As part of the news about the official launch of Twin1 to build digital twins for lawyers, which has just raised $20m from several major investors, Artificial Lawyer put some questions to Lewis Liu, the CEO and co-founder of the new startup, who also previously led Eigen. (Here’s the AL news story about the raise.) We look at the launch, the ‘why’ and ‘how’, and also explore the future shape of the legal tech market, as well as token costs.
What does this mean on a personal level?
In a sense, this is unfinished business for me. At Eigen, we pioneered enterprise AI before the current boom. We served half of the world’s largest banks and 20% of the AmLaw 100, digitized more than $100 trillion in contracts, and became the first AI company approved by the Federal Reserve and FDIC to process financial contracts without human intervention.
But we were early. Like most enterprise AI companies founded before ChatGPT, we did not capture the full upside of the AI revolution.
At Twin1 we are applying everything we learned. We have reunited some of Eigen’s best people, three of Twin1’s co-founders came from Eigen (including myself), and many of Eigen’s key investors have backed us again. More importantly, we bring hard-earned knowledge of how law firms and financial institutions actually operate, how they buy technology, and what it takes to meet their governance, security, and regulatory requirements.
After a year in stealth its great to be out open riding again. This time, we have the unfair advantage of having done it before.
— Advertisement —
Legal Innovators New York – Nov 17 and 18 – The landmark legal innovation conference!
—
How did you get to this concept from Eigen?
Twin1 really grew out of an unsolved problem I encountered at Eigen. Eigen was very good at extracting structured facts from complex documents, but we learned that the document is often only the final outcome. The most valuable knowledge is the context behind it: why a decision was made, how a negotiation unfolded and who had the relevant experience.
A project for at top 10 global law firm brought that home. They wanted to digitize the stock purchase agreements (SPAs) their M&A lawyers had negotiated. We successfully created a structured database, so an associate could see, for example, how often a buyer conceded a particular term. But that only told them what happened. What they really wanted to know was why the buyer conceded, under what circumstances and what negotiation path produced that outcome. That knowledge was scattered across partners’ and associates’ emails, calls, notes and memories.
That was the insight behind Twin1: in a professional organization, the atomic unit of knowledge isn’t a document, it’s a person. We built AI Twins to make each person’s knowledge and judgment accessible, then networked those Twins so expertise can move across an organization without giving everyone unrestricted access to private data. Eigen structured the documents; Twin1 captures and connects the human context behind them.
What is the growth plan?
Our growth strategy has two complementary parts: enterprise first, followed by self-service.
We are initially focused on regulated enterprises, particularly legal and financial services, where knowledge is highly valuable but fragmented across people and systems, and where governance and security are essential. This is our core strength. We have spent a decade learning how leading law firms and financial institutions operate, buy technology and govern sensitive information.
That enterprise strategy is already showing strong traction. We have a pipeline across more than 400 prospects, including major law firms, banks and other regulated institutions. We are already working with firms such as Linklaters, Orrick and Dechert, as well as financial institutions including Customers Bank. Our immediate priority is to deepen those deployments, expand within existing customers and turn that early engagement into repeatable enterprise growth.
But this cannot remain a product only for the largest institutions. The same knowledge problem exists at regional and specialist law firms, private wealth managers and community banks. Our next step is a lower-cost, self-service offering that allows those organizations and, ultimately, individual professionals to create and use their own Twins without a long enterprise sales process.
We have also deliberately built distribution into our investor base. In legal, that includes Antiportfolio Ventures, founded by former Kirkland & Ellis managing partner David Fox, alongside strategic backing from leading law firms such as Orrick. In financial services, our investors include Fidelity (F-Prime, their strategic venture arm), EJF Ventures and BTech Consortium, whose networks reach financial institutions. In energy, Aramco Ventures is a core investor. These investors give us more than capital: they provide market access, industry insight and direct input into product development.
So the plan is to win the top of the market through enterprise deployments, use that credibility and learning to expand across regulated industries, and then open the platform through self-service. Our long-term belief is simple: every lawyer and business professional should have a Twin that amplifies their knowledge and judgment.
How do you see the legal AI market evolving?
I think the legal AI market is moving from automating individual workflows to capturing and compounding the knowledge of the firm itself.
Kirkland & Ellis’s recently announced plan to spend $500 million building its own AI technology is a powerful signal. The stated ambition is to put the ‘collective intelligence’ of its lawyers into a technology platform. But even among the AmLaw 100 and Magic Circle, very few firms can invest that kind of capital. Twin1 makes that capability accessible more broadly, helping firms preserve and compound the knowledge of lawyers across the organization, including partners approaching retirement.
At the same time, workflow-focused platforms for legal research, document review, drafting and redlining will continue to make enormous progress. We do not see Twin1 as competing with that entire ecosystem. We want to complement and enrich it. Our enterprise MCP server gives existing AI agents and legal tools secure, governed access to the context held by an individual Twin or the wider Twin Network. That means those systems can draw on deeper client knowledge, institutional history and, most importantly, the know-how and judgment of the individual lawyer rather than producing generic answers.
The strategic concern I hear from law firms is that AI will continue to raise the floor and commoditize more routine legal work. As that happens, what becomes even more valuable is what remains distinctive: the experience, relationships, judgment and private knowledge held by individual partners and associates.
Twin1 helps firms stay ahead of that commoditization by identifying and amplifying each lawyer’s unique strengths, while preserving privacy, permissions, compliance and human agency. Our goal is not to flatten people’s expertise into generic output, but to help that expertise compound safely across the firm.
So I see tremendous progress ahead for today’s legal AI platforms. Twin1 can make those systems more context-rich and effective, but the bigger opportunity is to make the human more effective too. The future of legal AI should not just automate workflows. It should amplify the lawyers whose knowledge and judgment create the firm’s real differentiation.
How are you coping with token costs?
We designed Twin1 from the outset to be model-agnostic. We can route work across leading proprietary models and, as the market develops, incorporate open-source alternatives rather than being locked into a single supplier. The model is a component of our architecture, not the product or the moat. Our durable value lies in the governed context layer: the institutional knowledge, permissions and interaction history that remain regardless of which model sits underneath.
We have also been disciplined about unit economics from day one, so we would retain healthy margins even if token costs tripled. Over time, competition among model providers, better routing and the growing quality of open-source models should reduce costs further.
More fundamentally, we believe the application layer should remain independent from the foundation-model layer. If the same company controls both, it has less incentive to use models efficiently or pass savings on to customers. A competitive, modular ecosystem creates the right discipline and helps keep prices down for end users.
But this is about much more than cost. It is part of the broader question of AI sovereignty. We are hearing growing concern from legal and financial-services clients about where their data goes, which models process it and whether the frontier labs could ultimately use professional expertise to commoditize the very firms supplying that knowledge. Some clients are already asking us to avoid particular model providers or consider open-source alternatives.
We do not currently deploy open-source models for clients, but our architecture gives us the ability to move quickly when clients require it. We also offer deployment models in which data and processing can remain within a client-controlled environment, and customer data is not used to train our models or third-party models.
So our answer to token costs is technical discipline, supplier flexibility and healthy unit economics. But those same architectural choices also give clients something increasingly important: control over their data, their model supply chain and the expertise that differentiates their business.
Thanks and congratulations on the launch!
—
Two Major Legal Innovators Conferences this November
Come and join us in New York and London this November at Legal Innovators!
Legal Innovators UK – London, Nov 4 and 5
And, then Legal Innovators New York – Nov 17 and 18.
After another fantastic Legal Innovators California, where we had speakers from OpenAI, Y Combinator, Google, Meta, and many more pioneering organisations; and our stellar inaugural event in Paris this June, we are now looking forward to the landmark conferences in London and New York, both in November, and both across two days: Law Firm Day, and Inhouse Day.
Discover more from Artificial Lawyer
Subscribe to get the latest posts sent to your email.