Deploying legal AI assistants in health tech.

An interview with Zeno Capucci, GC @ Docplanner

Charlotte Kufus: Let me introduce our first speaker. Zeno Capucci is General Counsel at DocPlanner Group. Thank you for joining us. Could you introduce the company you work for? 

Zeno Capucci: Absolutely, my pleasure. And thank you for having me. DocPlanner is a digital platform for healthcare providers and patients. The best way to think about the basic idea is a booking platform for the digital healthcare sector: we have marketplaces that allow patients to book consultations with doctors, have online consultations with doctors, and exchange information with doctors, clinics, and hospitals. On the flipside, we run a SaaS platform that enables doctors, clinics, and hospitals to digitize the management of their practice: what we call a practice management software. We operate in 13 countries, and there’s almost 3000 of us employees and staff members. I’ve been at DocPlanner since January 2020, so it’s been four years and quite a journey. The beginning of the legal function was just me and another couple of brave individuals. Now it’s almost 25 lawyers and counting, as we’re aiming to hire more this year. 

Our mantra is growth, growth, and growth, launching new products, doing it fast, and in as many markets as possible. Keeping up with that pace is always a challenge for the legal team, because the ultimate role of a lawyer is to make sure that things happen while minimising risk. I would argue that minimising risk while not slowing down the business needs to be the priority (...)

CK: Can you tell me and the audience a bit about the challenges you face as a legal team?

ZC: They’re probably challenges that are common to everyone else. It’s repetitive tasks. It’s having to answer the same questions over and over again. It’s not having a proper way to organise the workload when you’re supporting teams from different backgrounds, different countries, with different goals, in a very high-growth organisation. I forgot to mention that we’re a venture capital-backed company, so our mantra is growth, growth, and growth, launching new products, doing it fast, and in as many markets as possible. Keeping up with that pace is always a challenge for the legal team, because the ultimate role of a lawyer is to make sure that things happen while minimising risk. I would argue that minimising risk while not slowing down the business needs to be the priority for a legal function in any corporation, especially a corporation that needs to grow fast in order to be appealing to investors and disruptive to the sector. Creating templates from scratch, finding information in policies, and making sure that employees have access to these policies––these are the typical challenges we face on a day-to-day basis as a team. 

CK: You were one of the first legal departments to implement a gen AI tool. What potential did you see, and why didn’t you wait any longer? 

ZC: It’s interesting, from reading posts on LinkedIn it sounded like everyone was doing it; we actually felt a little behind the curve! Basically, we are a tech company. Our management team is made up of many brilliant software developers and very tech-savvy folks, who are very enthusiastic about new technology, including gen AI. As soon as the gen AI hype started, in October 2022, with ChatGPT, the company ran a hack-AI-thon. This was a hackathon focused on AI solutions. And we started working on the most promising ideas to turn them into features that would better serve our users, customers, and colleagues. So, we felt like we needed to quickly catch up with the rest of the business.

If you think about it, how can you tell yourself you’re properly serving a business that, as a mission, provides best-in-class service and support to external customers and clients, when you’re not providing that level of same service to your own internal clients and stakeholders? That was the first motivator. We wanted to be at least as good as the rest of the company. Secondly, after playing a bit with ChatGPT, we realised there were some pitfalls or challenges slowing us down that could potentially be solved with the use of AI. We didn’t look at AI as the solution to all problems. We had specific problems we thought AI might help us solve. And it was a very nice product-market fit for us; what we were already looking for (a way to reduce our workload, eliminating as much as possible all repetitive, low value-add tasks from our daily to-do lists) became available on the market as a product, Legal OS. 

CK: What has your experience been so far with this technology?

The problem we were trying to solve was mainly tons of repetitive questions being thrown at us, the lawyers. Most of these questions could be answered by looking at our internal documents, policies, and guidelines. No one does it, and it's no surprise.

ZC: It’s already had a massive impact. When we started scoping solutions on the market, we came across many players, including Legal OS. The problem we were trying to solve was mainly tons of repetitive questions being thrown at us. Most of these questions could be answered by looking at our internal documents, policies, and guidelines. No one does it, and it's no surprise. Any legal professional in an organisation deluding themselves with the idea that employees will spend their days reading policies is probably in for a harsh surprise! We realised rather than having lawyers spend their valuable time copying & pasting or rephrasing answers they’d already given elsewhere (or that are already set out clearly in our internal documents), rather than having people wait for these answers (because a lawyer is not available 24/7, because we have other, arguably more important things to do..), we realised rather than that, we needed a smart solution, so there was margin to look for a provider.

And we came across Legal OS, who were solving this problem very efficiently, because their product integrates with our main communication channel with the rest of the business, Slack. This allowed us to say, "We don’t need to ask the business to use a different channel and ask the same questions in a different, more convoluted way, creating friction." It allowed us to say, "You business - stakeholders, salespeople, procurement people - do the same thing you’ve always done - asking us the questions on Slack. And there’s going to be someone, or something - an AI chatbot - answering on our behalf." And the impact has been massive. Since we rolled out for the whole company in October 2023, after a few weeks of piloting, in one quarter we had almost 1000 questions answered by the chatbot. We call the chatbot our “par-AI-legal”, a gen AI paralegal that answers questions on our behalf. The name is a bit cringe, I know. 

The speed at which this solution went from 0 to 100 was unbelievable. It replaced us on four or five different topics in four or five different languages. The speed at which we were able to roll this out, in a matter of weeks or months, was incredible.

CK: I think it’s great. Based on your experience, what is AI good at... and what can it not do? 

ZC: For us, what works really well is automating stuff that doesn’t require the brilliant, creative thinking of a lawyer, and making it seamless for the business to obtain answers without the lawyer having to get involved. The speed at which this solution went from 0 to 100 was unbelievable. It replaced us on four or five different topics in four or five different languages. The speed at which we were able to roll this out in a matter of weeks or months was incredible. It’s very good at taking a specific problem that requires a process-driven solution and solving it very fast, with very little input compared to other solutions we’ve tried. Anyone considering automation because certain processes are too cumbersome and require too much input from lawyers, I think gen AI can help you. What it’s not good at, at least so far, from what we’ve seen, is creative thinking; discussing with product, sales, and management, solutions to problems that are multifaceted and require some navigation of gray areas, or blue-sky gazing. But that’s what we enjoy doing the most, and arguably what we are paid for (and go through years of legal education for)... not copying & pasting from policies. So, in a way, I’m relieved that we can continue doing that part of the job. 

CK: Very true. What are some other challenges you’re facing, especially in regards to the repetitive, time-consuming, low-value work that you think gen AI should and will be able to help you with? 

ZC: There are tons of examples. As I said, we call the chatbots par-AI-legals. If you think about what a junior paralegal or legal secretary is able to do--obviously not the more technical, but the more admin parts of their jobs--I think all of that is ripe for disruption. From creating slides and presentations, first drafts of documents, and internal policies and contracts; to archiving communications, saving documents in the right folders, organising your calendars and agenda… There are so many little daily tasks we’d love to have someone running for us… like life admin, but in the professional sphere! That would be fantastic. Even finding information in a more efficient way--precedents, case law, communications we’ve already sent--and summarising it, creating a first draft of emails, policies, and documents for us to review. There’s so much that we hope gen AI will be able to do, so that we can hopefully focus our energy on more entertaining, challenging stuff… or simply lie on a beach in the sun and let gen AI do the work for us, for free - although I doubt that’s ever going to happen, unfortunately! 

CK: Our time is almost over, but I have one more question... What advice would you give to other teams who are considering implementing AI? 

ZC: As with anything, start with an MVP, a minimum viable product. Then test the pilot. Don’t start from the big picture, the journey, or the roadmap. Just start from a problem you have, experiment with a potential solution, and see if gen AI can help you. If it can, or if you see potential - if you see that it could work, subject to tweaks, or by working with a provider, or by putting in a little more effort - then continue. Otherwise abandon it. But definitely give it a try, because there is potential there. And sometimes it’s not even “just” potential: the solution may already be there and ready. Secondly, don’t underestimate the effort you’ll need to put in at the beginning, but don’t overestimate it either. Compared to other legal tech solutions I’ve tried, gen AI really learns, and the more it learns, the faster it learns. So the first week may be a bit more painful, or just as painful as implementing any other legal tech solution. But week four is going to be like week 15 or week 35 for a non-AI-driven product, at least in our experience. Don’t get discouraged on week one, because progress will come very quickly. 

The marginal impact is several times bigger, because we’re removing the lawyer from the equation, or at least allowing the lawyer to only get involved when they decide to.

CK: Thank you so much. We have another question from the audience: Why did you decide to deploy an AI solution that is directly used by your end users - DocPlanner’s employees and stakeholders - rather than by your legal team? 

ZC: It’s a good question. The marginal impact is several times bigger, because we’re removing the lawyer from the equation, or at least allowing the lawyer to only get involved when they decide to. A copilot for lawyers would make a lawyer’s life simpler, but doesn’t necessarily remove them from the equation. If you give a genAI solution to a lawyer, it requires the lawyer’s input every time it gets deployed. It may reduce the lawyer’s involvement by 30, 50, or 70%. But not by 100%. And on the users’ side, they’d still need to wait for the lawyer to wake up in the country they are based in (if in a different time zone), or to finish the previous task (or their to-do list) and jump on the question. Enabling the business to self-serve as much as possible, to the extent your strategy allows it, and that you can control/minimise/accept the level of risk (we are talking about low-risk activities, here), should be the end goal for most legal functions. Otherwise you will have to continue adding expensive legal headcount, for no good reason. Your CFO is not going to be happy, as the organisation may become bloated, costly, and inefficient. 

CK: One last question from the audience. What, in leveraging gen AI, is the next big thing for you? 

ZC: For us, it’s got to be document generation or document review automation. We’ve already reduced the burden on the legal team to fully review every single contract that comes in, using policies and defining our risk appetite; but the aim should be to again empower sales, procurement teams, and “buyers” within the organisation to self-serve as much as possible. Only escalate to the lawyer when it’s really needed. The lawyer should just be the one” bringing it home”... and obviously taking credit for the result! The heavy lifting should be done by gen AI. That’s one thing. Another would be, in my dream, a gen AI assistant that takes care of life and professional admin, so that I don’t have to save files in the right folders anymore, rearrange my calendars, and so on. And I can just feel cool about being a lawyer dealing with important stuff. :)