As AI continues to change the practice of law, Law Week Colorado is exploring how Colorado attorneys are using AI in their day-to-day work. This article is part of an ongoing series examining the ways AI is being incorporated into legal practice. We’re interested not only in what attorneys are using AI for, but also what’s working, what isn’t and how the technology is changing the way lawyers approach their work. Would you like to share how you use AI in your practice for the benefit of others in your profession? Please email [email protected].
Throughout my legal practice, I have experimented with a number of artificial intelligence tools designed for legal work, including platforms built specifically for tasks such as contract review and legal research. Some have been very good at the particular tasks they were designed to perform. But the AI platform I use most often, by far, is ChatGPT.
Because of the variety of problems I use it to help solve in my day-to-day work, I rarely begin by asking ChatGPT for a finished answer. More often, I use it to interrogate and narrow the problem itself. I might describe an issue and ask ChatGPT to identify the questions that need to be answered, organize competing considerations or suggest a research framework. If I am working through an argument for a brief, I may ask ChatGPT to attack the argument rather than support it. If I have several possible approaches to a problem, I may use it to compare them and identify the tradeoffs. That iterative back-and-forth is central to how I use ChatGPT, and the newer ChatGPT Work application extends that same approach beyond conversation into more complete, project-based work.
ChatGPT Work gives me a more structured environment for carrying a task from analysis through a usable work product. Instead of simply asking how I might approach an issue, I can provide the relevant materials, instructions, and constraints and work through the task in stages by reviewing information, developing an output, revising it, and checking the result. I have found myself using Work more often for projects that require sustained interaction, particularly when the task is too involved to fit neatly into a series of isolated prompts. ChatGPT Work makes the experience feel less like exchanging prompts with a chatbot and more like working through a difficult project with a capable collaborator.
Like many lawyers, I also use ChatGPT frequently for writing, but not by asking it to produce a finished, file-ready work product from scratch. Instead, I use it as an editor, sounding board and tool for reshaping information I have already developed. I may need to reorganize an analysis, tighten the language, identify gaps, adjust the level of technical detail, or tailor an explanation to a different audience. The real value comes from being able to refine the output through successive exchanges by explaining where a response is too broad, where it overlooks an important distinction, or where one section should be preserved while another is reworked entirely. In practice, the most useful result often emerges several exchanges into the process rather than from the initial response.
A substantial portion of my use is also technical. I regularly use ChatGPT Work while working with Python, automation tools, AI workflows and other technology. Sometimes that means debugging a script. Other times I may be determining how different systems could fit together, translating a manual process into an automated workflow, or testing the logic behind an AI-enabled solution before spending the time to build it. I can describe what I am trying to accomplish, test an approach, identify a failure point, revise the design, and continue from there. In that context, its value extends well beyond generating code. It helps me reason through how a technical solution should work and what the finished product should accomplish.
Prompt development is another area where I use the platform heavily. For complex tasks, an effective prompt is more than just a well-phrased instruction. Getting consistent results may require defining the model’s role, supplying the right context, establishing analytical constraints, specifying the desired output and determining what information should be introduced at each stage. I use ChatGPT to test those instructions, identify ambiguities, and see how changes to the prompt affect the result. That process has changed how I think about AI generally. I have become less interested in asking whether a model can produce an impressive answer from a short prompt and more interested in whether I can design a repeatable process that produces useful results consistently.
Of course, the AI hallucination horror stories floating around these days are a good reminder that none of this means I treat ChatGPT as an authority. For legal work in particular, there is an important distinction between using AI to assist with reasoning and relying on it for the underlying truth of a proposition. Legal authorities, quotations, factual assertions and other materials that matter to the analysis still have to be verified. The same is true of confidentiality and data governance. A useful tool does not eliminate a lawyer’s obligation to understand what information is appropriate to provide it and under what circumstances.
I also do not view ChatGPT as a replacement for specialized legal technology. I have used specialized platforms that were well suited to the work I was doing at the time. As my own practice has changed, however, I found myself doing less work confined to a single category and more work that crosses legal, technical, and operational boundaries. That is ultimately why ChatGPT became the platform I use most.
I now use the same tool to pressure-test an argument, reorganize an explanation, troubleshoot code, design a repeatable workflow, refine instructions for other AI systems, or carry a larger project through multiple stages. That breadth is ultimately what has made ChatGPT my default AI platform. It is not the best tool for every legal task, nor do I expect it to be. But because it can move with me through legal analysis, writing, technical work and workflow design, it has become the platform I return to when I need to think through a problem and turn that thinking into something usable. ChatGPT does not replace the work or judgment the job requires. Its value, for me, is that it has become a flexible part of how I approach that work, even as the problems in front of me change.
Trystan Melancon is an attorney with Fennemore whose practice focuses on construction law, business transactions, and litigation in Colorado. She advises contractors, developers, and subcontractors on contract drafting and negotiation, regulatory compliance, entity formation, and day-to-day business operations. Trystan also represents clients in construction defect disputes, mechanic’s lien claims, collections, and related litigation. She can be reached at [email protected].