Publications

Assessing When to Use Gen AI

Use Cases – Identifying use cases that return a positive investment is grounded in focusing on the best use cases for gen AI. This requires consideration of:

• Division of Tasks – Consider your process and push only subtasks best suited for gen AI to the machine.
• Risk Assessment – Identify privacy and ethical concerns, client sensitivities or potential biases.
• Impact vs. Investment – Where the machine is faster, use it to create more time to apply human judgment.
• Verification Needs – An interpretation of dense legal rules requires an intense audit. Brainstorming does not.
• Engagement – Consider the level of human involvement needed – training, oversight and feedback loops.
   

Tools – Evaluating tools, including the scope of their training data:

• Different tools may work best for different use cases.

Features of Large Language Models (LLMs)

Self-Learning – LLMs may be designed to ingest your inputs into the data they use for training, which has implications for security and data protection.

• We must always understand if we are using a public or private tool.
   

Probabilistic – LLMs don’t know facts; they model what is likely to be said based on patterns learned from massive data sets of human language.

• We call inaccurate output hallucinations.
   

Biased – Output generated by LLMs will include the biases present in the data set used for training.

Conversational – LLMs are designed to be conversational and confident, even when presenting statistically likely, but inaccurate, information.

Customized by Workflow – LLMs are trained to be best at certain tasks.

• Legal gen AI tools, for example, can be designed to ground outputs in the law and to cite back to legal databases that allow lawyers to audit generated output for accuracy. They can also be designed to protect sensitive client data.

Gen AI Leadership

Competency Trust – We will begin to rely on tools or teammates once we trust them.

• We learn to trust by testing against what we know.
• This process requires an investment of time.
• This applies to how our clients evaluate us, just as much as it applies to how we evaluate our tools and teammates.
   

Client Communications Practicum – Role-playing and practicing conversations with clients or supervising lawyers about gen AI.

• Using resource pages to field client questions and brainstorm best practices.
• Cataloging and sharing use cases within the firm (knowledge management workflows).

Innovation Leadership

Diffusion of Innovation – Innovation spreads through social groups, led by curious, influential early adopters. Law firms can drive practice innovation by empowering lawyers to embrace change.

Empathy, EQ and Early Adoption – At their core, lawyers are counselors, and the value they bring lies in judgment, persuasion, empathy and trust — qualities AI cannot replicate. Recognizing this helps lawyers overcome skepticism, as their fundamental role remains the same, whether assisted by gen AI or not.

Legal Ethics

Legal teams are bound by ethical considerations including the six “C’s”:

• Candor
• Client Consent
• Collaboration
• Communication
• Competence
• Confidentiality
   

Stradley Ronon’s internal Legal Ethics & Gen AI resource page offers a detailed breakdown of each concept, but most critically, the American Bar Association's Formal Opinion 512 states that before inputting client confidential information into a gen AI tool, counsel should inform the client about:

• Why the tool is being used.
• What kinds of client information will be inputted into the tool and how it may be used and disclosed.
• The benefits, risks and limitations of using the tool.

Process

Process Mapping – Identify workflow steps where AI can add value and improve efficiency.

Prompt Engineering Skills – Craft and refine prompts to get accurate, actionable AI outputs:

• Tailoring prompts by persona, task and tool.
• Auditing gen AI output, asking follow-up questions, and re-prompting for improved results.