Guest: Mike Anderson, Chief Product Officer, Filevine
Host: Seth Earley, CEO at Earley Information Science
Published on: March 20, 2026
In this episode, Seth Earley speaks with Mike Anderson, Chief Product Officer at Filevine, about what it takes to bring AI into one of the most demanding and high-stakes environments in the enterprise - legal operations. They explore why AI will not replace attorneys but will dramatically extend what legal professionals can accomplish, how real-time deposition analysis is transforming courtroom preparation, and why information architecture remains the critical foundation beneath every AI capability. Mike also shares why the interface - not the model - is the biggest unlock AI offers the legal industry.
Key Takeaways:
Insightful Quotes:
"The demand for legal services already outpaces supply, and it has for some time. We should be talking about the productivity and extensibility of legal professionals - not obsolescence." - Mike Anderson
"If only I had this analysis of the deposition during the deposition. That one customer comment kicked off an entire depositions platform for us." - Mike Anderson
"You still need the is-ness and about-ness. The interface changes, but the underlying information architecture is still what makes AI work correctly." - Seth Earley
Tune in to discover how legal teams are moving past AI skepticism and building the foundations that make AI accurate, trustworthy, and transformative in practice.
Links
LinkedIn: https://www.linkedin.com/in/michael-anderson-374299163/
Website: https://www.filevine.com
Ways to Tune In:
Earley AI Podcast: https://www.earley.com/earley-ai-podcast-home Apple Podcast: https://podcasts.apple.com/podcast/id1586654770 Spotify: https://open.spotify.com/show/5nkcZvVYjHHj6wtBABqLbE iHeart Radio: https://www.iheart.com/podcast/269-earley-ai-podcast-87108370/ Stitcher: https://www.stitcher.com/show/earley-ai-podcast Amazon Music: https://music.amazon.com/podcasts/18524b67-09cf-433f-82db-07b6213ad3ba/earley-ai-podcast Buzzsprout: https://earleyai.buzzsprout.com/
Podcast Transcript: AI in Legal Operations - Accuracy, Trust, and the Interface Revolution
Transcript introduction
This transcript captures a conversation between Seth Earley and Mike Anderson about the unique challenges and opportunities of deploying AI in legal operations. They discuss why the legal industry is past its initial AI skepticism, how real-time deposition analysis is reshaping courtroom preparation, why Boolean and semantic search must coexist, and why information architecture remains the indispensable foundation beneath every AI capability in legal.
Transcript
Seth Earley: Welcome to the Earley AI Podcast. I'm your host, Seth Earley. On the Earley AI Podcast, we talk about how organizations are getting value from AI, how they're running into problems and solving those problems, and how artificial intelligence is reshaping how organizations manage information, process information, and produce value for customers. Today we're looking at one of the most demanding real-world environments for AI - legal operations and litigation workflows. This is a domain where accuracy is critical, context is everything, and trust is not optional. Joining me today is Mike Anderson, Chief Product Officer at Filevine. Mike has spent many years helping legal organizations modernize how they manage cases, documents, deadlines, and workflows. In this conversation, we'll explore what it takes to bring AI into the legal world in a way that is usable, safe, and valuable, and what other industries can learn from those constraints. Mike, welcome to the program.
Mike Anderson: Hi Seth, thanks for having me. Glad to be here.
Seth Earley: I'd like to start with misconceptions. What do people not understand about AI in the legal world? What are the biggest fears or misunderstandings you run into when you talk to firms?
Mike Anderson: I'll start with one of the misconceptions I hear most in AI tech industry conversations about the legal industry. The first thing people often raise is whether AI could take the job of an attorney or a paralegal. And it's always interesting to me, because I want to ask - have you recently tried engaging legal services? It is very hard. The demand for legal services already outpaces supply, and it has for some time. The American Bar Association just published an article noting that over the last 10 years, the number of attorneys in the United States has grown just 5.6% - that is half a percent per year, basically flat. The world is quite a bit different today than it was 10 years ago, and the underlying changes actually increase latent demand for legal services. Even those who can afford legal services at current prices have difficulty getting what they need. There is way more work already than there are people and attorneys to do it. So what we should really be talking about is the productivity and extensibility of legal professionals - making them able to do more. Nobody's job is becoming obsolete here. We already have far too few of them. AI is going to give those professionals a superpower - the ability to do more things, at higher quality, at a faster pace.
Seth Earley: And I think a lot of people would be surprised by that, because there is this perception that AI performs as well as an attorney, that it can create contracts and so on. But what you're saying is that it's really an accelerator. Though I'll note when we used to sell efficiency to legal organizations, the response was often - wait, I'm going to be able to bill fewer hours? No thank you.
Mike Anderson: Right. The underlying pricing models may adjust, but if you're just talking about legal services output and the demand for that output, it is very high and it outstrips supply. There's also the less measured latent demand - the number of times households come up against civil legal issues and want help but can't get it. A few months ago I wanted legal help on a small real estate issue. I called four law firms recommended to me online. No one got back to me. They just have plenty of work as it is. The billable hour question is real, but legal professionals will have creative ways to continue to get paid for their services even if that model evolves. And there's still a strong case even within a billable hour model, because every attorney has non-billable hours - administrative burden, system-hopping, task coordination - and reducing those is where AI delivers immediate value without touching billable work.
Seth Earley: What else are you seeing in terms of misconceptions and concerns in the industry?
Mike Anderson: The other misconception is just how much business-of-law work there actually is. A law firm is a business. A legal department inside a corporation operates like any other department - it has to contribute and it is held accountable. When you get deep into the work of paralegals and lawyers, you realize that yes, a lot of it is case law and the things taught in law school, but a lot of it is also administrative burden: did the right document get to the right person, did we get it back, coordinating with clients, collaborating with outside counsel, making sure all of that happens in a streamlined way that does not put cognitive load on the people who are doing the higher-order legal thinking. If you help address those coordination and delegation costs, you actually free attorneys to do the knowledge work they are really paid to do.
Seth Earley: The legal industry is very cautious, as it should be. When did adoption begin to accelerate? When did people start to feel like they could trust these systems?
Mike Anderson: I've seen two transitions inside legal. One in the late 2010s, when law firms finally reached a critical mass of moving to cloud-based platforms, driven partly by younger generations who wanted access to case files on mobile devices. In some ways I think that transition was actually harder than the AI transition. In the initial years of AI there was real skepticism and risk aversion. But in the last 12 to 18 months we have really crossed into new territory - a sea change moment. Most of our customers now have whetted appetites and want to see what these tools can do for them. The story has been less and less about suspicion and more and more about genuine interest. As a Chief Product Officer I always try to be about 200 feet in front of our customers in terms of technology and innovation - not so far ahead that it becomes an ivory tower, and not behind where they're asking for things we can't deliver. Two years ago, AI conversations with customers were cautious. In 2025 and certainly now in 2026, the openness and quick comprehension of how AI can help them is very different from what it used to be.
Seth Earley: I'm sure some of that initial trepidation came from well-publicized cases of attorneys going to court with fabricated case law and fabricated testimony. Part of the maturation has been understanding the correct applications, the guardrails, how to prevent and check for hallucinations. One of the things we talked about was real-time analysis of depositions - an attorney can't keep everything in their head when deposing someone, but AI can flag inconsistencies with earlier testimony or a police report in real time. Can you talk about how that developed?
Mike Anderson: A couple of years ago, we had added AI widgets into our platform where customers could choose from a library of prompts - analyze this deposition transcript, analyze this contract, and so on. In those early years we had to seek out the customers who were really taking advantage of it and dig into what they were doing. One of those customers was a huge fan of our deposition transcript analysis. It was doing excellent issue spotting - surfacing things said during a deposition that the attorney had not even remembered, helping them prepare for pretrial matters and trial. Then one customer said to me: "Gosh, if only I had this analysis of the deposition during the deposition." We had our product ears on, and that comment kicked off an entire depositions journey for us. The first thing we built was live transcription during the deposition, running it through analysis in real time and comparing it against the outline and goals the attorney had going in. We flagged when the deponent said something inconsistent with what they said two hours earlier, or inconsistent with an affidavit or police report already in the case file. That has now grown into a full depositions platform with AI experiences pre-deposition for preparation and outline building, real-time transcription and feedback during the deposition, and post-deposition analysis of the certified transcript including pulling clips for trial. We took that one piece of customer feedback through many iterations and it became a core capability.
Seth Earley: When you understand the workflow and the objectives and what those manual processes look like, you can automate, accelerate, and provide new insights. Let's talk about Boolean search versus semantic search, because it is not an either-or in legal. Why does Boolean search remain essential?
Mike Anderson: Legal has something special. When you are responding to a public records request from an Attorney General's office, or satisfying a subpoena or a legal hold, you have to be accountable for how you define the population of documents provided to fulfill that request. That population definition needs to have external transparency - someone needs to be able to look at the logic and say, yes, that is a reasonable way to define that document set. That is the core underlying reason Boolean search cannot go away in legal. But semantic search and RAG have made everyone realize there is a much better way to find information you actually need to do the task at hand. You can take natural language and convert it into rich retrieval across a given matter. So you have to have both. And I would actually say there is a third lens. One is retrieval - semantic search for what you are actually after. Two is Boolean - defining populations with transparent, auditable logic. Three is attribute-based - using the fields and tags and structured attributes of your underlying system to find information. A legal professional needs all three, sometimes in combination and sometimes in isolation, and a great AI experience needs to support all of them.
Seth Earley: That third lens is really information architecture - defining the is-ness and about-ness of your content. It is a contract. What is it about? What parties, what transaction type? Before large language models, AI had been around for decades. We called it text analytics, machine learning, and many other things. The underlying requirement is the same: content structure, system integration, and defined attributes. Those still have to be defined, including for the case itself, which will have its own unique issues and attributes.
Mike Anderson: That is exactly right, and we have seen it play out in real time with our customers. We have an agent in our platform called Lois - Legal Operating Intelligence System. Customers go in and ask questions about their specific case or matter. A significant portion of those questions are things like, what are the insurance policy limits on this matter? There is an interface in the system where they could look that up directly, but they now choose to ask Lois instead. The important thing this illustrates is that you still need a place where the actual policy limit is stored and structured for accurate recall. The conversational interface is a new way in, but it does not replace the underlying data architecture. You can automate classification, but you still need the reference architecture - the fields, the entities, the vocabularies - defined in advance. That is still content operations, still information architecture, still building ontologies. It is as critical as it has ever been.
Seth Earley: What do you see as the biggest impact of AI over the coming years for legal organizations?
Mike Anderson: Interface is the biggest impact, by far. That is the big unlock. The forces of gravity still exist - you still need the is-ness and about-ness - but the interface changes everything. Eight years ago I was helping law firms move from legacy on-premise systems to cloud-based platforms, and the adoption curve around learning new click paths and navigation was enormous. With AI, that friction largely disappears. The ability to just ask a question, make an assignment, track a time entry, set a deadline and have the system surface the relevant statutory deadlines tied to that anchor date - all through natural language - means someone who does not know the click path in a particular system can now do their job in a really straightforward way. I had a neighbor who practiced law well into his 90s. The ability to just type out what you want to have happen - that changes access to these tools completely. Everything else is a distant second. The interface is such a fundamentally better way to work that I now get frustrated when I encounter a system that does not offer it. That conversational interaction is how things are evolving and will continue to evolve.
Seth Earley: And we will see the ROI from the efficiencies gained and the new capabilities unlocked, as well as from meeting that latent demand for legal services. Mike, this has been tremendous. Thank you so much for your time and for a really rich conversation.
Mike Anderson: This has been great, Seth. Really appreciate your time. I've enjoyed it.
Seth Earley: And thank you to our audience. Please subscribe, share it with colleagues, and we will see you next time on the Earley AI Podcast.