· AI × Law · Legal Tech

What the World's First AI-Native Law Firm Gets Right

Text size

I just listened to the latest episode of the AI and the Future of Law Podcast hosted by Bridget Mary McCormack and Jennifer Leonard with the founders of Garfield AI, the first AI-native law firm in the world, and their insights on building and running the firm are fascinating.

Key takeaways:

  • The infrastructure advantage

England's small claims court has an API that enables them to programmatically exchange data with the court system. This level of integration is incredibly rare globally and transforms how legal services can operate.

  • They worked closely with the regulators in the development of the firm (8 months)

The regulators were already concerned with people turning to unregulated LLMs for legal advice, which will lead to poorly prepared cases, and burdening an already stressed court system. Therefore, they decided to work with responsible players rather than fight the inevitable.

  • The hallucination solution

They built a hybrid system that combines a "deterministic expert system" with a "probabilistic LLM-based system". Just like how in a law firm where a junior associate's work must be reviewed by someone more senior, the outputs generated by the LLM system at each step must be reviewed and approved by the user.

  • Industry welcomed the innovation

Industry did not resist the innovation, but rather welcomed it. Some law firms even leverage Garfield in their own workflow to serve their clients more economically and scale faster.

  • The pie is getting bigger

The technology makes previously uneconomic work viable, expanding the overall legal market. For Garfield, they provides debt collection under £10k in the UK, where £6-23 billion in small debts go uncollected annually because it was previously uneconomic to pursue them.

My thoughts:

  1. Are there any similar opportunities in Canada? Practice areas or niche issues that are fact-heavy, process-heavy, and currently untapped or underserved because they're uneconomic. Law societies across the country have been pushing innovation and access to justice initiatives. Could we see similar regulatory openness for responsible AI solutions?
  2. If AI makes legal services more accessible and expands the market, we could see a massive increase in case volume. The Canadian court system is already backlogged as is. Will the courts also have to leverage AI in their workflow (fight AI with AI)?
  3. This is a reinforcement of my earlier post on how the AI replacing knowledge worker rhetoric is fundamentally untrue. The HITL (human-in-the-loop) system was a critical aspect of providing the necessary confidence and quality control needed to convince the regulator.

All in all, highly recommended for anyone interested in this space.

Further reading:

References & resources