For the last three years, the legal industry has been in a race. A race to adopt AI, to deploy tools, and to make more with less. To show that your legal teams are at the forefront of technology. Everyone in this room has been running that race and so far it has worked, kind of.
First generation legal AI has been amazing. Together with you, we have found better ways to execute legal work at unprecedented speed, quality, and scale. But we made lawyers better at the same work they've always done. We looked at AI and we applied the same model, the same structure, and the same ways of working that already exists. We built and implemented something that is smarter and faster, but isn't fundamentally different. I have a friend here on stage.
The legal industry is ready for something new. Today, I'm here to tell you that legal AI is dead.
Legal AI was instrumental in helping us understand how to apply technology, but it is no longer a fair reflection of where we are. The past six months have been a turning point for our industry. Agentic technology and new foundational models combined with improved harnesses have created a step change in power, performance, and intelligence. Think pre-ChatGPT to a post-ChatGPT world. And this isn't incremental. It is a generational leap and it changes everything for our industry.
This next era of legal technology we call agentic law. And agentic law isn't about making lawyers faster at the same work. It's a fundamentally different model for how legal work gets done and what legal teams can achieve. It creates space for us to rethink everything, to reimagine how the work gets done, from how a team delivers services to how it builds relationships, the role of a lawyer, and the new economic models that emerge in this new world.
We are building at the frontier of agentic law, challenging the old ways of legal work and finding new ways that were impossible until now. Legal AI made lawyers more efficient. Agentic law changes what a lawyer can be, and this transformation is already happening. And for a few years, we have had a front row seat to that very change. At Legora, we have a unique role called the legal engineer. Their sole purpose is to accelerate the realization of value for our customers, immediate, midterm, and long-term. And these are former lawyers who left practice because they believe in what AI can do for the profession. They are deep AI experts who understand the workflows, the risk tolerance, and the culture of change inside of legal organizations, and they are critical to our customers' success. In fact, we deploy as many legal engineers as software engineers, which is unheard of in software history. We have as many lawyers as software engineers.
This AI transformation requires more than just technology. It requires change management, and that is why we've invested so heavily in this team. But it's not just the role of a lawyer that's transforming. Agentic law is changing how our customers are operating, how they work, how they serve clients, and how they build their businesses. And a new operational model demands a new blueprint. In our experience, AI transformation does not happen overnight. And it's not always a linear path. But it typically happens in three large phases.
The first phase is about gaining leverage with AI. AI will handle the high volume, low complexity work which improves efficiency. Lawyers get capacity back, matters move faster, and stakeholders get better services. Most organizations are here today.
The second phase is when teams start to differentiate from the market. Service models start to adjust for the type of work that they're required, enabling new relationships and value functions, building real competitive advantage so that a team of five can hit hard like a team of 50. You compete on new ways of service delivery, not headcount.
And the last phase, I'd say still new territory. It's where legal teams reinvent themselves for an AI-first world. Knowledge is productized and reshaped, deployed at scale like automatic intake, triaging, and responses to legal queries. New business models emerge and AI-native legal teams rethink processes from the ground up rather than just an augmentation with AI. It's where the most forward-thinking organizations are headed. We have the privilege of getting to work with a lot of legal teams that are on different paths along this journey, but some of them are really pushing the industry forward.
Deboice is reimagining how they're delivering expertise through their star platform, which is built on the portal. They are scaling expertise to clients on demand, turning legal work into a continuous service. Mishcon is leveraging Legora to serve startup founders, a subject very close to my heart. They are turning static guides, handbooks, and market standards into AI-powered workflows that are accessible 24/7 because when you are a startup founder, you don't always work normal hours.
Ursa Group, one of the largest banks in Europe, deployed Legora to 250 lawyers, 30 legal departments, and across seven jurisdictions, one of the largest AI deployments in financial services. And these are just three of the many customers who are starting to reshape legal work together with us.
Now, a new kind of lawyer and a new kind of legal team require a new kind of system, an operating system. An operating system is exactly what the name suggests. It's a system for operations, for tools, for structures, for sequences that enable teams to produce outcomes previously thought impossible.
Operating systems have existed almost for as long as technology has been around. It's what William Caxton did with the printing press back in the UK. It's what Henry Ford did with the assembly line, Bill Gates with Windows, and Steve Jobs with the iOS. You'll see where we got our inspiration from. There are many, countless more, but they all have one thing in common, and they made the previously thought impossible possible. And today, I'm excited to introduce the next one. We call it Legora AOS, the agentic operating system for legal work.
The AOS is a single connected system that facilitates the flow of information, communication, and execution of legal work. It is something that we've been building towards for the past 3 years. And today, we are announcing the latest product innovations to bring this system to life. It's six layers, one connected system with every layer engineered to handle the depth and complexity of professional legal work at scale. This is the operating system that will power the next generation of lawyers, and it starts with the foundation models. They are necessary but they are insufficient for legal work. Just like Rory said, they power our platform and many others in different verticals. But an engine is not a car. And a platform built for everyone is at the end of the day built for no one. Horizontal AI is designed for general purpose office work. Legal is different. The horizontal LLMs with legal plugins are not the same as a deep vertical legal burger that I'll show you in a minute. Because what legal teams need is a system of data, domain-specific tasks, management, interfaces, governance, security, workflows, integrations, and increasingly the human partnership of legal engineers. And that's exactly what the additional layers add.
The first layer of our AOS is the agentic harness. Harness is the new cool word. A raw LLM is stateless. It has no memory, no guardrails, and no ability to act across multi-step surfaces. The harness is what gives it a spine. It decides which tools to call, routes actions through guardrails before anything touches client data, carries a task through retrieval, checking, flagging, drafting, reviewing all the way through completion.
Next is data and integrations, document management systems, DocuSign, third-party publishers, data sources, MCP connectors. The AOS plugs into the existing tech stack and unifies everything under a single roof. And on this layer, I'm very excited to announce that we'll be giving you access to our data site integration for all the transactional lawyers in the room very shortly.
On top of that, we are adding context and knowledge. No matter how intelligent an agent is, without the right context, it will never deliver high-quality legal work. Precedents, clause libraries, playbooks, and matter history are all stored in the AOS on top of the integrations. Then, legal skills and capabilities make the agent a legal expert. Ways of thinking, workflows, skills, instructions, long context handling. Try throwing in a 100,000 documents to a general purpose LLM and see what happens. All of the underlying skills and capabilities that are required to produce end-to-end legal work.
And the products and surfaces that live on top of it where lawyers are working every day and interacting with the system. Most of you are familiar with these. It's the surfaces in Legora where the work takes place. It's like tabular review, the research, Word integrations. This is enabling use cases beyond simple chat assistance. And lastly, this AOS is wrapped in enterprise-grade security and governance. This is the most important piece. Ethical walls, matter centricity, audit trails, permissions. This is the governing architecture that makes the platform safe and reliable for legal work in some of the world's most sensitive organizations.
But the system alone isn't what makes it work. It's the legal engineers I mentioned earlier. The people who deploy it, tailor it, and partner with you to bring it to life inside of your teams and organizations. That's the Legora AOS. Six layers, one connected system built for the depth and complexity of professional legal work. And at the heart of it, the engine that powers it all is the new Legora Agent.
The animators did a really good job with that one, but the engineering underneath it is even more impressive. The Legora agent is the most powerful AI ever built for legal work. The agent is remarkable on its own, but it's inside of the AOS that it becomes a legal powerhouse. It is a generational leap from assistive AI to agentic execution, redefining what a lawyer can achieve.
In the last 5 years, we have gone from rule-based software to simple prompting, advanced workflows, and now agents that can carry on complex work for hours at a time. The agent autonomously plans, executes, reviews, and completes legal work end-to-end with human oversight integrated throughout the system. Until now, every legal AI tool required the lawyer to drive the work forward. When you stopped, the work stopped. The agent changes that. It works with you and for you. Even when you're not there, while you're with clients, with your kids, playing golf with Ludvig Aberg, watching a great movie with Jude Law, or even while you're sleeping.
You set the objective. You approve the plan and it executes using any tool it needs inside of the AOS. And to make the agent your own, skills are the custom instructions that extend what the agent can do. You can customize exactly how the agent executes work across teams and practice areas. In other words, you can explain a process to the agent in plain English like you would to a new team member and it will execute work exactly to your standards. And turns out lawyers are the best prompt engineers in the world.
Skills allow legal teams to encode their knowledge and processes into the AOS. The agent proactively loads the right skill based on the task. And we offer a library of skills out of the box for common legal work. You can then take them and customize it to your style, processes, and practice areas. And our legal engineers are there to help every step of the way.
Now let's run through some use cases so you can see the agent in action, and this is a full live demo.
So here I've loaded up a matter. I have added a couple of files into this exciting project. We have an entire unstructured data room. Not the thing I wanted to see on a Friday afternoon. I'm going to ask Legora to reorganize this entire folder and every matter starts the same. So let's sort the documents into the subfolders by each custodian and let's fire it off. So without me giving any files to the Legora agent, it's able to find all of the files and the relevant context inside of this matter. And transaction teams know this. Data rooms with thousands of messy files, wrong folders, wrong names, trainees spending days manually renaming documents for a closing bible one by one by one. The prompt here asks to sort all of these documents into the relevant folder by custodian. And a custodian is basically a person or entity that a document belongs to, like the target company. Now, you can see the agent starting to read every document inside of this data room that I've given it access to. So, it's reading through and I haven't told it what the custodians are. It's figuring that out by itself.
So now we can see here that it started to read out all of them. It has found somebody that attendees wasn't clear in. And before it comes back to me with any finished work product, it is going to produce us with a plan. This plan is very important because when you're sending off these agents to do work at a longer period of time, you really want to make sure that you set the right intent with the model from the start up. So now Legora seems to have everything it needs and it's going to put together a full plan that we can review before it goes off to work. One of the behaviors that I started noticing in our engineering team is that when they leave their computers to go to lunch, they never close the top all the way down. They leave just a little bit of a gap because somebody's working under there.
Now, Legora has produced us with a plan for how it's going to go about this folder reorganization. Instead of just reading the short version, I want to see the full plan. So, inside of its own editor, it's created a plan for me here. Step one, create subfolders. Step two, move and rename. And then it's basically renaming and moving based on every custodian that it found. And it has its own to-do list of what it needs to do. I'm going to close the plan and I'm going to say approved. If we want to, it's that moment we can change and influence the plan before Legora goes off to work. And now I actually want to open this. Oh, there we go. Started to create one folder for each of the custodians. There's zero files in each of them, as you can see. And all the documents are lying very, very messy here. Oh, now six files in Vidal. There we go. Renamed very nicely. So, we know what we're looking at. Oh, there's something again. It's so refreshing just to watch it go to work.
But it's quite soothing. You can just sit there and maybe have a bit of a water. This is what it must be like to be a partner, right? I'm just joking. So what's happened here is this new agent, we've given it access to every tool or capability inside of the platform. So everything that you can do as a human user, now the agent can do. And turns out the agent is actually a lot better at navigating this UI and knowing what to do than a lot of people. It's a lot easier to onboard the agent than to onboard all of you guys.
So, here we go. It's found all of the folders. It's done with its thing. And it even flagged a duplicate document for me. And now we're done. This obviously extends to data rooms and with our data site integration to thousands and thousands and thousands of documents and different document types, etc. The cool thing about the agent is that it's really flexible and if it's unclear, it's going to ask you for clarifying questions. So, let's move on to the next task here. In this case, we have some intercompany agreements and I want help visualizing them in a diagram. So, one of the other cool things about the agent is that we can give it access to a lot of new tools that the previous assistant was too stupid to use. The old assistant had a really hard time understanding when it should use a certain tool to solve a certain task. But basically the new agent has just received this very, very long list of tools that it's allowed to use in its planning layer and you can add your own custom tools there. It's where skills come in and things like MCP servers and MCP servers is just a cool word on LinkedIn for APIs for LLMs.
So our agent here is using a pre-baked skill that we've given it access to which is to help it visualize data. So we're effectively asking it to do a tax transfer pricing here. And these intercompany agreements I'm asking Legora to show basically which company owns which. Haley is like yes this looks good. So different companies and who owns which and where is business flowing between these organizations and then I'm asking it for three things. I'm asking for a corporate structure. I'm asking for a diagram. I'm asking for a risk heat map that scores each of the agreements across pricing, documentation, structural and audit risk. And then I'm also asking for a radar chart across this. So there we go. Now it's created this pretty graph which is a little bit hard to see in the display but I promise the lines are there as you can see and it's showing me with flags because I asked for that. It helps me present things to my stupid CEO. Now it sees here like what is going where, right? And this is the cool thing about agents because the legal space when we arrived was so full of these point-based solutions that had, you know, motivated their existence by solving a very, very, very narrow problem. Turns out these agents can solve a lot of these narrow, narrow, narrow problems out of the box even. And when we build all of that together under a single roof inside of the AOS, that is when it comes to life. Risk heat map here fully interactive. And this is also something that I noticed. I only survived nine weeks at McKinsey, but they're really good at packaging their work, so they can charge a lot for it. And this is something I think a lot of you in the room can get better at. Receiving an Excel file or a boring Word document doesn't get me going. And so instead sharing something exciting like this in the portal. Now we're talking and here we also asked it to complete a radar chart.
So this is all exciting but listening to the previous panel there was a dream about being able to complete end-to-end due diligence. That has been so hard. We have worked on it I tell you for three years but I think we finally got it.
So here I've given Legora a bit of a longer prompt actually. And I've also given Legora access to all of the what do I have here? The RFI, the LOI, the due diligence review guidelines and the entire due diligence report template. And the reason I've loaded this into a video is because it takes 40 minutes to run and I didn't get 40 minutes. So this is a full legal due diligence report. We are asking it to complete a corporate, IP, commercial, employment, real estate, data protection and tax all in the same go. And as you can see, you have all the unstructured virtual data room files over here and we've fired it off. Okay, so Legora is going to start. It's going to start to go read all the documents. It's again going to create a plan. This is a bit of a longer plan because in this case it has to use sub-agents for many of the different phases here like employment, real estate, data protection, tax litigation, blah blah blah blah blah. It is going to launch a sub-agent because watching these things complete sequentially would mean that this would take six hours, right? So basically the main agent launches many sub-agents that goes off and does their little part of the work kind of in a law firm. And then when they're done, they're like, 'Hey, I'm done.' And they go back to the main agent and they receive that work product. So now we can see it's like spinning very quickly and it's because the video is sped up a little bit, but here the Legora agent has decided to create a tabular review for each of the different parts. So it's the corporate, it's the IP, it's the employment, it's going through them. But now this is the We did this video way too fast, I'm realizing. Here's the really important part. The agent is like unclear. And it's like, 'Oh my god, I found a bunch of issues. I need your help. I need your judgment. I am not fit to make these decisions for you. So, I am involving you.' That is so important here. It's like asking us what we want. And we've added a bunch of answers here. And I feel like this is also how you stay engaged because you can't just send off a task like this and then just expect the results to come back at the end of the sausage factory and be completely done yet. And here again, it's like a lot of different tasks now that it's going through. And I really do mean it when I said that this took 40 minutes to run, but it's created a due diligence report for us. So, I'm opening that in full mode. And here we have all of the edits into our existing template using our firm guidelines and language just as we wanted it. And it's gone through every single piece. Now, I don't get to take any credit for this, but our engineers do. So, I think that's worth a round of applause.
So when agents are this powerful, when working with this agent internally and many of our legal engineers here in the front row have been pushing this to the limit, we had to ask ourselves, how do you solve this problem of human and agent collaboration? Thank you so much. Today, legal teams use a variety of tools to keep track of tasks, checklists, chronologies. Some use Excel, some use Word. I've even seen a few of you use physical post-its. And none of these tools are right for the job because they inherently carry no structure. It's just stupid text. So, we built lists.
And you might say, what's so exciting about lists? But at Legora, we have list fever. Lists is how work gets organized, assigned, and executed inside of the agentic OS. It's the connective tissue between the agent, the lawyer, and other stakeholders. Every matter has structure, tasks, dependencies, handoffs, and deadlines. Lists make sure that this structure is visible and executable so that you can collaborate with fellow lawyers and the Legora agents. Closing checklists, due diligence trackers, case chronologies, in-house intake, anything that needs to be tracked, assigned or worked through, the list can handle it. Now you get to now you understand why we have fever. The magic is what happens when the agent gets to work together with the lists. So dropping in a contract, the agent will figure out the closing checklist for you. Assigning a task to the agent inside of the lists makes it start working on it immediately. And when it's done, the work product returns to the list for your review and it flags you.
And this is exciting to talk about, but it's even more exciting to show you.
So inside of Legora, we now have a new primitive called lists. I'm going to make a new list. And the first thing I get asked here is, do I want to create something from scratch or do I want to use a template? I'm going to use a template. So in this case, I want a closing checklist. The closing checklist then loads in and it's figured out what type of data I might want extracted on every point here. And each item effectively has a title, assignee, parties involved, etc. Now I need to add some context files. So in this case, I'm just going to give Legora a credit agreement. Attaching this file into this template loads it with a pre-written prompt about what I want to do. This is important because it means that the underlying software is not rigid. It's actually adaptable in terms of the instructions that you provide it. So now I've asked Legora to look through the closing checklist and populate this entire list for me. Now this takes a couple of minutes to run and like any good cooking show, I have the final dish ready for you.
What it does is it extracts all the different items. Each one of these items is turned into its own atomic unit, an object that can live and have data attached to it. So let's say we open this oh actually let me just do one thing here. So it's got the status, the assignee, start date, end date, we've got a reference and the parties. But maybe let's also ask it to add an attribute. So here I'm asking Legora to add an attribute, a new tag effectively called jurisdiction and populate it for every checklist item. So instead of me opening all of the documents and going oh what's that? What's that? What's that? Now I've just told Legora to go do it and it's going to figure it out on its own. So let's have that keep working. Can close that down and let's inspect some of these items. Let's open here corporate resolution and offer certificates. So we have a short description. It's on the right hand side here. We have a short description of the item and then we have a couple of different data points that Legora has extracted for us. And what it also does is it points to the oh there we go jurisdiction. Nice. It's cooking fast. It's also giving us the underlying source document which always points to the specific source citation. When we founded Legora, we had four priorities in terms of how we develop product. And they went, is it secure? Is it accurate? Is it fast? Is it cheap? And this is accurate and secure.
Oh, what was that cough? Let's also maybe add some point here. So if we want we can go and say let's add let's add some I have reviewed great please check right this is how you write and then I might attach a file here so let's add and attach the corporate resolutions great so that somebody can go review that and then instead of the status to sort of to be provided I'm going to say that this is under review and this has been assigned to my colleague Jules so in Stockholm they're currently busy building a notification center which means that Jules will get pinged and says hey this is now under review time for you to go do that perfect so have we gotten some action on the jurisdictions here it's still adding okay it's going to keep working on that now if we continue I want to show you one more thing so this isn't only about managing tasks and managing the status of things this is also a data structure that can be used to create a database that has more metadata and objects than just text. So what I did here was I uploaded all of the chronology files that I have. So basically I uploaded an entire case and I asked Legora to produce an entire chronology. And so what it does here is it just adds all of these items. It gives me a description of what exactly has happened. This is basically the fact intake if you will. And we can see here again that clutch allegedly said sort of it'll be fine. And we can then open that up in the witness statement and see exactly where that was.
I'll be damned. It'll be fine. So I'm really excited about lists. We have list fever at Legora and because they're so flexible, it means that we're going to be able to use them for a lot of different things. And I am certain that this is going to become a central matter in terms of how we track and keep both humans and agents on the same page about what work has been done. But the agent, it can do remarkable things. And with lists, I think that work is easily manageable. But without the right access to information and legal data, even the most powerful agent will be handicapped. And today's legal research wasn't built for the world that we're heading into. We're rebuilding it for the agentic era, for the way that legal work actually happens across borders, across languages, across systems. And two weeks ago, we took a major step towards this future. We acquired Kura, the agentic legal research company that has built, we met with all of them, the best technology for this kind of system. And their knowledge graphs and taxonomies connects millions of data sources across jurisdictions and languages, which basically gives the AI a map so that it can find the right things inside of the legal data. And you can expect to hear a lot more from us on this, especially our American friends.
Finally, I am excited to introduce you to true regulatory horizon scanning. There's a lot of people who claim they do this. We actually do it, and we're doing it in the form of monitors, which is powered by our acquisition of GraceView. As we announced yesterday, regulations change constantly, and most businesses miss important updates and have to face the consequences. The monitors change that. You simply set up a monitor, select the topics and jurisdictions that you care about, like UK employment law or financial services regulations, and anything that's covered by our taxonomy, the agent will track relevant changes in real time. And then it stitches together those changes with your firm's policies and documents to update them or run a gap assessment or compliance brief, all cited back to the original source.
And for in-house teams, real-time compliance monitoring at scale has previously been impossible. And for law firms, this will enable a managed service and a deeper, more scalable client relationship that was previously too cost prohibitive to deliver out of the box. We cover over a thousand, this is a 10,000 official sources, 100 jurisdictions updated hourly. And by partnering with many of the people in the room here today, we're adding new sources and capabilities to the system every single day. And monitors is just a great example of how the AOS is unlocking new possibilities that were previously wildly inefficient by bringing everything together. So let's look at a short demo live.
I'm going to head over to Legora and I'm going to make my first monitor. When I create a monitor, the first thing I have to do is I have to select the jurisdictions and the topics that I care for. In this case, I might decide to opt for EU and I want to get antitrust and competition. So now I've selected my ingredients and we can put it into the oven. Maybe I should start a cooking show.
Legora will now start to grab all of the sources from EU that has anything to do with the topics that we selected and inside of the preferences we can go very deep in terms of what data we want to keep. Now what we can also do on top of this data is work with the agent. So in this case, I'm going to ask Legora for a quick review of all the monitors that it finds and I want it to find the relevant articles and oh actually I have an idea. Let me redo this. I'm going to attach you can just ask it to do anything on top of the monitors itself, but it's more interesting if you actually give it some data as well. So let's add all of the policies that my company already has and then let's ask Legora to do a gap analysis on top of them. So let's see here we go. So now I wanted to look for gaps for the most meaningful regulatory or proposed changes that it finds. And for that specific gap I'm asking Legora to draft a short board briefing that includes a table detailing the changes. And then I just want a short response because otherwise it's going to go on forever.
The cool thing about all of these sources is they are also objects and they are not just a document. We are actually using an LLM to look at all of the data, create our own summaries and then we obviously source back to the original sources. The pipelines that the GraceView team has lied for the past three years. It's what's powering this system. But the data only becomes really useful when we can put the agent on top of it. And as you can understand, it's not far from us giving Legora access to all of this incoming data, all of the company policies, and then asking it to do this every single day on a scheduled task. And here, as you can see, Legora has actually identified a gap. It's citing back to the original source as it always does. Number two on the list and it's giving us some required policy changes that I can easily bring to my board. Awesome. But we're in the UK so we have to do something with the UK jurisdiction. Here I've asked Legora to find all of the financial services and I have selected all of the topics basically and this use case is tailored at a law firm. So a law firm watching UK sources relevant for their practice areas. And what I'm going to ask Legora to do now is I'm going to ask it to find an update and prepare a short financial services practice group update with everything that's happened over the past week. So I want a brief summary of the key regulatory developments, what changed, why it matters, what clients should do, and obviously if Legora has context about who your clients are, it will tailor this specifically for them. And then I also wanted to create a table of key dates to watch out for in deadlines. But again, because everything is tailored together, imagine this could create deadlines and obligations inside of a list that's shared with your client in the portal proactively every single day as something new comes out. We are building all of these different pieces and then it's up to you and our legal engineers to build it all together. And this is why we're so excited about the future. So now Legora is giving me an update here and I know that there's a really good financial lawyer somewhere here in the audience. So they can stand up or scream no if this is completely irrelevant.
Yes or no? Yes. Fantastic. That's it for monitors.
Starting today, every product and feature that you have seen is starting to roll out to our customers. And I'm so excited for all of you to tell us what you think. Three years ago, we all started the same race. Learn about AI, adopt it, move faster, stay competitive. That race is over. It was not enough. The next race is about who defines what a legal team can be in the era of agentic law. Who reinvents themselves first? Who uses AI in a way no competitor can replicate? And who delivers outcomes that makes every other legal team feel like a decade behind. The individuals and organizations that lead this movement will be the ones who reshape legal work for the next generations.
Legora is an AI company, but I like to think that we're not just a vendor. We also throw great parties. I'm joking. We are the partner that walks every step of the way together with you into this wild uncharted world. And we did not build the AOS just to sell software. We built it because we believe that AI and law is one of the most important and most knowledge-intensive professions in the world. So while others are building better, faster assistants, Legora is enabling a new kind of lawyer and a new kind of legal team. We believe that this is the most exciting time in history to be a lawyer and the Legora team is extremely proud to be your partner in this defining moment in time. Thank you.
