5 Ways In-house Legal Departments Are Using AI 

Banner artwork by LariBat / Shutterstock.com

AI adoption in legal departments is often discussed in the abstract.  

The conversation starts with how “using AI” can make the legal department more efficient, free up bandwidth, or help attorneys make informed decisions.  

What tends to be missing are the more useful questions of what got built, by whom, and whether it worked.  

Many law departments have rapidly moved from exploration to adoption. 

A recent ACC survey found that 85% of legal departments currently use AI within their practice. An uptick from 53% that reported using AI last year.   

How are your peers adopting AI? We asked ACC members to share their stories.

Legal teams provided detailed success stories covering contract self-service, compliance training, knowledge management, and NDA review.

While the tools differ, these use cases have common implementation processes that cut across industries and department sizes.  

Here is a roundup of what each team did.  

---

1. Letting business users draft their own routine agreements (Palo Alto Networks)

A familiar problem many legal departments face is being buried under high-volume, low complexity drafting requests.  

Palo Alto Networks’ legal department responded to its bottleneck of drafting requests by building narrow, task-specific tools in Google Gemini.  

The AI tools provide simple forms where users input key variables (such as counterparty name, contract date, termination date, and reason for termination).   

The AI then uses pre-approved templates and playbooks to generate a first draft, which can be used by the business or sent for expedited legal review.  

Result: Turnaround on routine documents dropped from three-to-five business days to one. 

“The tool saved my three-person legal team an estimated 20-40 hours per month,” Hayden Creque, Director and Senior Corporate Counsel at Palo Alto Networks, told ACC.   

Tip: “The key was not to try building one big ‘AI Lawyer,’” Creque said. “We started with a single, high-pain, low risk use case (Termination Letters). Its success built the momentum and trust needed to develop the next.” 

2. Creating compliance training courses employees will watch (Elastic)

Does your organization have a content problem rather than a workflow problem? 

Elastic N.V.’s challenge was to create training courses that would resonate with a busy, highly technical workforce. Employees often viewed such training as a bureaucratic hurdle.  

The organization used Google's NotebookLM to synthesize its trade compliance training guide. The tool turned its dense guide into a five-to-seven-minute video script tailored specifically for its engineers.  

They repeated this approach to distill enterprise-wide privacy policies, guidelines, playbooks, and control requirements into a four-part training course on data privacy and the use of AI.  

Result: Higher engagement with the content and related resources.  

"It has transformed what once felt like a chore into a tool our employees can use," Tim Bown, Senior Global Trade Compliance Manager, and Melissa Garré, Corporate Counsel for Business Integrity Operations, of Elastic told ACC. 

Tip: “When the AI tool generates a script or other material that is closest to the messaging or content you’d like to deliver, feed it back into the tool as source material.” Bown and Garré said. “This ‘feedback loop’ allowed us to quickly guide the AI toward an output that was most closely aligned to our requirements.” 

3. A closed-universe knowledge base on international compliance (Salesforce)

GenAI can also enable legal teams to quickly provide answers to internal clients without having to do time-consuming research or searches. 

Salesforce Inc. needed a fast but reliable way to answer HR compliance questions across jurisdictions. 

The team had the answers, but they were scattered across outside counsel emails, regulations, and treaties, Yves Nguyen, Director, Global Labor & Employment, at Salesforce told ACC. 

Nguyen’s team uploaded years of outside counsel advice, statutory language, and a master spreadsheet of legal requirements into a source-grounded AI research assistant. 

The tool created a closed-universe repository that the team can query like a legal research tool, complete with citations back to the uploaded sources.  

Result: “Our team is saving a total of about 16-20 hours per month by using this tool,” Nguyen said. The team also avoids outside counsel spending on questions the department had already paid to answer.  

Tip: “When uploading your sources, include dates or other identifiers in the source name that will allow you to more easily switch out or update the source if it becomes outdated,” Nguyen said.  

4. Institutional knowledge management for large teams (Mars Inc.)

Mars Inc.’s legal department includes more than 230 professionals who either generate knowledge or obtain it from outside counsel.  

The department’s problem was that none of that knowledge was stored in a way that made it easy to find or reuse.  

Lawyers spent too much time hunting for information by either asking colleagues or reengaging with outside firms for questions that had already been answered, Kelly Mickelson, Head of Legal Operations and Chief of Staff at Mars Inc. told ACC. 

The team had tried to create a centralized knowledge base before. But that effort stalled out over debates related to taxonomy, metadata, and naming conventions. 

This time the team tried a different approach: get everything into one place first, then use Microsoft Copilot to make it searchable. 

The team built the “Mars Legal Beagle,” an AI agent on Microsoft Copilot. The agent uses a centralized internal knowledge repository to answer queries.   

It also provides direct links to underlying source documents so users can validate outputs and serve as the human in the loop under the company’s Responsible AI Policy. 

Result: “We expect the Legal Beagle agent to reduce our reliance on outside counsel by eliminating duplicate requests for information we already have internally, which will translate into meaningful cost savings,” Mickelson said.   

Lawyers can also access relevant knowledge much faster, which improves responsiveness and allows us to better support the business, she said.  

Tip: “Instead of asking people to contribute source materials on their own over time, we ran one-hour ‘rally sessions’ with each practice group,” Mickelson said.   

“During those sessions, we identified recent work, outside counsel advice, and training materials, and had people upload them on the spot. That made the process manageable and ensured it actually happened.” 

5. Self-service NDA review (Liberty Mutual Insurance Co.)

Liberty Mutual Insurance Co.’s shared service team Enterprise Legal Solutions (ELS) reviews roughly 2,500 NDAs a year for the company's underwriting teams.  

Prior process improvements had cut lawyer handling time from more than an hour per NDA to 15 to 20 minutes.   

The problem was that the NDAs still can create friction for the business because underwriters still had to file legal intake requests and wait 24-48 hours for the return SLA, Nora Marantz, Vice President and Assistant General Counsel at Liberty Mutual Insurance Co., told ACC. 

Marantz shared how ELS is piloting an internal tool that lets underwriters submit NDAs through a self-service platform.   

The AI NDA tool, built by Liberty’s Legal Data Science team, lets underwriters receive immediate AI-generated feedback against US-specific requirements.  

The tool is aimed at helping underwriters self-serve on routine NDAs and reserving attorney review for agreements that require legal judgment.

Result: The pilot is still being evaluated. Given that ELS reviews approximately 2,500 NDAs annually in support of the underwriting teams, even modest reductions in legal review time and business wait time can create meaningful impact, Marantz told ACC. 

Tip: “Start with a high-volume, rules-based workflow where the legal standards can be translated into clear guidance and escalation criteria,” Marantz said. “For this type of use case, success depends less on simply introducing AI and more on designing a workflow that the business can confidently use.”