LegalAgento: Researching the Access to Justice Crisis

Before I write a single line of code for a new product, I research. I talk to people, read industry reports, study the competition, and try to understand the problem deeply enough to know whether technology can actually help. For LegalAgento, that research phase has been one of the most eye-opening experiences of my career.

The access to justice crisis in the United States is staggering. And the more I dig into it, the more convinced I become that this is one of the most important problems AI can solve.

The Numbers That Changed My Mind

I started researching legal tech because I was looking for the next vertical for the Agento suite. I had a hypothesis that legal services were ripe for AI disruption, but I didn't fully appreciate the scale of the problem until I looked at the data.

According to the Legal Services Corporation, 92% of the civil legal problems reported by low-income Americans received inadequate or no legal help. The World Justice Project found that 1.4 billion people globally can't resolve their justice problems. In the US alone, roughly 80% of civil legal needs go unmet.

These aren't just statistics. These are people losing their homes because they could not navigate an eviction defense. Parents losing custody because they could not afford a family lawyer. Workers accepting wage theft because they didn't know they had legal options.

The legal system was designed with the assumption that everyone would have a lawyer. Most people don't and cannot.

Why Existing Solutions Fall Short

Legal tech isn't a new industry. LegalZoom has been around since 2001. Rocket Lawyer since 2008. There are hundreds of legal tech startups. So why is the access to justice gap still so wide?

Document assembly tools solve the wrong problem. Most legal tech focuses on generating documents: wills, contracts, LLC formations. These are real needs, but they aren't where the crisis lives. The crisis is in litigation, housing disputes, family law, employment issues, situations where people need guidance through a process, not just a template.

Legal marketplaces have the wrong economics. Platforms like Avvo and LegalMatch connect people with lawyers. But the people who most need legal help can't afford the lawyers on those platforms. Reducing the friction of finding a lawyer doesn't help when the fundamental barrier is cost.

Self-help resources assume too much. Court self-help centers and legal aid websites provide information, but they assume users can parse legal language, understand procedures, and apply general guidance to their specific situation. Most people cannot.

The gap isn't in information. It's in navigation. People need a guide that understands their specific situation and walks them through what to do, step by step.

The Unbundled Legal Services Model

During my research, I discovered a concept that became the foundation of LegalAgento's approach: unbundled legal services.

Traditional legal representation is all-or-nothing. You hire a lawyer, they handle everything, and you pay for everything. Unbundled legal services break that model apart. A lawyer might only draft one document, or review one filing, or coach you on how to present your case, charging for only the specific task rather than full representation.

The American Bar Association has been encouraging unbundled services for years, and most state bar associations now permit it. But the adoption has been slow because the administrative overhead of managing many small engagements is high for solo practitioners and small firms.

This is where AI fits perfectly. An AI agent can handle the parts that don't require a licensed attorney (understanding the user's situation, explaining options, drafting initial documents, guiding court procedures) while connecting users with attorneys for the specific moments that require human legal judgment.

The economics flip. Instead of paying a lawyer $5,000 for full representation in a housing dispute, a person might pay $200 for an AI-guided process with a one-hour attorney review of the final filing. That's the difference between accessible and inaccessible for millions of people.

What I Learned from Talking to Lawyers

I spent weeks talking to attorneys: legal aid lawyers, solo practitioners, BigLaw associates, retired judges. The conversations were illuminating.

Legal aid lawyers are drowning. Every legal aid attorney I spoke with described the same situation: overwhelming caseloads, impossible triage decisions, and the constant guilt of turning away people with legitimate legal needs. They aren't opposed to AI help. They are desperate for it. Their concern is accuracy, not displacement.

Solo practitioners want efficiency, not replacement. Small firm and solo attorneys see AI as a tool to serve more clients at lower price points. Several told me they would love to offer $500 housing dispute packages if they had AI handling the initial intake, research, and document drafting.

The bar associations are watching carefully. Regulatory concerns are real but navigable. The key distinction is between legal information (which anyone can provide) and legal advice (which requires a license). LegalAgento needs to operate firmly in the legal information and guided self-help space, with clear handoffs to licensed attorneys when legal advice is needed.

The Technical Research

Beyond the market research, I spent significant time evaluating whether current AI technology is actually capable of the level of legal reasoning LegalAgento would require.

I tested extensively with Claude, running it through legal research tasks, document drafting, and procedural guidance. The results were promising but revealed important boundaries.

What works well: Explaining legal concepts in plain language, identifying relevant areas of law based on fact patterns, drafting standard legal documents from templates, and walking through procedural steps for common court actions.

What doesn't work yet: Citing specific case law reliably (hallucination is still a real problem), understanding hyperlocal court rules (each courthouse has its own quirks), and making strategic judgment calls about how to argue a case.

These boundaries map cleanly onto the unbundled services model. The AI handles what it does well. Licensed attorneys handle what requires human judgment. The user gets help that was previously inaccessible.

What Comes Next

The research phase has given me confidence that LegalAgento addresses a real, urgent need and that current AI technology is capable enough to make a meaningful product, as long as I am honest about its limitations.

The next phase is prototyping. I am starting with housing disputes, specifically security deposit recovery and eviction defense, because they are common, the law is relatively standardized across states, and the stakes are high enough that people are motivated to act but low enough that full legal representation is often not cost-justified.

I'll write about the prototyping process as it unfolds. For now, the takeaway is this: the access to justice crisis isn't a problem that needs a breakthrough technology. The technology exists. What it needs is a product that connects AI capability to real legal needs with the right guardrails, the right user experience, and the right integration with the existing legal system.

That's what I am building with LegalAgento.