Corporate Legal Departments Start Embracing AI, Cautiously

Generative artificial intelligence is on the verge of transforming how corporate legal departments deal with routine functions—even as it creates new obstacles that must be navigated.

By harnessing AI to do the grunt work on tasks like reviewing simple contracts, in-house counsel can reduce the time and money needed to perform those tasks to a fraction of what would be needed by humans.

But generative AI is new, and the industry is navigating critical questions around the security and quality of data.

“We are at a pivotal moment” in the evolution of legal operations technology, said Mary O’Carroll, the former president of the Corporate Legal Operations Consortium and former head of legal operations at Google. O’Carroll is now the chief community officer at Ironclad, a contract management technology company that’s building generative AI into its products.

A recent increased focus on legal operations technology has been driving investment, she added, “and AI is just this rocket propelling all of us.”

‘Pilot Mode’

Many in-house departments are already using automation tools to streamline tasks like contract management.

For example, companies are automating “self-serve” non-disclosure agreements to easily swap in language reflecting different state laws, said Brian McGovern, executive director of corporate legal and claims at Mitratech, an enterprise software company that sells legal workflow and automation products.

These tools are useful for “high volume, low complexity tasks,” he said. “NDAs a couple of years ago were taking a month to negotiate, now they’re getting done in five minutes.”

Generative AI moves beyond those automation tools because it isn’t just operating according to specific instructions from humans. It can interpret language and produce original text—such as red-lining or suggesting edits based on its interpretation of other contracts a company has negotiated.

Using AI, legal departments can carry out certain work as much as 20 times faster, O’Carroll said

Currently, faced with a crowded market of AI and automated legal technology vendors, companies are still weighing and testing their options, said Sean Monahan, a senior director at Harbor, a legal technology and operations services company.

“The vast, vast majority of legal departments are in pilot mode or evaluation mode, much more than application mode,” he said.

Legal technology companies, including Ironclad, said many of their customers are already using AI tools.

Early adopters are likely companies that are already well advanced in adopting technology into their legal operations, O’Carroll said.

Security and Accuracy

But adoption of the technology brings its own set of concerns.

Legal departments worry that sensitive company or customer data could be compromised by feeding it into a third-party platform, or that they could run afoul of data privacy regulations in the US or abroad. There’s also a risk that proprietary information could become part of the platform’s training model, if the service is built directly on an open platform and the information isn’t walled off.

Companies could also run into intellectual property law issues. When information is fed into a large language model, Monahan said, the question is: “Who owns it?”

Because generative AI can “hallucinate”—turn out incorrect answers—humans need to check the work. That could mean hiring dedicated staff, Monahan said.

Training Data

The AI can only “learn” from the data it can access—but privacy and security constraints limit the amount of real-world data, such as actual contracts, the tool can be fed.

Finding good training data “is a major problem” for training the models, said Jason Boehmig, CEO and co-founder of Ironclad.

When generative AI first burst onto the scene, Monahan said, the industry was saying, “If we can point this at our stuff, it’s going to be awesome.”

But, he added, “What are you pointing? What stuff? Are you sure it’s yours?” and, “Are you sure it’s not your client’s?”

There’s publicly accessible data—contracts in SEC filings, for example—but “that would be like training a car dataset on cars that wound up in the impound lot,” Boehmig said. “It’s highly not representative.”

Bad training data can also create bias, a potentially expensive problem.

Once there’s bias in the model, it’s difficult to un-corrupt it without millions of dollars of training, said Jerry Ting, founder and CEO of Evisort, a contract management technology company that focuses on AI.

Companies can train models themselves, but that can be costly, Monahan said.

“Fine-tuning” the data going into the model, and its responses, is a challenge, he added.

What’s Next?

Currently, sentiment toward the new technology runs “a wide gamut” across the legal ops industry, and even within companies, Ting said.

He said he expects to see most companies within a year make an initial investment in automated negotiation and automating the manual tracking of when contracts expire. In five years, Ting said, “the legal industry will be completely changed. But I think it’s going to take some time.”

The speed at which generative AI is evolving is one reason to be hesitant for now, said Ed Sohn, head of solutions at Factor, a legal services company.

“We believe it has insane potential, really a massive opportunity,” Sohn said. But, he added, “we think it’s really important—because of the speed at which it’s developing—that we’re cautious to not over-build around the present state of the art because what’s really awesome in June is outmoded in July.”

Given the pace of change, “obsolescence” of the new technologies is a major concern for legal departments considering investing in them, Monahan said.

For now, the industry is grappling with questions kicked up by the new technology.

“This came on us really fast,” McGovern said. “We haven’t had time to figure out all the things that are going to go wrong.”

As adoption moves forward across the industry, “the laggard won’t be the technology,” he said. It’ll be the law, the precedent, and attitudes toward it, he said.

Like any new, shiny object, AI is the center of attention now, said Brittany Leonard, general counsel at Civix, a public-sector technology and services provider. Once the hype dies down, it’s likely to become comfortable and useful for in-house counsel, she said.

Meanwhile, the technology is also evolving rapidly.

“Stuff I thought might be achievable in my career got achieved last month,” Boehmig said.

Bloomberg Law competes in this market and sells AI-based tools that provide contract solutions.