Most conversations about legal technology start from the same premise: law firms exist, and technology is introduced into them.
The question is usually how well those tools are adopted, how much efficiency they create, and whether they meaningfully change how lawyers work. However, a smaller, but increasingly relevant, question sits just behind it: what happens when a firm is built around the technology from the outset?
That is the perspective Conrad Everhard brings. As a founding partner of Flatiron Law Group LLP, he is not approaching legal tech as a buyer or evaluator of tools, but as someone designing a law firm where technology is part of the operating model itself.
We spoke with Conrad to understand how that thinking translates into practice, and what it suggests about the direction of the legal services market.
Looking at legal tech from the inside out
Flatiron is, on paper, a law firm. It focuses on high-stakes M&A and complex transactional work, led by partners with backgrounds in large firms. But the way Flatiron is described leads you to think less about what it actually does, and more about how it is structured to do it.
Rather than selecting tools to support existing workflows, Flatiron’s systems sit at the centre of how work is executed. Flatiron built an AI infused deal operating system called Deal Driver which manages all of the firm’s transactions from end to end (Deal Driver is used by Flatiron internally but may be separately commercialized in the future). The founders of Flatiron are also founders of a separate company called Deal Mentor, which has a built an AI training platform that simulates live negotiations in complex transactions and can be deployed for any learning purpose. Deal Driver and Deal Mentor provide the foundational architecture for Flatiron’s practice model. They are not layered onto practice but embedded within it.
What that design choice reflects is that if technology is treated as infrastructure rather than an add-on, the starting point is no longer “how do we improve the current model?” but “what would the model look like if built differently?”
This week, we spoke with Conrad to delve deeper into how that model operates in practice and what it means for the relationship between AI, legal technology, and the firms being built around them
TLW: You’re building from a different starting point than most firms. How much of Flatiron’s model is a response to the limitations you’ve seen in traditional firms trying to layer technology onto existing structures?
Conrad: “Almost entirely. When you spend decades inside Big Law, you see clearly why technology never takes root the way it should. The billable hour model rewards inefficiency. The partnership structure resists change. And the cultural incentives are all pointed in the wrong direction. You can layer any tool you want onto that foundation, and the institution will find a way to absorb it without changing. We didn’t want to improve the old model — we wanted to build a new one. Deal Driver is not a productivity tool bolted onto a traditional deal team. It is the deal team’s operating system. That’s a fundamental difference.
We concluded that, like in other industries, radical business change has to be driven from “outside the firewall”, from start-ups and innovative firms, which are free from the cultural baggage of Big Law. That is why we founded Flatiron. We are free to innovate, reimagine and experiment without constraint.”
Less focus on tools, more on sequencing
One of the more consistent themes in our discussion was sequencing.
Flatiron’s “HI first, AI second” approach is not framed as a marketing distinction, but as a practical one. Human judgment defines the parameters of the work, and technology operates within those boundaries.
This is an intentional departure from the way many tools are positioned. The question is not how much work can be automated, but where automation sits within the process.
For Conrad, that sequencing appears to be a question of control, to ensure that technology amplifies decision-making rather than displaces it prematurely.
TLW: There’s a lot of emphasis in the market on autonomy and automation. How do you decide where AI should take over versus where human input remains essential?
Conrad: “AI should not ‘take over’ anything — that framing is the wrong one. What we’ve built is a structure where human intelligence sets the direction, defines the parameters, and makes the judgment calls, and then AI handles the extraction, organisation, and acceleration of the work within those boundaries. Our lawyers are not reviewers of AI output — they are the architects of how AI operates on each deal. That sequencing is not a constraint; it is the design. In a complex M&A transaction, the stakes are too high for any other approach.
Look, we compete with Big Law at the top of the food chain for high margin, complex deal work. In that rarified air, you need elite talent to be competitive. You can’t compete with Big Law without it. Now, under our AI/tech leveraged/infused model, we need less labor than Big Law does, and we empower our talent to do more, better and faster than Big Law, with less inputs and less overhead. But we will always be HI first.
There are other so-called ”AI native firms” that purport to be AI first. As far as I am concerned, those firms are more akin to legal apps, automating select, repeatable tasks like NDAs and commodity licenses. That’s a fine model. But it’s not our model. That model doesn’t work at the top of the food chain.”
The firm as the product
A recurring pattern in legal tech is the separation between those who build tools and those who use them.
Flatiron collapses that distinction.
The systems that underpin its work are built internally, for its own use, and continuously adapted as part of delivering live transactions. That creates a different feedback loop: the product evolves through practice, rather than being deployed into it.
It also raises a broader question about where innovation in legal services will come from. If the most effective tools are those built closest to the work itself, firms like Flatiron may be operating with fewer of the constraints that shape traditional adoption cycles.
TLW: Do you think meaningful innovation in legal tech is more likely to come from firms building internally, rather than from external vendors? And if so, why?
Conrad: “I think the answer is both, but in different parts of the market. External vendors build for breadth — they solve problems that exist across many firms. That is valuable, but it produces general-purpose tools that no single firm has a structural advantage in using. What you cannot buy off the shelf is the deep, deal-specific intelligence that comes from building a platform inside a live transactional practice. Deal Driver was built deal by deal, transaction by transaction. Every workflow reflects a real problem we encountered and solved. The firms that will move fastest are the ones that stop treating technology as a procurement decision and start treating it as a core competency.”
Training, but in a different context
Another area where this model diverges is training.
Flatiron’s Deal Mentor platform suggests a view that legal training should evolve alongside the systems lawyers use. If workflows are increasingly structured and supported by technology, training may need to reflect that environment rather than sit apart from it.
This is not a widely explored area in legal tech, but it speaks to a broader point: if the way work is done changes, the way lawyers learn that work may need to change as well.
TLW: You’ve been in this space for many years now. How does working within a technology-driven environment change how lawyers develop judgment and expertise over time?
Conrad: “The concern you hear most often is that if AI does the heavy lifting on diligence and document review, younger lawyers miss the repetitions they need to build pattern recognition. I think that concern is legitimate if you let AI displace learning. But that’s not what we do. When a lawyer at Flatiron works with Deal Mentor, they are engaging in simulated negotiations against AI personas built from decades of real deal experience. The feedback loop is immediate and realistic. You can accelerate the development of judgment in this environment — but only if the training is built around the work, not apart from it.
Fundamentally, you need to train lawyers to practice differently. We train our lawyers to resist the impulse driven into their psyches by Big Law to maximize inefficiency to produce more hours. Precisely for that reason, we don’t bill by the hour. We bill flat fees by the project. Our objective is to train our lawyers to leverage the full suite of productivity tools available to them (Deal Driver, Deal Mentor and others) to achieve the same or even higher levels of quality than Big Law but faster and more efficiently. It’s a completely new mindset and requires retraining and rethinking everything about the way that we practice.”
A different kind of competitive pressure
What makes this perspective relevant is not that Flatiron is unique, but that it may not remain so.
If firms can be built around technology rather than gradually adapting to it, the source of competitive advantage shifts. It is no longer limited to efficiency gains within an existing model but extends to how the model itself is designed.
This may also explain why some of the friction seen in traditional firms, around adoption, integration, and pricing, is less visible in firms built from scratch. There is no legacy system to work around, and fewer structural constraints on how work is delivered.
What the future holds
Much of the legal tech conversation centres on tools: what they can do, how they are used, and whether they deliver measurable returns. A more structural question is also emerging: if technology is capable of reshaping how legal work is executed, it may also reshape the organisations built to deliver that work.
Conrad’s perspective offers one view of that shift, not as a theory, but as something being tested in practice.
It is still early. But it suggests that the next phase of legal tech may be defined less by adoption within existing firms, and more by the emergence of firms built with different assumptions from the start.
