Version Up

Denisa Kopandi returns to Version Up to discuss what happens when legal AI agents meet the uneven, iterative work of an actual transaction. Drawing on her experience testing legal AI tools and building workflows, she explains why a sequence that looks convincing in a demo can be difficult to trust in practice. We distinguish between agents working behind a chatbot’s answer and workflows intended to carry a legal matter through multiple stages. Denisa describes the gap between a neat progression from intake to analysis to drafting and the reality of documents arriving in batches, incomplete information, and clients changing the structure or timing of a deal. A central theme is the legal analysis that happens before anything gets written down. A lawyer may open an apparently routine document and notice something that changes the entire assessment. That makes it difficult to specify every task or decide in advance where an agent should stop for review. Checking the final legal analysis may not reveal that an earlier agent overlooked a document or extracted the wrong information. We explore a more manageable approach: specialized agents for narrower tasks, with lawyers deciding when to use them and how the results fit together. Examples include assessing a merger notification threshold or preparing a request for information using what is already known about the client. The question is how much useful work can be delegated without creating a verification burden that cancels out the time saved. Denisa also shares an experiment in which an agent began deleting earlier versions of a due diligence report, leaving her to transfer the analysis manually. We discuss what that experience says about reliability, alongside the difficulty of trusting a batch of contract analyses after finding an error in one. There is a training dimension too. Denisa asked interns to use AI for research and submit their full conversations, so she could review how they questioned the model. We consider whether narrower tasks and visible review can help junior lawyers develop judgment, and how quickly improving tools affect the decision to invest in building a workflow. Denisa Kopandi is a senior associate at bpv GRIGORESCU ȘTEFĂNICĂ in Romania, working across technology, competition, and commercial matters. Alongside her client work, she evaluates and implements legal AI tools and explores how they can support specific legal tasks and the development of junior lawyers. You can follow her on LinkedIn. * * * Version Up Members is a free, invitation-only community for lawyers and legal professionals putting AI to work at law firms and legal departments. Request an invitation: versionup.ai/members

Christopher Smith joins to discuss what it takes for law firms to turn their accumulated documents, data, and experience into knowledge that lawyers can actually use. Drawing on his previous work at Sullivan & Cromwell and Freshfields, Chris explains why many of the challenges behind today’s AI ambitions have been familiar to knowledge management teams for decades. We start with the practical problem of collecting information from busy lawyers. Chris describes why matter intake and document profiling need to fit around getting the work done, why firms should avoid asking for the same information twice, and how overly restrictive systems can push lawyers into workarounds. As collaboration spreads across email and other platforms, keeping track of authoritative documents becomes more complicated. That leads to a question I’ve often had when looking at legal AI tools: who is going to make sure they’re working from the right material? Chris sees improving tools for organizing and cleaning data, but someone still has to decide which information is correct. He shares how he brought teams at Freshfields together around shared firm data, with clear responsibility for particular decisions. We also discuss the value of a vetted collection of useful material and why lawyer attention remains a constraint on evaluating what AI finds. The incentives matter just as much. At Sullivan & Cromwell, Chris describes connecting knowledge contributions to performance evaluations and looking at whether other lawyers actually used the submitted material. We explore the tension between building a firm’s intellectual capital and lawyers’ ability to take their experience and client relationships elsewhere. Looking ahead, we consider whether broad legal AI platforms will become a common baseline, with differentiation coming from how lawyers combine tools, data, and judgment. We also discuss the apprenticeship model, emerging roles that combine legal practice and technology, and whether changing client expectations and billing arrangements will create stronger incentives to work differently. Chris closes with a candid explanation of why he is pursuing opportunities outside law firms and the limits he sees in their rewards for people who build technology. Christopher Smith is the former Chief Information Officer and Chief Data Officer at Freshfields and former Chief Knowledge Management and Innovation Officer at Sullivan & Cromwell.  **Corrections: TensorFlow was released in 2015. Generative AI emerged in 2022/2023.** ALL VIEWS EXPRESSED OUR OWN * * * Version Up Members is a free, invitation-only community for lawyers and legal professionals putting AI to work at law firms and legal departments. Request an invitation: versionup.ai/members

Findley Penn-Hughes, a Senior Associate at Mayer Brown, comes on to help provide a practical check-in on where AI actually stands in legal practice. Findley and I compare our experiences adopting AI over the past few years—from early hallucinations that undermined trust to the point where tools like Harvey, Copilot, Claude, and ChatGPT have become part of everyday work. We also dig into why the next step, agentic AI, is proving much harder than the demos suggest. Findley shares his experiments building AI agents to monitor regulatory developments and draft client alerts, along with the technical and reliability problems he encountered. A major part of the discussion is cybersecurity. For law firms and legal departments, the issue is not just whether an AI system produces accurate work, but what happens when it is given access to sensitive client data, internal systems, and the ability to take actions autonomously. We talk about the risks of data leakage, third-party model providers, increasingly capable cyberattackers, and the possibility that more autonomous agents introduce entirely new attack surfaces. Even where those risks can ultimately be managed, the uncertainty itself helps explain why legal organizations are moving more cautiously than the AI hype cycle might suggest. We also spend time on incentives. Lawyers are still largely rewarded for producing billable work, not for spending time building systems that make that work faster or eliminate it altogether. That creates an obvious tension for associates and partners asked to invest non-billable time in AI workflows. Firms may need to rethink billable-hour credit, compensation, staffing, and the role of dedicated AI specialists. And as clients increasingly expect AI-driven efficiencies to translate into lower fees, firms will also have to work out how they price legal work and capture value from the technology they build. From there, we explore other barriers slowing AI adoption across legal: messy and unstructured data, questions about how much proprietary “secret sauce” law firms really have, uncertain economics, and a technology market with no clear winners yet. We close by discussing what AI could mean for law firm business models, competition between firms, junior lawyer training, and the future shape of the profession. Our conclusion is cautiously optimistic: AI is already changing legal practice, but the industry is still much earlier in the transformation than the hype sometimes suggests. https://www.linkedin.com/in/findley-penn-hughes   ALL VIEWS EXPRESSED OUR OWN * * * Version Up Members is a free, invitation-only community for lawyers and legal professionals putting AI to work at law firms and legal departments. Request an invitation: versionup.ai/members

In this episode of Version Up, I am joined by Melanie Hallas Kirsch, a senior in-house lawyer, for a candid look at her personal journey adapting to generative AI. Mel shares how an everyday practitioner in a high-stakes practice area moves from basic curiosity to active experimentation, navigating the overwhelming flood of new tech tools without losing sight of practical realities. Rather than treating AI as a futuristic silver bullet, Mel breaks down how her perspective on legal technology has evolved, offering an honest look at how corporate attorneys can realistically integrate these tools into their daily work. Mel’s journey with technology began long before the recent generative AI boom, drawing on years of using early AI-like tools such as technology-assisted review (TAR) during large-scale document reviews. As general purpose and legal-specific AI tools came into being, she adopted a deliberate strategy to deploying them in her practice. A key highlight of Mel’s journey is her realization that senior legal experience acts as a crucial superpower when adopting AI. Because senior lawyers possess deep practice intuition and can instantly spot when an output fails the "smell test," they can use these tools safely as drafting assistants or advanced search engines. Successful adoption requires changing one's mindset to actively question and push back on the tool rather than accepting its first output. Learning how to leverage AI hasn’t come from generic, click-through corporate webinars, but from continuous self-education—listening to podcasts during long commutes, exchanging practical tips with peers, and taking time to experiment hands-on. In her day-to-day practice, Mel’s biggest productivity gains have come not from chasing niche legal software, but from personalizing general-purpose and off-the-shelf legal AI tools to match her personal workflow. She also openly shares her ongoing experiments and friction points—from attempting to build custom multi-step workflow agents to navigating data connectivity, security boundaries, and information siloing. On how AI will impact the client-attorney relationship, Mel offers a grounded framing. She explains how law firms will continue to do work for corporate clients, but that will move up the value chain to higher-stakes and higher-judgment work. She also outlines the ways that AI can strengthen the client-attorney relationship, whether it's a law firm and their client or an in-house lawyer and their business colleagues within a corporate.  Ultimately, Mel’s journey offers a grounded perspective: while AI is an incredible aid for streamlining repeatable work and research, human judgment, risk tolerance, and client relationships remain the true core of legal practice. https://www.linkedin.com/in/melanie-hallas-kirsch/   ALL VIEWS EXPRESSED OUR OWN * * * Version Up Members is a free, invitation-only community for lawyers and legal professionals putting AI to work at law firms and legal departments. Request an invitation: versionup.ai/members

In this episode, I sit down with Professor Kipp Coddington from the University of Wyoming College of Law to discuss how legal education is adapting to the age of AI and what skills tomorrow's lawyers will need to succeed. Kipp teaches an innovative course on AI and Internet Law that combines substantive legal issues with hands-on technology training. Students are expected to use AI throughout the course, experiment with tools such as Gemini Notebook, complete programming exercises, and analyze the legal, policy, and technical challenges emerging from AI adoption. His view is straightforward: lawyers do not need to become software engineers, but they do need to understand and embrace the technologies that are increasingly shaping legal practice.  A major theme of our conversation is the debate taking place across legal education about how much AI students should be allowed to use. Kipp argues that law schools are trying to strike a balance between exposing students to transformative technology and preserving the critical thinking, reasoning, and judgment that legal education has always sought to develop. In his view, the goal is not to outsource thinking to machines but to teach students how to use AI responsibly while continuing to strengthen their own "wet neural network" through rigorous analysis and problem solving.  What stood out most from the discussion was Kipp's experience teaching students in an AI-enabled environment. Despite concerns that generative AI might make student work more generic or less original, he observed the opposite. In a class assignment where students used AI tools to develop presentations on emerging legal issues, students brought a unique perspective, resulting in a surprisingly diverse range of topics and insights. Rather than producing forty versions of the same answer, students explored issues ranging from healthcare data and privacy to international AI regulation and business risk. Their own judgment, creativity, interests, and experiences shaped the output.  That observation led to one of the most important takeaways of the episode: even in an AI-powered world, junior lawyers and law students continue to create value through their perspective. A student or first-year lawyer may not have decades of legal expertise, but they can bring fresh thinking, creativity, problem-solving ability, and a unique lens to legal challenges. AI can accelerate execution, but it does not replace the human judgment that helps lawyers identify risks, generate insights, and develop novel solutions.  We also explored what skills law schools should be emphasizing in this environment. Kipp believes the most important traits are not technical. Instead, he highlighted humility, effective listening, empathy, intellectual curiosity, sound judgment, and the ability to operate in legal gray areas. Lawyers are fundamentally risk managers and counselors. They rarely provide simple yes-or-no answers. Their role is to help clients understand risk, evaluate alternatives, and chart a path forward. Those skills become even more important in a world where AI can generate answers but cannot fully replace professional judgment.  The conversation also touched on how AI is changing teaching and assessment. Traditional take-home assignments and research papers present new challenges because of the ease with which AI can generate written work. Kipp discussed emerging approaches such as oral defenses, in-person examinations, and more interactive evaluation methods designed to ensure students truly understand the material. While these challenges are real, he remains confident that legal education can adapt and continue to identify and develop talent.  Perhaps most striking was Kipp's optimism about the future. While acknowledging that some traditional legal business models may come under pressure, he sees a lot of opportunity for lawyers who embrace technology. He pointed to growing demand for professionals who understand both law and technology, including opportunities within AI companies themselves. Just as the internet created entirely new career paths for lawyers, he believes AI will create new roles, new industries, and new ways for legally trained professionals to add value.  If you're a law student, young lawyer, legal educator, or anyone interested in the future of the profession, this conversation offers a thoughtful and optimistic perspective on how legal education can evolve while staying rooted in the timeless fundamentals of legal practice.  https://www.linkedin.com/in/kipp-coddington/   ALL VIEWS EXPRESSED OUR OWN * * * Version Up Members is a free, invitation-only community for lawyers and legal professionals putting AI to work at law firms and legal departments. Request an invitation: versionup.ai/members