Many in the legal industry have now heard of large language models like ChatGPT or Claude, and naturally, curiosity has driven a wave of experimentation. The technology produces a certain “wow effect”: generate a draft, classify a document, summarize a case in seconds. Impressive, and in some contexts genuinely useful. But the real question is not whether AI tools can be tried — it is whether law firms can transform their processes with AI, rather than simply adopt a new product.
The First Challenge: From Demos to Reality
On the surface, AI performs well on the tasks typically shown in demos — summarizing, drafting, basic research. And indeed, as Bloomberg’s 2025 report shows, a clear majority of lawyers, 63%, already use AI in their daily work, whether in research, memo writing, or document review. Usage spans the profession, from junior associates to senior partners.
But when these same tools are applied to real, high-stakes work, their limitations quickly surface. In law, small details matter. The “last 20%” of precision often decides the outcome of a case or negotiation. A misplaced citation or a hallucinated precedent can have catastrophic consequences. We already see examples of “AI slop” in courts, where judges were forced to withdraw opinions containing fabricated references. What works as a time-saver in drafting memos may collapse under the weight of professional responsibility.
So while adoption is widespread, effectiveness in critical tasks remains limited.
The Second Challenge: Adoption vs. Transformation
Here it is important to draw a line between adoption and transformation. Adoption means treating AI as another tool—something that can generate a draft faster than a paralegal. Transformation is something else entirely: it is about rethinking the way work itself is done.
MIT’s State of AI in Business 2025 calls this gap the GenAI Divide. On one side are the many firms experimenting with tools — pilots, demos, trials. On the other, the few who have managed to re-engineer their workflows with AI. According to the report, more than 80% of organizations have piloted generative AI tools, but only about 5% of enterprise pilots reach production and create measurable business value.
The reason is not model quality alone. The problem lies in how work is structured. Just as the internet transformed the very way lawyers searched and processed information, so AI requires a rethinking of workflows: what should be automated, what must remain under human control, and how the two interact.
From my perspective, making this shift demands at least three areas of competence:
- Legal expertise — not just abstractly, but deeply enough to know where precision, compliance, and ethical responsibility make automation dangerous.
- AI understanding — knowing not only what LLMs can do, but also their weaknesses, where hallucinations appear, and how to build safeguards.
- IT integration — since no single tool covers the entire workflow. AI must be combined with existing systems, databases, and compliance processes.
It is the rare intersection of these three that enables real transformation. And it explains why many law firms, despite enthusiasm, remain in the adoption phase. Bloomberg’s 2025 survey confirms this gap: lawyers expected AI to revolutionize billing, workloads, and efficiency, but a year later most report only modest improvements, far short of predictions.
The Practical Path Forward
The temptation for many firms is to try to solve this internally — set up pilots, experiment, allocate time from associates. But this quickly consumes scarce, expensive human resources without delivering structural change.
A more effective approach is to work with partners who combine all three competencies: legal knowledge, AI expertise, and IT integration (which we have in Denovo). With this combination, the firm is not simply shown how to use a tool — it is guided through the redesign of its workflows. That includes identifying where AI genuinely adds value, where risks lie, and how to mitigate them before they surface in practice.
In other words, the goal is not to sprinkle AI across existing processes, but to transform processes themselves, with AI built into the logic of how the firm operates.
Most lawyers are already using AI. Adoption is no longer the issue. The challenge is transformation — restructuring legal processes to reflect the strengths and limitations of AI. As both Bloomberg’s report and MIT’s study show, enthusiasm alone is not enough. The firms that succeed will be those that move deliberately beyond experimentation, combining legal, technological, and AI expertise to carry out a true transformation of practice.
Sources
- Bloomberg Law Report: Artificial Intelligence – The Impact on the Legal Industry (2025)
- Bessemer Venture Partners: The State of AI 2025
- MIT Project NANDA: State of AI in Business 2025