Scaling AI-Enabled Legal Teams

With each day bringing a fancy new AI tool and use case, what’s the best way to scale an AI-enhanced legal team? It turns out that there are a few repeatable steps that can help an organization expand its AI adoption to improve efficiency and costs without sacrificing coordination and risk management. Here is a playbook that will help your team move forward with AI adoption plans in a systematic way.

Define Your AI Vision and Risk Appetite

Successful scaling begins with identifying what success looks like. AI can bring benefits like: faster execution, reduced costs, and increased coverage. Define the outcomes your team is optimizing for, and which metrics you will deprioritize when tradeoffs occur. Lastly, get explicit about the organization’s risk appetite for AI. If you operate in a highly regulated space or collect sensitive data, additional mitigations may be needed, which will be discussed below.

Once the vision is in place, the next step is preparation. Inventory the legal team’s workflows and segment them first by risk level. This may look like:

• Low Risk: NDAs, intake forms, internal FAQs and summaries

• Medium Risk: Commercial contracts, playbooks

• High Risk: Employment matters, regulated products, sensitive data

The organization’s AI risk appetite determines where (and if) AI can be used. Establish principles for how each risk category will be handled. For example: a human must be in the loop for all AI use, or GC signoff is required for AI usage in high-risk scenarios.

Train, Pilot, and Scale in Waves

Scaling AI-enabled teams works best in phases. These phases will take place across both: 1) the work product where AI is being used, and 2) across the teams involved as AI implementation scales.

Scaling Work Product

To find the work product that is the best fit for AI, categorize low risk tasks by urgency and impact. The best candidates for a pilot are tasks that are low-urgency and high-impact. As the pilot matures and the team accumulates success stories, then AI can be integrated into medium-risk workflows following a similar process.

Popular examples of legal tasks that are good candidates for the first wave of AI implementation are often internal-facing bots. These automations can answer questions about company policies, provide status updates to other departments, route sophisticated questions to the proper team member, and create first drafts of repetitive documents based on the company’s style. Try to have an owner for each tool so the team can keep a pulse on experiments, and also prevent duplication of tools for similar use cases.

As the rollout expands, continue to track its effectiveness, particularly on the metrics the team is prioritizing (e.g., speed, spend, coverage).

Scaling Teams

Once the necessary alignment is secured for the pilot (e.g., leadership, InfoSec), continue to deepen skills across the organization in three segments: legal team only, close cross-functional partners, and the broader organization.

Within the legal team, conduct internal training on new AI use cases, create a shared repository of successful prompts, and continue to build a spirit of experimentation through brown bags as the team finds (and shares) new applications with each other.

Scale from legal-only learnings to building cross-functional use cases as AI fluency and trust increases across the organization. Here, legal can collaborate with teams such as sales, customer success, and HR to develop joint workflows. Many of these will be the low risk, high impact scenarios identified in the previous section, but now both parties have the AI knowledge and buy-in to get an experiment off the ground.

Finally, AI implementations can be rolled out to the broader organization, depending on the company’s risk appetite. Ongoing learning, templates, and access permissions can help the organization stay current and also surface new opportunities for leverage. Continue to clarify roles and responsibilities as AI usage expands. For example: the general counsel identifies AI risk appetite in partnership with business leads; practice area leads identify default positions and escalation criteria for playbooks; legal ops will design and iterate on workflows; and all lawyers are expected to use AI in their work where permitted and surface any new issues or risks.

Sidestepping Common Pitfalls

Here is how to avoid some of the most common pitfalls that can arise while scaling an AI-enabled legal team:

1. AI Experimentation Without Leverage

AI can be fun to experiment with, but a frequent challenge is translating those fun experiments into repeatable processes that improve outcomes. To fix this, teach the team when AI should be part of the standard operating procedure, and how to do it consistently. For example, the team might align that all NDAs are reviewed by a specific AI tool (and prompt) before human review.

2. Regulatory Hypervigilance

There are valid concerns about AI, especially for companies in highly regulated industries or dealing with sensitive data. The pitfall, however, is when valid concerns overextend and prevent AI learning and usage in lower risk areas of the organization. To avoid this pitfall, conduct a deeper inventory of low risk scenarios where the organization can start building AI fluency, such as brown bags and hackathons.

3. No Iteration

An important element of scaling is continuing to refine AI prompts, processes, and tools after they are deployed, instead of allowing the system to stagnate and mistakes to repeat. Carve out time each quarter for a quick retro on where AI showed the greatest benefits and failures, and where it could be improved next quarter.

4. Lack of Ownership

Odds of a thoughtful and coordinated roll-out are slim if nobody is explicitly tasked with driving AI in the legal department and beyond. Name one DRI (directly responsible individual) or identify a working group that can meet targets and keep the organization's AI integration on track.

5. Disorganized Tooling

As more tools become available and team members get more comfortable doing AI experiments, your organization’s tooling can become disorganized in two ways. First, tools might become fragmented, with a different tool needed for each step of the workflow. Second, the team might also wind up with duplicative tools that can accomplish the same goals. As the organization matures in its AI use, designate a standard tool for each use case, and periodically reevaluate that tool based on its performance.

Conclusion

Scaling AI-enabled teams begins with understanding the organization’s AI vision and identifying the specific benefits the technology will ideally unlock. Inventory legal workflows to find low risk processes where AI could have an impact and begin the pilot there. 

As familiarity and momentum build, AI can be scaled outward to more complex workflows and broader teams. Standardizing AI processes, defining owners, and continuing to iterate while staying within the organization’s risk appetite are keys to reaping the benefits of an AI-enabled team over time.