Don’t Worry if AI Is Good Enough Yet. Focus on People, Strategy

Most commentators agree that artificial intelligence will have a significant effect on legal services. But people really want to talk about implementation, and they want details. We will provide an implementation framework derived from our experience. Bottom line—the technology will take care of itself, so focus on people and process.

These ideas aren’t a step-by-step guide suitable for all situations, but they will kickstart the conversation. The hope is to go from “How are you using AI?” to “Let’s make this specific change and measure the impact.”

The technology of large language models is actually pretty advanced for most use cases. LLMs are essentially a tool to process unstructured information in a way that scales. That is what makes them special. The real problem is lack of coordination between the right people, the time available, and the hazy complexity of matching the tech to the process.

This mismatch between a technological solution, the people who need it, and the processes to get it to them isn’t a new phenomenon. Penicillin was discovered in 1928 but only became recognized as a medical miracle after they figured out mass manufacturing and World War II created demand for it. The question is how to speed up application development.

To begin the conversation on people and process, we suggest two things. First, establish the overall business strategy before looking at individual software solutions. Second, when you find yourself at a crossroads defining that strategy ask yourself the same questions—what information do I want to process and what is the value in automating it?

Once you have a general answer to these questions, do the following:

Identify the particular workflows where there might be a return on investment. It’s okay to say “might” at this point about the ROI. This return might be about efficiency or effectiveness. Just make sure you know which.

Use those workflows to select particular tools that might be helpful. Limit your consideration to four or five options and find the differences between them that matter. Press the vendors to articulate these differences. One tool may be best for first drafts that rely on precedent documents. Another might be better for the negotiation process. It may require more than one tool, and that is fine. Most workflows are actually a bundle of different tasks.

Navigate the build versus buy decision. The line between these is actually quite blurry. Much that you “buy” is in the process of development and needs to be customized, and most law firms and legal departments that “build” will outsource the work. Many of us fall into one camp or the other, but there are ways of keeping costs down whichever you choose.

Pilot a tool. This can easily fail for want of structure, setting of expectations, and product support for users. Some users will try it once or twice then abandon it when it doesn’t perform the way they imagined—i.e., doesn’t automate the whole thing. You will need some handholding for this group, a way of pushing without being pushy to determine what tasks or steps in the process the technology enables.

Take the findings of the pilot and feed them back into the strategic conversation. Even if the tech works, you need to justify the cost. Some technology will be interesting but not accretive. Some will be clear winners, but now you need to have the people, process, and strategy discussion again. Just because the tech helps doesn’t mean attorneys are motivated by acute need.

Adjustments can be made when this process involves both clients and their outside law firms, but the basics still hold. You take a wider view of the relevant workflows and provide additional structure around use and evaluation. Solutions might present themselves when people aren’t only looking at their part of the problem. The point is to avoid the default to inaction that comes attempting a complex effort of indeterminate value involving multiple organizations.

When law firms invest in AI that process usually occurs with limited input from clients, if any. This means investment decisions are made on a theoretical basis rather than an empirical one and don’t incorporate the whole picture of the business relationship.

Much of what is outlined above sounds pretty obvious when you think about it. But its obviousness highlights its utility compared to having no structure in place. We’re in the early days of working with technology that will bring about profound change. People need to stop thinking about one top-down killer app and engage with bottom-up point solutions that move things in the right direction. The steps outlined above are a framework for thinking about implementation that can show the way forward. Just don’t get lost.

This article does not necessarily reflect the opinion of Bloomberg Industry Group, Inc., the publisher of Bloomberg Law and Bloomberg Tax, or its owners.

Author Information

Matthew Dunne is Katten’s senior innovation and data science manager. He evaluates and implements AI tools into law practice and is a technical adviser to attorneys.

Andrew Sprogis is Katten’s chief innovation officer. He and his team enable the firm’s attorneys, business professionals, and clients to innovate their approach to law and business.

Write for Us: Author Guidelines