized process. That is valuable. It is also limited.
The honest framing: AI expands access at the margins where cost is the primary barrier and matters fit standardized processes. For the hardest access-to-justice problems — where complexity, vulnerability, and structural inequality intersect — we need policy interventions that go far beyond technology.
Anyone who tells you AI will "solve" the access-to-justice crisis is either naive or selling something. Anyone who dismisses it as irrelevant is ignoring real people getting real help that they could not afford before.
The truth, as usual, is in the middle. And the middle is worth building toward.
Key Takeaways
- Garfield's seven-pound-fifty letters before action versus one-thousand-pound traditional fees represent a genuine cost barrier breakthrough
- 1.73 million county court claims annually in the UK, most without professional assistance — the unserved market is enormous
- Scope limitations are real: AI serves standardized, high-volume dispute types but not complex, contested, or novel matters
- The digital divide means the populations most in need of legal assistance may be least able to use AI platforms
- Honest measurement requires tracking incremental access, comparative outcomes, error rates, and true cost per resolved dispute