'Hire 30 data engineers tomorrow': How law firms win the AI race

If Alexander Kardos-Nyheim had five minutes with a Big Law managing partner, his advice would be this:

“I’d say hire 30 data engineers tomorrow and get them to live in your iManage, in your SharePoint,” he says. “You don’t need more associates, you need data engineers.”

Kardos-Nyheim is the founder of AI research company Safe Sign Technologies and now co-leads AI research at Thomson Reuters following its 2024 acquisition. Speaking on The Non-Billable Podcast, he argues that the next competitive advantage in law will come from turning institutional know-how into something machines can actually use.

“If you are able to turn your know-how, your precedents into something that LLMs can read, then your LLMs are going to be your biggest champion,” he says. But if firms fail to structure and curate their data, “I’m sorry to say you’re not going to win the battle.”

The real moat is data

That view reflects a broader evolution in legal AI over the past two years. The early wave of “wrapper” products - tools built on top of general-purpose models - is giving way to a deeper competition over models, data and defensibility - including from the Big AI giants.

“I wouldn’t build a wrapper,” Kardos-Nyheim says. “Do not build the ChatGPT wrapper.” A feature advantage, he argues, creates a moat measured in “days to weeks to months.” That’s not how you build a company “that’s going to be around in five years.”

Instead, he believes the battleground is moving under the bonnet - to legal-specific models, retrieval systems and proprietary data. Kardos-Nyheim’s own team focused on building the “engine” rather than the interface, developing a legal LLM trained on domain-specific data.

Legal-specific models, he argues, are “more familiar with what legal reasoning is” and better equipped to deal with “safety, robustness, hallucinations” and “know what good citations look like.”

What does it all mean for law firms? “We will need fewer lawyers, we will need better lawyers,” Kardos-Nyheim says, predicting that the pyramid structure of a traditional law firm will change as AI becomes a “force multiplier.”

For legal tech startups chasing quick wins, he warns that the barriers to entry are becoming much higher. “But if you manage to break beyond that, you can go to the moon and beyond.”

The full conversation is available now on The Non-Billable Podcast.