How Mars Legal Is Leveraging AI for Growth
The age of generative AI has the Mars Legal Department shifting into higher gear. Under Kelly Mickelson, experiments are in, and glossy roadmaps are out.
As Head of Legal Ops at Mars, Mickelson personifies agility. She has to – her large legal department serves a growing 150,000-employee global business with thousands of veterinary practices and such iconic snack brands as M&Ms, Twix, Snickers, Pedigree, and Royal Canin.
That agility definitely means pushing herself out of her comfort zone. A few months ago she surveyed her LinkedIn network for inspiration for applying generative AI to knowledge management. Then she built an AI agent. Herself. In half an hour, she produced a prototype of the Mars Legal Beagle, a chat-based interface for the company’s entire corpus of legal knowledge.
“We're just moving that fast,” she said. “The legal department is trying to be more about agility and testing and learning, so that whatever we're doing today is better than what we did the day before.”
Mickelson sat down with Everlaw to discuss the knowledge management challenges that the Mars Legal Beagle helps solve, her strategies for overcoming resistance to change (which include “tough love”), and how the value of legal work is evolving.
Where is GenAI having the biggest impact for your legal department?
We are extremely bullish about GenAI overall and we're finding exceptional outcomes in knowledge management. With the deployment of the agent — which I built pretty quickly thanks to how great the tech is — we're really unlocking our corporate legal department's ability to leverage the knowledge we already have.
That's especially important for us this year. If we are to be successful with a large pending transaction, our legal department will grow. And these new lawyers and legal professionals will join us and say, "How do we do things in the Mars Legal Department?" And we need an evolved KM strategy to support this.
Tell us about some of the KM challenges that you designed the tool to address?
Like many companies, we pride ourselves in being relationship-driven and service-driven. We really like when we feel helpful, both to the business and to our other legal colleagues. We take pride in and feel warm about those moments that we have together as humans. Resultingly, our knowledge was often exchanged via brokering and conversations and meetings.
But when we peel that back — we know that Mars has 150,000 associates, our legal department resources are finite, and that the scale and the complexity of the business is increasing every day — we know those moments cannot be the way we primarily handle and exchange knowledge.
You’re describing a common challenge for large companies. What were some of the implications of not having legal knowledge stored in an easily accessible central location?
Across our business, we engage trusted outside counsel, and if we fail to bring back the advice of the law firm and store it in an agreed-upon way, we risk receiving duplicative and maybe even inconsistent answers from law firms who have approached those questions differently.
Within the legal department, there are other implications. For example, we risk not being able to flow talent the way we want to because we worry that a newly promoted legal associate will take all of her knowledge with her. We worry about onboarding new legal associates in the most effective way. We risk not capturing the knowledge of legal associates who decide to move on outside of Mars.
As you can see, there are so many reasons why it was critical to nail down the knowledge first before I actually built the agent. In other words, we had to first "Feed the Beagle," before the Beagle could be useful. [Edit. Note: A campaign slogan created by one of the lawyers.]
What does Legal Beagle help your legal team do today?
With “rally hours” — which I created to encourage lawyers to upload their templates, checklists, and other documents — we have uploaded over 1,200 pieces of knowledge. Now we’re able to meet the lawyers where they're at. Rather than asking them to go to a SharePoint site and hope the search results there are good, we ask them to work with the Mars Legal Beagle just the same way they would a human.
Suppose they say, "Prepare an email for me to advise a factory manager on how to approach welcoming visitors to the factory." The agent is generatively grabbing from the knowledge, creating that body of work. But here's the key: the agent always cites the knowledge asset from where it got the answer. And there's a convenient hyperlink, which not only makes it easier, but we think it obligates the requestor to hop out to the knowledge, make sure again that they're applying their human-in-the-loop judgment under our responsible AI guidelines to ensure that it's the correct answer, and the framing of the issue, before they send said email to their factory site manager.
The lawyers who thrive will be the ones who pair the wisdom of the law with the power of the machine.
How do you approach training and support on the new tool?
We held a full Mars Legal Department team meeting to launch our KM strategy and the Mars Legal Beagle. But there wasn't a whole lot of training per se for understanding how to use it because it's just like texting or chatting with any other colleague.
What we've been doing is showcasing — through embedded champions and evangelists — where this has saved them time with storytelling. I was just on a call with our employment lawyers, one of whom was getting ready to kick off a review and evolution of our Americans with Disabilities Act guidance. And one of the other employment lawyers (from a different part of our business) said, "Well, don't do that. I've already done it and I’m feeding it to the Beagle now. Just go look there."
It is far more powerful for our legal associates to hear from other legal associates how these initiatives and tools are working than to have to listen to me as just some legal ops talking head who thinks she knows it all.
What’s been the reception from your lawyers? Are they using it?
Even with a strong knowledge foundation, an agent that's looking into it and delivering the right answers, and me following up and closing knowledge gaps to support the Beagle — we have to encourage the lawyers to use it, and this is, probably, the hardest part. We don't have it all nailed. There are still some legal associates who haven’t installed the agent (in Teams) and even those that have installed it need some gentle reminders about using it as their first stop on their knowledge journey. Getting them to really change the way they exchange and use information is absolutely the real work.
How do you overcome people’s natural resistance to change?
We require folks to go to the Beagle first. Every day someone in Teams asks me a question, either substantive or operational, about this legal department that I know is answerable by the Beagle. And so I say, "Good morning, Legal Associate X. Happy to help, and your first stop is the Beagle." I do not answer the question. I require them to go to the Beagle, ask the question. And then I say, "If you should find that the answer there isn't correct, isn't full or complete, you have doubts or vulnerabilities, you come right on back."
We call that a little bit of tough love around here. But guess what? Very few of them come back because the tech is just that easy. And when they do, that is our spirit of continuous improvement. With each new piece of feedback about the Beagle, it gives me an opportunity to go fill the knowledge gap that was preventing the Beagle from returning a correct or thorough answer.
The billable hour is not going to disappear anytime soon, but the easy hours are going to be the first ones to go.
The right AI technologies promise to make legal teams more effective and more fulfilled. What kind of evolution are you seeing in legal roles as a result?
New roles such as data analysts, legal technologists, and data privacy specialists are going to continue to have a shining moment. Their stock is about to go up, if it's not already high enough. Some corporate legal departments are also talking about hiring legal prompt engineers and AI-specific legal knowledge managers.
How is this going to change the traditional roles? I don't believe that AI will replace lawyers, but in some practice areas, lawyers who use AI may very well replace lawyers who do not. AI is your new superpower. You've got to learn it and you've got to amplify your impact. The lawyers who thrive will be the ones who pair the wisdom of the law with the power of the machine. And that is the way we're trying to approach the message within our traditional generalist or even our specialist roles.
You also have a lot of experience managing outside counsel relationships. How will GenAI tools change how legal services are valued?
We feel like AI dilutes the time-equals-money calculus, and that is especially true for tasks like document review, privilege logging, issue coding, depo prep, and other high-volume, repetitive work. In the case of ediscovery, if AI can cut review sets by 60% to 80% with good predictive filtering and clustering, the time-based billing model really starts to look inflated.
We believe that we're going to have more leverage to push for alternative fee arrangements tied to outputs and outcomes, not hours, or flat-fee doc review even, or project-based pricing. The billable hour is not going to disappear anytime soon, but the easy hours are going to be the first ones to go.
