Introducing Index Intelligence

Frontier legal AI, built for the people who need the law.

The first generation of legal AI was built for lawyers. It helps them search the law, review documents and draft work faster. This is real progress, and it will change the profession.

But it leaves the larger problem almost untouched.

When something legal happens to you, you often have no idea where you stand. A contract is broken. A threatening letter arrives. A business relationship turns into a dispute. Something important is suddenly at risk, but before anyone explains the problem, you are expected to find the right lawyer, judge advice you are not equipped to judge and agree to an hourly rate.

The lawyers have the information. The person with the most at stake often has the least.

We believe the next generation of legal AI should change that.

Today, we are introducing Index Intelligence, our research and engineering effort to build frontier legal agents for the people and businesses who need the law.

Its first public application is Assess, available now at indexlaw.com.

You describe what happened in your own words and upload the documents you have. Assess asks the questions that matter, studies the record and builds a clear view of the situation: the important facts, the evidence, the law that may apply, the questions that remain open and the credible paths forward.

The aim is simple. Before someone spends thousands on legal advice, they should be able to understand the shape of their position.

Building from the client backwards

Most legal technology begins with the lawyer. It asks how research can be faster, how more documents can be reviewed and how drafting can become more efficient.

Index begins with a different question: what should happen when an ordinary person or business encounters a legal problem?

They should not need to know the correct legal words. They should not need to guess which type of lawyer to call. They should not need three consultations simply to learn whether the matter is serious.

They should be able to explain what happened and receive a rigorous, useful view of where they stand.

I started Index after experiencing the legal system as a client. What stayed with me was not only the cost, but the lack of clarity. The professionals understood the language, the process and the possible outcomes. The people living through the matter did not. Yet they were the ones expected to make the decisions.

Access to justice does not begin in a courtroom. It begins when you can understand what is happening to you.

Clarity is not a minor feature of justice. It is one of its foundations. And yet clarity remains one of the most expensive products in the legal market.

A serious technical mission

Making legal reasoning widely available is not a matter of putting a friendly interface on top of a language model.

A real legal matter may contain years of messages, conflicting accounts, contracts, court papers and missing evidence. The decisive fact may be buried in one attachment. The answer may change because of a deadline, a jurisdiction clause or a detail the client did not know was important.

A useful legal agent must do more than retrieve a rule and produce a confident paragraph. Confidence, as lawyers and language models regularly remind us, is not the same thing as being right.

The system must construct an understanding of the matter. It must separate evidence from assertion, identify what is missing, ask the questions that could change the analysis and explain uncertainty rather than hide it. It must distinguish between what is legally possible and what is sensible in the real world. Most importantly, it must know when a qualified lawyer needs to take over.

To do this, we are building our own agent harnesses, matter-level reasoning systems, evaluations and benchmarks. We are testing whether an agent can understand a complete legal matter, identify the facts that truly matter, find the right law, compare credible strategies and recommend a useful next step.

Many existing benchmarks test narrow tasks such as finding a clause or recalling a rule. These are valuable, but real legal work is less tidy. Several answers may be legally plausible. Only one may make sense for the client.

We want to build systems that can tell the difference.

We will publish more of this work as it develops, including our research, evaluations and failures. Legal AI will become too important to be judged only by polished demonstrations and carefully chosen screenshots.

The standard should be whether it helps someone make a better decision in a real matter.

From intelligence to action

People do not seek legal help because they want an impressive answer. They need something in the world to change.

A contract must be negotiated. A notice must be sent. A claim must be defended. A company must be protected. A dispute must end.

That is why we are building Index Intelligence inside Index Law.

The understanding created at the beginning of a matter should become the foundation for everything that follows. The client should not have to tell the same story again each time a new professional becomes involved. The lawyer should receive a structured picture rather than a pile of documents without context.

Assess is the first agent. We will release more systems for research, drafting, review, diligence, negotiation and disputes. They will share context across the matter and work with qualified lawyers wherever judgment, advocacy and accountability are required.

This is what we mean by an AI-native legal institution: not a chatbot placed on top of an old process, but a new way of delivering legal work when intelligence can become dramatically less expensive.

The first generation of legal AI made lawyers more productive.

The next generation can make the law more understandable and more accessible to everyone whose life is shaped by it.

The law will always be complicated. Understanding where you stand should not be.

Assess is available now at indexlaw.com.

Before you hire a lawyer, Index it.

Join us

We are hiring exceptional lawyers who want to help build the future of their profession, and AI researchers and engineers who want to work on hard problems in agents, reasoning, evaluation, retrieval and long-context systems.

This is still the beginning, and there is an unusual amount left to build.

We would like to hear from people who want to build it with us.