By Sameena Safdar, CEO, Amplify Your Voice & Customer Success Manager, Thomson Reuters
A Recap of the Women’s Bar Association of DC Litigation Forum Panel | March 25, 2026
On a Wednesday evening in late March, Caplin & Drysdale’s offices hosted one of the more candid conversations happening in the legal profession right now. Presented as part of the Women’s Bar Association of DC’s AI Advantage Series and sponsored by the WBA Litigation Forum, “Is AI the New Second Chair?” brought together a sharp, diverse panel to do something most AI events don’t actually do: get specific.
Moderated by Serafina Concannon, Member at Caplin & Drysdale, the evening featured Kenya Rothstein (Associate Attorney, Aqua Terra Aeris Law Group), Funmi Owolabi (Principal Attorney, The Data Privacy Lawyer PLLC), and Deedra Smith (Managing Director, FTI Consulting Technology LLC). Their technology stacks and use cases differed considerably, which made for a richer conversation than most — and Concannon kept it grounded by opening the floor to audience questions throughout the evening rather than holding them until the end.
Where AI Is Still Falling Short
Owolabi flagged something every lawyer who has wrestled with an AI output will recognize: the lack of human nuance. Because AI processes prompts algorithmically, output can be “quite robotic,” requiring multiple rounds of refinement or breaking a single prompt into several components to get something usable. Prompt engineering, she noted, is a real skill — and one most attorneys haven’t been formally trained in.
An audience member raised the question of agentic AI — systems designed to understand and autonomously execute multi-step tasks without requiring the user to manually decompose each prompt. Smith responded that the trajectory of the capabilities within FTI Technology’s suite is moving in that direction. It’s an area worth watching closely — and one where purpose-built AI platforms designed specifically for legal professionals are already delivering, with agentic workflows trained on legal content and the governance frameworks the profession requires.
The courts are watching too. Concannon offered an observation that landed with the room: the AI certification language appearing in some jurisdictions is “too vague and overbroad.” After all, she noted, you’re using AI today even in versions of Westlaw or Lexis that don’t contain generative AI. The profession will need more precision in how it defines and regulates AI-assisted legal work.
The Data Privacy Dimension: What’s Actually in Your AI Contract
Owolabi’s practice area put her at the center of the evening’s most substantive exchange. Her message was clear: if your organization is using a third-party AI tool, you need to understand what’s happening with your data. What training data was used? Is the vendor training their model on your data? Many third-party tools include terms of service provisions allowing exactly that — “dangerous,” she said flatly. Metadata logs and derived information mean third parties are continuously learning from organizational behavior in ways most users simply don’t realize.
The distinction matters more than most attorneys realize. General-purpose AI tools hallucinate citations, obscure their sources, encrypt nothing, and — as Owolabi made clear — may be training their models on your client data right now. Purpose-built legal AI platforms were architected from the ground up to solve exactly those problems: verifiable sources, data governance built for confidentiality obligations, and legal content that was curated — not scraped. The gap between the two isn’t closing. It’s a different category of tool entirely.
Geography adds another layer. If your organization does business in Europe, China, or Brazil, processing data in those regions may subject you to local legal frameworks. Smith confirmed that Switzerland takes data residency so seriously it is essentially criminal to move certain data across its borders. Owolabi’s summary was concise: “Every use of AI has to have its own analysis.”
Junior Attorneys, the Billable Hour, and What Comes Next
Smith was emphatic: incoming attorneys must know how to prompt effectively — comparing it to the eDiscovery literacy that became baseline competency a generation ago, except now it applies to every lawyer. But she added an important counterpoint: junior attorneys still need to learn the underlying tasks manually. You can’t perform quality control on something you don’t understand.
The billable hour question proved equally unresolved. One audience member asked whether clients are already requesting that only partners or senior counsel be billed, on the premise that everything else can be automated. Another extended the frame further: in-house legal teams are also investing in AI and may soon decide which portions of a matter they’ll keep in-house rather than send to outside counsel. That’s a structural shift in how legal work gets allocated — not just how it gets billed.
The Cost We’re Not Talking About
Perhaps the most unexpected contribution came from Rothstein, whose environmental law practice gives her a lens most AI discussions lack. “Every 20 to 50 searches take up approximately 16 ounces of water,” she told the room — roughly 7 million gallons per day in aggregate. Layer on the greenhouse gas emissions required to maintain temperature-controlled data centers, and the environmental cost of AI at scale becomes very real. An audience member offered the hopeful counterpoint that technology tends to improve its own efficiency over time. It hasn’t yet — and the profession should be honest about that.
Final Thought
“Is AI the New Second Chair?” didn’t resolve that question — and it wasn’t trying to. What it offered was something more valuable: a ground-level view of how thoughtful legal professionals are navigating AI right now, with honesty about what works, what doesn’t, and what still keeps them up at night.
One through-line emerged clearly: not all AI is created equal. The risks around data governance, hallucination, prompt brittleness, and lack of legal nuance are real — but they point toward a solution, not away from one. The legal profession deserves AI built for it, not borrowed from somewhere else.
The second chair isn’t going anywhere. But the memo it produces, the cases it cites, and the data it touches? That’s a different conversation — and this room was ready to have it.
This article reflects the author’s observations and notes from the March 25, 2026 panel discussion at Caplin & Drysdale, Chartered. Statements attributed to panelists reflect the author’s contemporaneous notes and may be paraphrased.