Want more content like this? Sign up for our monthly newsletter.
Legal teams feel the pressure to use AI in their daily work—whether the push is from leadership or a to-do list that never gets shorter. No matter how much teams want to incorporate AI, though, this is still uncharted territory for many. And that means that figuring out the next step is tricky.
According to Gartner, 53% of legal and compliance leaders consider identifying AI use cases a top priority, but only 33% feel confident in their ability to do so. The team at Ironclad is navigating this gap, too.
Caroline Wang, Head of Legal Operations at Ironclad, and Léo Murgel, VP of Product Partnerships and Community Engagement, shared how Ironclad is scaling its own legal AI adoption.
Here’s how they measure success, build out an adoption program, and choose contracting use cases worth starting with.
Define the metrics that prove AI’s impact
Rising AI adoption looks impressive at first, but it has to make an impact to stick around. Results can also help increase adoption. When someone sees that their teammate or another department offloads their most-hated admin task, they may want to try it, too. Tracking outcomes also reveals what isn’t worth the lift so that you can make an educated decision on next steps.
The Ironclad team monitors AI’s impact across three categories:
- Commercial performance: Legal’s work shows up in how fast the business closes deals. As AI takes on more first-pass work, deals should move faster with less hands-on effort.
- Average contracting cycle time
- Percentage of business contracts completed with no legal touch
- Legal efficiency: AI tools should free up legal team time for nuanced work over routine tasks while cutting down on outside counsel engagements.
- Percentage of routine legal questions resolved by AI
- Outside counsel spend
- Risk and operations: Spreadsheets and scattered legacy tools make it harder to catch problems early. Centralizing that information with AI tools helps legal spot issues before they grow.
- Compliance issues identified proactively
- Legacy system configurations retired
Léo Murgel singled out the first category as the one where legal touches business performance most directly. A few hours saved on a small legal team seem insignificant at first glance, but it turns into a lever for more revenue when it cuts the time sellers spend chasing approvals. Those numbers “tend to be bigger than the legal numbers by themselves,” Léo said, and they show leadership what legal contributes to the company.
Start here: Pull your current cycle time and no-touch rate as a baseline before you roll out any new projects. If you want to see how you stack up to your peers, check out our Contracting Benchmark Report.
Build a structured AI adoption program
Getting entire teams or departments to use a new technology is a massive lift, and it can’t happen ad hoc. That’s why Ironclad used an organized six-part program to scale AI adoption:
- Write an AI charter
- Map out an AI adoption plan
- Organize your tool library
- Build shared resources
- Define a learning curriculum
- Have everyone build their first skill
Here’s how the team approached each step.
Write an AI charter
An AI charter defines the rules: where AI helps, where humans stay accountable, and how to measure success. The legal team worked with Jasmine Singh, Ironclad’s General Counsel, to build the charter. Since everyone is expected to use AI in their role, there had to be rules about the line between AI and human tasks.
- Reserve judgment for humans:
- Humans remain accountable
- Live trade-offs where motivation matters
- Context AI can’t supply on its own, like internal vs. external stakeholders, relationship history, and the broader commercial picture
- Let AI handle the first pass:
- Intake and triage for new contracts
- Playbook-based redlines at scale
- Flagging risk and surfacing missing clauses
- Routing approvals and tracking obligations
- Surfacing signed agreements that need renegotiation
Start here: Sort your team’s recurring work into these two columns. Wherever the AI column runs long, you’ve found a potential first use case.
Schedule an AI adoption plan
Ironclad treated AI adoption like any other project, with owners and deadlines so no one fell behind. The team kicked off in June with three monthly milestones:
- Month one: Complete a baseline survey on how the team uses AI today, and finish all foundational AI courses
- Month two: Finish advanced agent-building courses
- Month three: Every team member deploys at least one live agent or skill
In addition to the monthly milestones, there are also weekly touchpoints. Meetings feature show-and-tell time with AI demos, including the ones that didn’t work as expected. The team also tracks metrics and targets while staying up to date with how other departments are using AI.
Start here: Survey your team. Ask how they use AI today and where they see room to automate in their practice area.
Organize your tool library
Most legal teams already have more AI than they realize. Ironclad mapped every tool the team could access by practice area, then looked for gaps.
- General purpose (Claude, ChatGPT, Copilot): Drafting, summarizing, editing, and open-ended research. General AI tools lack legal-specific guardrails, so they may need more human review.
- Commercial (a CLM like Ironclad): Playbook-based drafting, redlining, and negotiation, plus intake, review, signature, and renewal tracking.
- Corporate (entity management, e-billing): Policy drafting, compliance review, board and governance documentation, and regulatory filings.
- Product and privacy (IP management, legal research): Regulatory scanning for GDPR, CCPA, and the AI Act, privacy impact assessments, and IP portfolio monitoring.
Léo’s rule for general-purpose tools is to use the enterprise version your company already procured, never the free version, because of privilege and data-sharing risks.
Start here: List every AI tool your team can access today, by practice area. The empty spots show you what to evaluate next.
Distribute shared resources
Two charter principles drive this part of the process: “learn in public internally” and “participation isn’t optional.” Everyone takes a show-and-tell slot and contributes to the team’s shared resources.
Those resources include a prompt library and an agent brainstorm log, where people document what they build so others can reuse it. Ironclad’s Contract Review Agent has its own prompt library, so the habit carries into contract review.
Start here: Open a shared prompt doc this week and ask every team member to add one prompt that works.
Define learning curriculum
Required courses give the whole team a shared vocabulary for working with AI. Ironclad designated three tiers, with advanced paths for the team’s AI champions:
- Foundational: Contract review prompting, AI fluency, and AI governance
- AI Fluency: Framework & Foundations
- Ironclad Academy: Introduction to Ironclad AI and Learn to Use Ironclad Contract Review Agent (Jurist)
- Your company’s internal course on its enterprise AI tools, if it has one
- Advanced: AI capabilities and limitations, plus prompt exercises
- Anthropic AI Capabilities and Limitations
- Google Skills AI Boost Bites, a series of short prompting videos
- Building skills and agents: Automation, tools integration, APIs, and MCP
The advanced tier covers habits that make prompts work better, like giving instructions one at a time instead of all at once or starting a fresh chat for a new task. The top tier introduces MCP, the protocol that lets AI systems talk to each other. Léo pointed out that MCP is how Ironclad’s contract intelligence shows up inside tools like Claude, so a seller can ask about a contract without leaving the tool they work in.
Start here: Assign the foundational tier with a deadline, the way Ironclad did in month one.
Have everyone build their first skill
Ironclad planned a hackathon for month three but needed a gentler start. At the team offsite, everyone copied a shared prompt and customized it into a start-of-day digest that scans their email and Slack and prioritizes their work. Building the same skill side by side made the first one less daunting.
From there, each practice area builds its own, because Léo argued that only the people doing the work know it well enough to find the best use cases. A super user in each practice area can help colleagues get there.
Every team member adds ideas to a shared agent brainstorm log. Ironclad sorts them into four categories:
- Contract intelligence: Negotiation analysis and renewal management.
- Task assistance: Report builder and contract review.
- Decision intelligence: Patent review process and international expansion.
- Autonomous execution: Ask Legal Agent and marketing review.
Start here: Run one group build session before asking anyone to build on their own.
Start with purpose-built contracting agents
For leaders who aren’t yet confident identifying AI use cases, contracting is the easiest place to start. The work is high volume and repeatable, and purpose-built agents already exist for it.
Léo warned against relying on open-ended prompting alone. “Open prompting AI is often a bit of a trap,” he said. Agents built for one specific job see much higher adoption. Here are the six use cases Ironclad recommends starting with:
| Use case | What it does | Outcome |
| Contract intelligence (Ironclad Agent) | Analyzes portfolio trends to reveal leverage, negotiation patterns, and risk, and serves as the team’s first place to search and answers how-to questions for newer users | Reduced risk and better margins |
| Intake and triage (Intake Agent) | Centralizes contract requests, extracts required details, prefills intake forms from third-party paper, and routes submissions to the right owner | Faster contracting cycles |
| Intelligent archiving (Archive Agent) | Reviews, corrects, and verifies contract data before it reaches the repository | Reduced risk and manual effort |
| Risk review (Contract Review Agent) | Flags legal, financial, and operational risks based on your playbooks and historical positions, with an attorney approving its output before anything goes out | Reduced exposure |
| Drafting and negotiation (Contract Review Agent) | Suggests clauses and fallback language to speed up negotiations and enforce standards | Faster time-to-revenue |
| Renewals and expansion (Renewal Agent) | Surfaces upcoming renewals with history and terms, summarizes risk, and reduces unwanted auto-renewals | Reduced spend and value leakage |
With Contract Review Agent, a legal team uploads a vendor agreement and runs a redline prompt from its library. The agent flags non-standard language and explains the reasoning behind each suggested edit, and the reviewer accepts or rejects every change.
Start here: Pick the agent that matches your biggest bottleneck, whether that’s intake volume, review backlog, or missed renewals. Roll it out to one team, track the metrics you defined at the start, and expand from there.
Built for legal teams of 10 or 500
Ironclad’s legal team has 10 people. Léo’s legal team at Salesforce had nearly 500, and he says the same principles applied there. Whatever your team’s size, the program starts the same way: set the rules, build the habits, and give AI the first pass on work that’s ready for it.
See how Ironclad’s Contract Review Agent takes the first pass on your team’s contracts by requesting a demo today..
Ironclad is not a law firm, and this post does not constitute or contain legal advice. To evaluate the accuracy, sufficiency, or reliability of the ideas and guidance reflected here, or the applicability of these materials to your business, you should consult with a licensed attorney.
