On-demand webinar: inside the data on how Legal really uses AI

AI adoption across in-house legal teams has moved quickly. But how much of that usage translates into genuine AI fluency, and can legal teams really trust the outputs they receive?

In our latest webinar, we dive into the data behind our AI Fluency Report with Summize General Counsel Lexi Lutz, Legal Director at CodeRabbit Rory Douglas and legal ops leader Patty Corey for  a look at how Legal is really using AI.

Drawing on Summize’s own first-party research from more than 500 in-house legal professionals across the US, the discussion explored AI adoption, confidence, governance and the role of business context in producing reliable AI outputs.

The discussion highlighted an important gap: using AI regularly doesn’t necessarily mean using it fluently. Here’s what the panel shared about closing that gap.

AI adoption is high, but confidence isn’t the same as fluency

Our AI Fluency Report found that AI adoption among respondents had reached 100% - up from 23% in 2024, according to ACC and Everlaw’s research. Meanwhile, 91% rated their understanding of generative AI as advanced or expert.

These figures suggest that AI has become a familiar part of legal work. However, the panel agreed that adoption alone doesn’t tell the whole story.

As Rory explained, the question isn’t just if lawyers use AI, but how deeply they use it. Summarizing an email or researching a legal issue is very different from building a repeatable workflow that applies a team’s established legal positions.

Patty made a similar distinction between confidence and competence. Genuine AI fluency means being able to assess an output, recognize nuance and apply professional judgment when something is missing or doesn’t reflect the intended outcome.

As Lexi explained, the level of human oversight should reflect the potential cost of an error, whether a decision can be reversed and how much human context is needed.

Our CEO and Co-founder Tom Dunlop recently wrote about how lawyers can continue to use AI without losing the critical thinking skills that make them great lawyers in the first place. Take a look at his article for how you can get the balance right.

Why does business context matter?

One of our AI Fluency report’s most significant findings was the gap between perceived context and actual access to it. Although 75% of respondents believe their AI reflects their organization’s policies and previous decisions, only 14.5% said their AI was automatically connected to business context, such as contracts, playbooks and precedent language.

Without this context, AI tools might struggle to reflect an organization’s established positions, negotiation history and commercial priorities.

Rory shared how his team at CodeRabbit used previous contract redlines to help build a playbook reflecting preferred language and negotiation rationale. However, he stressed that this still requires human review – a position accepted for one deal shouldn’t automatically become the standard for every future negotiation.

The goal is to reduce repetitive work and make institutional knowledge more accessible, while leaving lawyers to focus on the judgment and context that each deal requires.

Building your business context can feel like a daunting task, but the panel agreed that starting small with getting information out of people’s heads and down somewhere (even just a shared document) is the best place to start. Find out more about where to start with your business context in our guide, written by our Director of Strategic Implementations Lara Trope.  

How can legal teams build AI fluency?

The panel also explored AI governance, agreeing that while Legal is well placed to establish guardrails, responsibility needs to be shared across the business. Clear policies, accessible guidance and appropriate oversight can help employees use AI responsibly without driving its use underground.

For teams looking to improve their AI fluency, the panel recommended starting with a repeatable problem, documenting preferred positions and fallback options, and working with legal operations to turn expertise into reusable workflows.

AI-generated playbooks and processes also need ongoing reviews as contracts, policies and business priorities change. Building AI fluency is therefore not a one-off implementation project, but an ongoing process of testing, learning and refinement.

Catch up on the AI Fluency Report

To hear more of the panel’s insights on AI adoption, governance and verification, and the practical steps legal teams can take to make AI more reliable, watch the full on-demand webinar.

Alternatively, catch up on the full findings by downloading the AI Fluency report.