A year ago, I wrote a series of articles, GC x AI, arguing that artificial intelligence would remake the substance, process, and economics of legal services — with the general counsel in the eye of the storm. A year later, I’m picking up a red pencil to mark up those predictions.
My accuracy varied. Some predictions have settled into the industry’s working assumptions. Some were wrong, though less about where the technology was going than about how fast institutions would follow. On others, especially economics, the verdict is still out.
READ THE SERIES: GC x AI: Reinventing the General Counsel Role in the AI Era
What Came True
The foundational call held: AI would reinvent legal work, not merely automate it. Legal work is moving from single-purpose tools to connected workflows and agents that work less like software than like members of the team: researching, drafting, and managing the flow of matters and documents.
Several structural bets landed harder — and sooner — than I expected. Firms and clients now work inside shared AI environments, so collaboration itself is becoming a product. And legal memory — systems that retain not just what a team decided but why — has moved from speculation to product frontier. That raises a question the series didn’t ask: Will legal departments control that memory and be able to take it with them, or rent it from whomever holds it?
The series also argued that legal tech winners would bridge coder and counsel. The market didn’t just agree; it created a whole new profession to do it: the legal engineer. Often not lawyers at all, legal engineers don’t practice law. Working alongside AI engineers, they build the systems legal work runs through.
“Off-label” use was one of the series’ most important predictions. The tools have proven to be adaptable well beyond their marketed purpose, and in-house teams have been adept at bending them to new uses. Some of our most valuable innovations this year came from repurposing tools for our own idiosyncratic needs.
That adaptability, together with legal’s visibility across the enterprise, has moved legal — GCs and legal ops leaders especially — toward the center of enterprise AI adoption. A year ago, I argued the GC would lead change within the legal department and with outside counsel. I didn’t anticipate how far beyond the department that role would reach. Several GCs have since taken on enterprise AI leadership roles — some with new titles, others simply with a broader portfolio.
At our company’s senior leadership retreat this fall, I gave the keynote on enterprise AI. I opened with the preemptive question: Why is the CLO the lead speaker on AI? A big part of the answer is how much changed in a year. AI tools built for legal multiplied and improved, and legal adopted and adapted them faster than anyone — including me — expected. But the deeper answer goes to what the series argued from the start: Legal AI’s ultimate value lies less in speed than in strategy. And legal already sees risk, transactions, and governance across the enterprise.
The market has noticed. Peers and recruiters tell me GC searches increasingly emphasize AI fluency and change leadership. The “Steampunk GC” — my shorthand a year ago for the analog lawyer leading digital change — is becoming a job description.
What I’d Revise
I wrote that diligent validation by the lawyer in the loop was the only safeguard against hallucination. That’s still true. What’s surprised me is the persistent lack of rigor. Lawyers who once failed to Shepardize now cite cases that never existed. Nor did I foresee “AI slop” finding its way into advice, briefs, and contracts. Sloppiness was always there. But the scale is new.
In many cases, AI’s efficiency strains the essential human in the loop: more output, same reviewer. The encouraging response is that governance has become an even richer discipline, as the profession builds the structures to match AI’s volume and capacity.
I predicted prompt engineering would become a prized, durable skill for junior lawyers. Instead, it split. As the tools learned to take plain instructions, the mechanics of phrasing a request came to matter less, and building those systems became the job of a legal engineer — now an alternative path for some junior lawyers. A different but crucial part of the workflow has remained with lawyers: prompting on questions of strategy and judgment — framing problems, challenging premises, and testing arguments.
That is the move from document excavation to insight that I had envisioned for junior lawyers, and it still looks right. But it has also sharpened the training question: How do associates develop judgment when the routine work that once built it has become automated? I still see promise — for different reasons. Stanford Law’s Megan Ma has written about building digital versions of mentorship, so junior associates can learn how individual senior lawyers spot issues.
What Hasn’t Happened — Yet
My economic predictions fared worst, at least for now.
Law firm rates haven’t fallen. They rose, and the largest firms had a banner year. I’ve called the billable hour the Twinkie of the legal profession, forever outliving its predicted shelf life. AI’s productivity gains are real, but firms are banking them as margin.
Yet rates and profits are rear-view indicators. Elite boutiques keep multiplying and gaining ground. AI-native firms have arrived: Norm Law, chaired by Sidley’s former chair, Mike Schmidtberger, is taking Big Law work from Wall Street clients with private equity behind it. Two more launched this fall: Arceus in the US and Falcon in Europe, the latter pledging to “bury the billable hour.” The fragmentation I predicted is underway. The tectonic plates are rumbling.
Where I was flat wrong — also for now — is the lateral market. I argued that institutional memory would make partners less portable. Instead, the lateral market is white-hot. But the causation may now run the other way: The lateral market may be shaping legal AI. As profits go to rainmakers, lawyers with deep expertise but no portable books are heading to AI-native firms and Big Law platforms that use AI to rework law firm economics. I thought AI would keep talent in place. It may be giving the experts somewhere new to go.
I wrote that legal departments would become leaner. They mostly haven’t shrunk, but they haven’t grown, either. AI is filling the capacity gap, and in some departments creating new capacity from existing headcount. The challenge for GCs is where to direct that capacity as legal’s portfolio widens, the technology advances, and headcount doesn’t budge.
The Big Question Now
In a legal webinar last fall, a questioner pressed me on AI’s risks to humanity. I steered the conversation back to legal technology. When he persisted, I finally said, I’m just a GC. My focus was on transforming legal work. The larger question, I suggested, was beyond my responsibility.
His question has only grown more consequential. I’m less certain that my answer holds up.
A year ago, I made a great many predictions about how AI would change legal work. Looking back, I wonder whether my biggest miss wasn’t a prediction I got wrong, but a question I thought wasn’t mine to answer.
Columnist Eric Dodson Greenberg is executive vice president, general counsel, and corporate secretary of Cox Media Group. Eric writes about leadership, legal operations, and the intersection of the legal industry and AI for Bloomberg Law’s Good Counsel column.
Read More Good Counsel