In last Friday’s session, Vadym Kuzmenko, Legal Engineer at Legora, one of the top legal tech companies in the world, unpacked what it really takes to build and implement legal AI at speed – from the foundational models to the day-to-day reality inside law firms.

From Law graduate to Legal AI Engineer

Kuzmenko trained as a lawyer in Germany before realising he was more energised by fixing inefficient processes than by traditional legal practice. A decade in legal tech has taken him from start-ups to one of Germany’s largest law firms and now to Legora, where he helps hundreds of firms and in-house teams implement AI in practice.

The move from law firm to legal tech vendor was a natural one: the work is similar – building AI systems, running workshops and educating teams on what large language models (LLMs) can and cannot do – but now carried out at global scale rather than for a single firm.

Speed at three layers: Models, Applications, Law Firms

For Kuzmenko, “speed” in legal AI plays out across three distinct layers:

1. Foundational models
   At the base level, providers such as OpenAI, Google and Anthropic are locked in a high-velocity race. New models and capabilities appear so quickly that even specialists struggle to keep up. This layer is not especially “sticky”: most customers can switch providers, so healthy competition is inevitable and desirable rather than a winner-takes-all outcome.
2. The application layer
   The furious pace at model level cascades into the tools built on top of them. Kuzmenko notes that every couple of months, Legora feels like a different company in terms of product capabilities. Over the last year, the team has grown from around 40–50 people to more than 250, and the customer base from roughly 250 to close to 600, driving exponential demand for implementation and support.
3. The law firm and in-house layer
   The next wave of speed will hit the organisations themselves. Until now, many firms have watched developments “with popcorn in hand”, but Kuzmenko believes that comfort is temporary. As AI becomes embedded in how legal services are delivered, firms and legal teams that cannot match the pace of adoption will simply be outpaced by those that can.

Why consultative support matters

The conversation returned repeatedly to the role of expert support. AI-specific consulting is no longer a nice-to-have: it is often the only way busy partners and GCs can cut through the noise.

Kuzmenko sees three core elements:

• Cutting through hype – distilling market trends, separating short-lived experiments from durable shifts, and sharing pattern-recognition from hundreds of implementations.
• Change management as a real discipline – not just a buzzword, but a structured approach requiring empathy, planning and intentional leadership.
• Building internal champions – identifying and empowering individuals who understand the technology and the business, and who can carry the change into their practice groups and teams.

Not everyone needs to be an AI builder, but firms do need a critical mass of champions – and they can be at any seniority. Kuzmenko has seen partners in their sixties emerge as some of the most enthusiastic innovators, undercutting the stereotype that only younger lawyers will drive change.

Clients, AI-first firms and the build-vs-buy question

Surprisingly, Kuzmenko does not believe client pressure is yet the decisive driver of AI adoption, even though many in-house teams are unhappy with pricing and inefficiencies. Longstanding relationships and perceived “team extension” dynamics mean GCs may be slow to push hard for change.

At the same time, AI-first boutiques are emerging without legacy processes, re-imagining staffing, pricing and service delivery from the ground up. For them, investing in R&D is natural; many have built their own GPTs or experimental agents over 2023–2024.

On the build vs buy question, Kuzmenko draws a clear line:

• Building highly specific, deeply integrated tools in-house for niche workflows can make sense.
• Trying to build a horizontal, general-purpose AI workspace internally is increasingly hard to justify economically when specialised vendors are iterating at such speed.

Trust is the real barrier

Asked about the biggest barrier to entry for legal tech vendors, Kuzmenko’s answer is simple: trust.

Mindset matters – if a firm is fundamentally uninterested in innovation, the conversation never starts – but once discussions are underway, progress stalls if there is no trust in the brand, the people, and the roadmap.

Kuzmenko highlights:

• Credibility and track record – many firms have been burned by early-stage tools that pivoted or folded after one or two years, wasting internal time and money.
• Radical honesty about limitations – overpromising features that are “on the roadmap” and then slipping timelines is a fast way to destroy trust. Clear statements of what works today, what does not and what is realistically coming next are essential.
• Human connection – workshops, in-person visits and face-to-face time with client champions still matter enormously. Legal is a relationship-driven market; AI does not remove the need for humans who show up and deliver.

What changes by 2026?

Looking ahead, Kuzmenko expects that by 2026 most firms and legal departments will have made the critical strategic choices:

• which tools or platforms they are backing
• whether they build selectively in-house or rely primarily on vendors
• how far they are prepared to invest in AI skills and experimentation among associates

The firm-side “speed layer” will then accelerate: organisations will re-engineer processes, rethink the classic leverage pyramid and use proprietary data and expertise to differentiate. Partners will increasingly act as enablers, using their market knowledge and client relationships to steer how AI is deployed, while empowering the lawyers below them to build and iterate.

The overarching message was that this is a once-in-a-generation chance for firms to re-evaluate everything they do and how they do it – and to use technology to flatten hierarchies, engage younger lawyers and rebuild the last mile of trusted, human legal advice around a far more intelligent core.