Augment or empower? The new split in AI for legal.
There are still no “AI lawyers,” no bots running matters end to end or standing up in court.
What we’re seeing instead is far more interesting: hundreds of focused solutions that chip away at specific problems. Taken together, they’re quietly reshaping the way legal work gets done.
And there’s a clear split emerging.
Some tools are built to help lawyers do their day jobs better, acting as co-pilots that draft faster, summarize smarter, and make the day a little easier. Others are built to deliver work directly to clients or to the business through self-service tools.
That split matters because it shapes who buys what, how those tools are built, and how much impact they really have.
Two different incentives: law firms vs in-house teams
It’s no surprise that most law firms are investing in co-pilots. Time-based billing rewards efficiency only up to a point. Tools like Harvey and Legora let lawyers do the same work faster without changing the business model too much.
In-house teams, on the other hand, have different incentives. They’re measured by impact, not billable hours. They’re under constant pressure to scale, to say “yes” more often without burning out their team.
That’s why they’re reaching for both types of tools: co-pilots that make their people more effective and self-service platforms that help the business help itself.
This is where the real transformation is happening.
Why trust, control and visibility matter
Self-service AI tools change the stakes.
When a lawyer uses a co-pilot, they can check the work before it goes out. But if a self-service tool is answering questions or generating contracts on behalf of legal, it’s representing the function to the entire business. That’s a whole different level of responsibility.
To make self-service viable, the AI has to be something legal and compliance teams can actually control. It needs to be accurate, transparent and explainable.
In practice, that means designing for three things:
Verification: confidence that the answers your tools provide are right and stay right.
Control: the ability to edit, update and govern how those tools behave without relying on a developer.
Visibility: clarity on what users are asking and what the AI is saying back.
Cityblock Health’s General Counsel, Wendy Chow, put this in real terms in conversation with my Josef co-founder Sam Flynn. The key wasn’t just speed, it was trust.
“The business copying the EIN out of a document is no different from me doing it, but they still want Legal to give them the answer so they’re 100% sure they’re not messing things up.”
“With self-service tools, they still get that it’s coming from Legal, but we don’t have to spend our time doing it at scale.”
Augmentation vs empowerment
AI in legal isn’t a single category; it’s a spectrum.
Co-pilots are inward-facing and augment the work of legal professionals. Self-service tools are outward-facing and empower the business.
Augmentation helps lawyers themselves move faster. Empowerment changes how legal expertise is distributed.
To my eyes, when all the right controls are built in, it’s the latter kind of tools that offer far greater transformative potential for legal teams.
Picking the right tool for the job
When legal teams take ownership of AI, not just as users, but as creators and stewards, something shifts. They move from reactive to proactive. They stop being the bottleneck.
The business gets reliable answers when it needs them. The lawyers get to focus on complex, strategic, human work.
The future of AI in legal won’t be defined by fully autonomous systems replacing lawyers. It will be defined by empowered professionals using the right tools for the job and complete control over how they are used.
Tom Dreyfus is cofounder and CEO of Josef and leads the company strategy and global relationships with customers and technology partners. He brings a global perspective on the legal industry, with experience working in law firms, universities, in-house teams and the court system both in Australia and the United States.
