- Build Club
- Posts
- Build Club Spotlight #1: 👩🏻‍⚖️ Mary Technology with Co-founder Daniel Lord-Doyle
Build Club Spotlight #1: 👩🏻‍⚖️ Mary Technology with Co-founder Daniel Lord-Doyle
Building AI for law firms - automating chronology creation
Daniel Malkinson
April 22, 2024
Welcome to the Build Club Spotlight Series, where we shine a light on some of the most promising Aussie AI startups!
This week, we have the privilege of speaking with Daniel Lord-Doyle, a co-founder behind the innovative Mary Technology.
Get ready to dive into Mary Technology and Daniel's perspective on the challenges, triumphs, and aspirations propelling Mary Technology's quest for AI disruption in the legal tech arena. ⚡️
Talking points
Let’s dive in!
What is your background and your story founding Mary Technology?
Hi I’m Daniel Lord-Doyle, 👋
I am a 2x founder, working over nine years in startups. I've led sales, marketing, and engineering efforts during my journey, scaling companies to generate $4.8M in ARR.
I started Mary with my co-founders Rowan McNamee (a lawyer turned entrepreneur and Techstars alumni) and Harry Raworth (ex-Virtually Human Studios, backed by a16z and TCG for $25M+).
Rowan's insights from his legal background were the catalyst for founding Mary.
He experienced a critical problem in the legal industry. Lawyers today are overwhelmed with admin and manual review of huge volumes of documents.
This often leaves little time for strategic legal work, while also inflating the cost of legal matters for both the clients and the law firms.
As someone who had personally spent tens of thousands of dollars on immigration lawyers, I deeply empathised with this pain point. This experience, combined with Rowan's industry expertise, inspired us to create a solution that can make legal firms more efficient and enable them to offer more transparent pricing to their clients.
We are team of lawyers and AI engineers, helping law firms to safely and accurately automate the process of chronology creation
How would you describe your product offerings to a new lawyer and to an AI engineer?
To a lawyer: We understand the core challenge you face - creating an accurate and comprehensive chronology, which can take days, if not weeks, and costs your firm hundreds of thousands of dollars a year in time. Our solution reduces the time spent on this task by 90-95%, allowing you to focus on the strategic legal work that truly matters.
We extract the key events, people, and dates from the thousands of documents involved in a legal matter, and present them in an interactive, collaborative chronology. Our purpose-built user interface is designed to minimise the time-to-confidence - the time it takes for you to feel 100% certain that the chronology is accurate and complete.
The chronology is a living document, enabling you to seamlessly add new information, edit, and remove event lines. You can review the source of each event on hover and access the original text. It's also version-controlled, allowing you to share and collaborate with your team.
In an upcoming feature, you'll be able to ask your chronology questions, such as "How many times did John go to the doctor?", and receive answers alongside the relevant source documents. Importantly, we deliver this solution with no implementation time or technical team required, and the output will be 100% conformant with your house style.
To an AI engineer: At our core, we leverage a combination of OCR, traditional machine learning, and large language model (LLM) techniques to extract valuable insights from large document sets. We efficiently store this data and push it through multiple processing layers, applying various LLM design patterns to uncover the most relevant information.
We expose this information through a purpose-built user interface, where lawyers can view, filter, and interact with the data. The UI is designed to help lawyers quickly become confident in the accuracy and completeness of the produced chronology, enabling them to make strategic, informed decisions about their legal process.
Building a world-class LLM operations team is a key focus for us. We're developing best-in-class evaluation pipelines that continuously monitor the quality of our outputs, alerting us to any changes that may improve or reduce the accuracy and reliability of the data we provide. We also leverage these pipelines in real-time to ensure the information delivered to users is consistently accurate and complete, regenerating it as needed.
We believe we're tackling a perfect use case for AI, as the written laws and legal documents that form the basis of our legal system can be considered a codebase for the English language. By working with this highly structured, grounded data, we can explore and address critical AI challenges, such as hallucination and accuracy.
Importantly, the economics of the legal industry make this an ideal problem to solve using AI. The current manual processes are extremely expensive, allowing us to thoroughly explore and develop the most effective solutions.
How has the journey been so far and how has Mary Technology evolved ?
It's been a challenging, but rewarding journey so far. Our team has shown remarkable resilience, working 14-hour days, 7 days a week, for 10 straight months to get us to where we are today. Despite the long hours and occasional setbacks, we've made significant progress and are starting to see the light at the end of the tunnel.
We spent the first 6-7 months deeply immersing ourselves in the problem, conducting over 100s of hours of interviews with lawyers from SME firms. We even spent 4 months embedded in a mid-sized law firm, speaking to lawyers every day to truly understand their pain points. This intensive research phase allowed us to pivot to our current product and audience, which has been incredibly well-received.
Over the past 3 months, we've built and launched our SaaS product, onboarding 8 sophisticated legal firms as alpha users. The feedback has been incredibly positive, with comments like "People here are very interested and excited by the tool!", "Is there anything more we need to do - or can we just pay and go?" and "I just tried it - and it is very impressive!"
In 6 weeks, we're releasing our Beta version, which will take our product to the next level for our users. At that point, we're going to aim at scaling to 50 users here in Australia and 100 users in the US as a major milestone.
What advice would you give to aspiring AI builders?
I’d like to give pragmatic advice, so I would say:
Think big and disruptive, but focus on a small subset of a large market first. I think Peter Thiel and Sam Altman’s advice on this is hyper relevant for startups with limited resources: focus on micro markets, create a dominant position, and use this momentum to whip out to pursue a broader opportunity.
Velocity, velocity, velocity. Build something and get it into your users hands as soon as possible. Perfection is the enemy of progress.
Be delighted when you are wrong. It’s ok to not get things right, but try to be humble about your hypothesis. Search out real value for your users, even if it isn’t where you thought it was.
Be hyper user-focused. This is the way.
Join the Build Club. If you are ideating, tinkering or working full time on your project, there is a place for you. Build Club is a startup hub with over 600 builders, with chapters all across Australia where you can find a community of other AI startups in the arena. Go to www.buildclub.ai and apply now: the Slack is a great place to ask questions and get great advice.
I would just say that this isn’t an easy path, and “If you're going to try, go all the way.” Read Charles Bukowski’s poem.
How can our audience help you build and grow Mary Technology?
If you are a lawyer who wants this, contact me and I’ll get you in the alpha. If you are a startup founder or builder, please give me advice. If you are an investor, let’s chat - we’re about to fly!
You can stay up to date with Mary Technology on their LinkedIn :
Or follow Daniel’s LinkedIn here
Liked this post and want to be featured next? Reach out to [email protected]