It’s become the norm for law firms nationwide to launch pilot program after pilot program, testing generative artificial intelligence platforms to see if the technology does what it claims.
Yet at Faegre Drinker, the time for testing is over.
The firm recently made the leap from limited trials to full enterprise-wide deployment, rolling out both Microsoft Copilot and the legal-specific AI platform Harvey to its entire workforce.
Moving Past the ‘Two-Tier Firm’
When Faegre Drinker’s Chief Technology and Innovation Officer Scott Angelo joined Faegre Drinker less than a year ago, his mandate was clear: advance the firm’s innovation program. At the time, the firm was involved in numerous AI pilots. While those programs answered basic questions about the functions of the technology, Angelo quickly realized the limitations of the pilot model.
“What pilots are bad at is telling whether a firm can absorb the technology,” he said. “Pilots also create a two-tier firm. You end up with a small group of people who have the tools and a much larger group who hear about them. That’s not a strategy.”
Faegre Drinker decided to move to full deployment when three critical pieces aligned:
- Guardrails: Robust technical, procedural and cultural boundaries were established.
- Training: A comprehensive AI-use training program was created (including testing).
- Data Readiness: The firm’s internal data house was ordered to ensure AI tools were pointed at a trustworthy foundation.
“The model was never the constraint,” Angelo noted. “Trust, training and data were.”
The Economics of Adoption: Everyone Gets a License
A common pitfall for firms adopting new technology is the rationing of software licenses due to cost, which can create adoption bottlenecks. Faegre Drinker bypassed this issue by securing enterprise-wide access for both platforms, thanks to favorable market competition.
The firm’s AI strategy relies on a two-layer approach:
- Microsoft Copilot: Everyday enterprise layer used to catch up on meetings, summarize email threads and draft routine correspondence.
- Harvey: Legal-specific work layer used for tasks such as reviewing large document productions, analyzing credit agreements, conducting complex research and drafting initial memos.
This simultaneous deployment has shifted how Faegre Drinker attorneys structure their days. The biggest practical change, according to Angelo, is “where the blank page starts.”
He explained, “Six months ago, an attorney sat down to a blank screen and spent the first hour producing something rough. Today they start from a draft and spend an hour on the part only they can do, which is deciding whether the draft is right.”
He added that some tasks previously requiring a hard-to-find three-hour window can now be completed in 20 minutes.
Curiosity, Not Age, Most Predicts AI Adoption
Thus far, the firm’s internal data has revealed that attorneys of all ages have enthusiastically embraced AI.
“I’m happily surprised to say the age thing wasn’t as big as what I thought it was going to be,” Angelo said. “Curiosity was a better predictor than practice area or age.”
Angelo observed that when senior partners in particular adopted AI tools, implementation was more quickly adopted by others. A partner or senior member using AI signaled to associates, paralegals and legal assistants that leveraging AI “was not a shortcut,” he said.
Strict Guardrails and the 96% Training Triumph
To mitigate the risks of hallucinations and data leaks, Faegre Drinker implemented a strict three-layer guardrail system:
- Technical: The tools run in a closed environment where data is never used to train external models. Matter-level isolation and ethical walls are strictly enforced, and attorneys can click directly through to verified sources cited by the AI.
- Procedural: AI output is treated strictly as a rough first draft. “We hold AI output to exactly the standard we would hold a first-year associate’s memo,” Angelo said. “You check the citations; you own it when it goes out.”
- Cultural: The firm reminds staff that a tool can be highly confident yet completely wrong — and that confidence and accuracy are unrelated.
Reflecting high-profile public instances of AI hallucinations in the legal sector, Angelo said, “One hundred percent of the time, it wasn’t the technology. It was that they didn’t have the right technical guardrails, processes or culture.”
At Faegre Drinker, access to AI is restricted until personnel complete an onboarding process that includes a scenario-based foundational AI course, a professional responsibility course revamped by General Counsel for AI and a formal security policy attestation.
The firm’s approach has paid off: 96% of the firm has completed the training.
The Client Reality: Value Over the Billable Hour
Questions from clients about AI have moved from asking whether a firm uses AI to technical compliance questions about adoption rates, exact use cases and economic benefits.
While Angelo believes the use of AI won’t single-handedly kill the billable hour, he suggested firms will need to reframe how they communicate their worth.
“We have to go to different models, a value model that works for everyone, like firm fixed pricing,” he said. “We need to start talking about the value, not the hour.”
Advice for Mid-Sized and Small Firms
For smaller firms looking to leverage AI without the capital of BigLaw, Angelo shared what he has learned through the deployment process:
- Buy, Don’t Build: Building proprietary tools is expensive to create and maintain. With commercial tool prices dropping, “there is no prize for building your own,” Angelo said.
- Target One Pain Point: “Identify one specific workflow causing genuine pain in the organization, solve it well with AI and let that success build momentum and budget,” he said.
- Mandate Training: To prevent costly mistakes, train users to understand the tools before granting access.
- Establish Simple Governance: Start with a simple, one-page document. “Clearly define who can use what, under what circumstances and who ultimately remains accountable for the output,” he said.
As for where the legal industry will stand five years from now? Angelo doesn’t pretend to have a crystal ball but given the rapid acceleration of firm-wide adoption, he is certain of one point: “It’s going to be something special.”