As we write this column, we are just back from ABA TECHSHOW. It was fantastic, as usual, and needless to say, there were many sessions that focused on artificial intelligence (AI). And it is probably safe to say that AI was mentioned at every single session at TECHSHOW. On the vendor floor we saw various products that have incorporated AI to varying degrees—and many others that are in the process of doing so.
In our personal and professional lives, we are already seeing examples of AI daily (e.g., Siri, Alexa), and there is no doubt AI is working in the background in ways most of us don’t see. To help you move beyond the hype of AI and see some real-world applications, we would like to give you some practical examples of how you can use AI today. Many of our readers use Teams on a daily basis, which includes deep integration with Microsoft’s AI tool, Copilot.
At TECHSHOW, we bumped into Austin-based attorney Mitch Zoll. He showed us how he is using AI—including Copilot and ChatGPT—to complete repetitive and mundane everyday tasks. Mitch uses a combination of Microsoft Copilot and ChatGPT’s Custom GPTs to complete tasks in a way that provides him with greater efficiency and responsiveness.
A: Most lawyers’ first experience with ChatGPT and Microsoft Copilot is in the initial prompt window. Using their experience with Google searches, most lawyers will start by asking simple questions. Simple questions will result in simple answers. But prompts are very different from web searches. As you work with these platforms you quickly learn prompts are interactive and that more detailed prompts yield better results. For example, my prompts are like the narrative of a partner talking to a first-year associate. I provide context, guidance and even sample case holdings before sending the AI on its task. This helps guide the AI to the information it should consider and the format I want for the response.
Think of Copilot Agents and Custom GPTs as an enhanced way to create detailed prompts with specific knowledge. Instead of putting rules in a prompt window, lawyers can build their own interfaces by preloading the custom AI with very detailed prompt instructions and rules. As a lawyer you likely want to give the custom AI a default personality and perspective (“Expert Texas business lawyer”) and provide it rules on the format of responses to future questions (e.g., paragraphs, bullets or full memos). The custom AI can also learn statutes, memos, templates, or articles it must consider each time before it responds to an inquiry. Every time you interact with this module in the future, the “knowledge” is already in place and guiding its responses.
Custom GPTs are trained on the rules in the profile that you provide, as well as documents you load into the template. But be aware, this involves loading documents onto OpenAI’s servers, and you lose control over confidentiality. Do not put private documents into the system that you intend to keep private. Alternatively, you can train Microsoft Copilot on internal data and point it to firm folders and documents within your Microsoft tenant, but Copilot Agents are not quite as responsive as ChatGPT. In other words, you must choose when to trade a bit of performance for full data privacy.
A: The easiest way to think of custom AIs is to imagine a very smart college student, on their very first day of college, interning at your law firm. They are personally competent but will know absolutely nothing about the law. I like to throw in the assumption that they are also trying to build their influencer lifestyle to help me remember to think about privacy!
I try to think of tasks that I need to complete but where I cannot charge the client my full hourly rate. For example, my clients do not want to pay for two hours of “attorney time” for me to search through a 50-page lease just to get the variables needed to create a lease assignment. They just want a strong lease assignment. With that in mind, I try to find tasks that a custom AI can accomplish quickly to help me produce a better legal product.
Custom AIs are great for tasks where you need it to: 1. summarize substantial amounts of text, 2. brainstorm creative options, or 3. process traditional “secretarial” work. For example, custom AIs can review that long commercial lease and extract pre-identified variables needed for the lease assignment. A custom AI can also review a proposed article and produce 15 possible titles to find the best one for a blog. Finally, custom AIs are great at taking simple, time-consuming tasks—like retyping notes or organizing files—and creating actionable summaries. In each of these situations, the custom AI doesn’t need to know the law; it just needs to understand the task. But these tasks help the lawyer complete legal work.
A: Sure! Here are some custom AIs I built and use regularly. They are not groundbreaking. In fact, that is the point. I am not trying to sit back like the wizard behind a curtain having the computers be my associate lawyer. These are real-world applications that I use regularly to serve clients and help my business run more efficiently and effectively.
Here are some agents I use daily—no black magic, just practical efficiency:
A: My advice to lawyers using AI: Use AI to keep up—not to catch up or give up.
Use AI to take time-consuming but mundane tasks off your plate. Proper use of AI will help you “keep up” with the tasks that previously created bottlenecks in your workflow and free you up to do the real legal work. Used regularly and at the time you are doing the “real” work, AI can be a tool to help you be more efficient and effective.
But do not use AI to “catch up.” I have seen lawyers lean on AI as a crutch to finish legal work they should have done themselves. If you are late on an assignment or overwhelmed, that is not the moment to churn out mediocre work drafted by AI. The output must be something you can check and verify—and when you are racing to catch up, there is too much temptation to glance over mistakes. This is where AI can be a trap, not a tool.
And do not use AI to “give up” your law license. I spoke with one lawyer who marveled that AI had drafted his entire response to a motion in 15 minutes. It “looked great” and was “ready to file!” he told me. That is not how to use AI. These tools can only produce what they believe you want. We have all seen matters where a single footnote in an email or document can turn a case. It is fine to summarize—but you still need to do the real legal work. If you hand over the practice of law to AI, you will miss those critical moments. The thought of “efficiency” is not worth it.
Thanks for sharing your insights, Mitch. Your examples will help our readers have a much better understanding of how they can use AI agents in practical ways in their daily practices. To help future proof your practice, we encourage you to experiment with AI agents. You likely already have Copilot on your desktop––dive in and experiment.
