We sat down with our friend and colleague Sarah Cole, a mediation scholar and practitioner, and the Michael E. Moritz Chair in Alternative Dispute Resolution at the Ohio State University Moritz College of Law. Sarah is now in her third year as a Scholar in Residence for the International Academy of Mediators (IAM), a role she designed herself around a single, deliberately narrow question: not how AI affects the parties in a mediation, but how it affects the mediator.
“I raised the possibility of focusing on AI,” Sarah told us, describing her interview for the residency, “which I’m humbled to be doing in the presence of two AI aficionados.” She wanted to push herself into unfamiliar territory, and she found good company. IAM’s members, in her words, are “distinguished mediators who are very experienced, from all over the world, all confronting the same questions: To what extent should I use AI? If I use it, how should I use it? How should I go about disclosing my use, if at all, to the parties?”
The Deskilling Divide
What surprised Sarah most was how sharply mediators split over a single issue: whether AI is sharpening or eroding their skills. Many mediators feel AI is undermining their abilities — Cole used the term “deskilling,” a word that shows up throughout the literature. But at a recent IAM group session, she heard the opposite reaction voiced just as strongly. “There’s the group that says it’s ruining me, I’m not as effective a mediator… and then the other group that says, I love having AI there to engage with.”
That second group’s enthusiasm, Sarah suggested, has less to do with the technology than with something structural about mediation itself. “One of the hardest things about being a neutral is how solitary it is,” she said. An arbitrator sitting on a panel can turn to a colleague and ask what they heard or think. Mediators do not have this privilege. “As a mediator, you’re all alone… a lot of times you almost feel like you’re under attack,” fielding pushback from parties in real time with no one with whom the mediator can workshop ideas. For the mediators in that second camp, the appeal of AI is having something to bounce ideas off during a break: “The party said X to my Y, and then I tried this, and that didn’t work — do you have any other ideas?” That said, she was careful to add a caveat: confidentiality must be protected by a genuinely closed system, something many mediators don’t yet fully understand or trust.
Where Mediators Are Actually Using AI
Pressed on where AI shows up in real practice — during prep or live in a session — Sarah’s read was clear: almost entirely in prep, not in the moment. “Nobody I spoke with has used it that way,” she said of real-time use between caucuses, although she suspects some mediators would welcome an AI presence sitting with them during a session if the tools (and appropriate confidentiality measures) existed. Instead, she pointed to high-volume practitioners, particularly employment mediators in Southern California juggling multiple cases a day, using AI to summarize briefs. “That’s the area I’m seeing it used more,” she said, while flagging an obvious tradeoff: “Even summarization leads to some anchoring for the mediators.”
The other common use is drafting mediators’ proposals, especially in commercial and employment mediations, where that practice is routine. Mediators feed anonymized case information to AI and ask what it thinks the proposal should be — sometimes before forming their own view, sometimes after, to check their instincts against it.
One story stood out to us, and clearly to Sarah as well: a mediator affiliated with a large ADR firm has been feeding pleadings into a closed system and having it draft a full arbitrator’s opinion from each side’s perspective, then showing a party the opinion written against them during caucus. “He’s very clear the AI is the one that writes these, and he tells them that,” Sarah said. “And it’s very compelling when they see, in black and white, a loss they weren’t expecting.”
The Comfort of an Incomplete Picture
We pushed Sarah on a worry that comes up often on this podcast: that a summary can miss the fact that actually mattered, because AI models are grounded in probability, not understanding. Her answer reframed the problem in a way we hadn’t considered. Mediators, unlike arbitrators, were never working from a complete record to begin with. “There’s no way you’re possibly going to be able to read all the pleadings in a lot of cases,” she said, recalling her own volunteer mediation days at federal court, staring down “a pile of pleadings” no one could fully absorb before a session. When she uses AI to review pleadings, she said she rarely even looked at the output mid-mediation: “You’re so on, you’re so engaged… there were no real breaks.” In her view, the imperfection of an AI summary sits inside a process that was already imperfect. “That worries me a little less with mediation, because you’re never going to have the total picture.”
She’s also watched mediators turn AI into a de-escalation tool for a newer problem: parties who arrive convinced because “AI told me I would win my case.” Rather than confront that certainty head-on, some mediators told her they simply sit down with the party and run the scenario through AI together, shifting the facts slightly and generating the other side’s view alongside it — “a good way of having a non-confrontational discussion” with someone who has dug into an AI-reinforced position.
Disclosure That Doesn’t Scare
Sarah’s survey of IAM’s roughly 200 conference attendees — about 36 responses so far, with the link to participate included alongside this column — is designed to map how mediators are actually disclosing AI use, and what response they get. What she’s found so far surprised her: several mediation firms now include model disclosure language directly in their agreements to mediate, and the mediators using it report no pushback at all. “I thought the parties would be a little more skeptical,” she admitted, adding that she suspects the hesitation mediators feel about disclosure has more to do with their own anxiety than the parties’ reaction: “If the parties know I’m using AI, will they still feel they need me?”
One finding cuts against the efficiency story AI is usually sold on: several mediators told Sarah that using AI is making their prep take longer, not shorter, as reading pleadings gives way to extended back-and-forth with an LLM.
Cole’s Concerns
Looking ahead, Sarah’s biggest concern isn’t today’s tools but tomorrow’s — specifically, real-time sentiment and emotion analysis. She worries about accuracy (“You’ll think that they’re nervous, but they’re not… some people laugh at funerals because that’s how they’re expressing emotion”) and about overload: an AI flagging that “party A looks a little upset,” on top of everything else a mediator is already tracking, risks burying the very judgment it is meant to support. She described watching a colleague use AI to generate thirty settlement options spread across two full columns for a single case — technically impressive, practically useless. “We humans are not in a position to internalize that,” she said, which is why she teaches her own students to walk into an impasse with one or two tools in mind, not a menu: “In the heat of the moment, you don’t remember anything except caucus.”
Asked directly about Renee Jackson’s fully AI-driven mediator platform, which we featured in an earlier column and which Sarah has encountered on a shared panel, her view was measured rather than dismissive. Renee, she noted, is candid that the tool is built for straightforward disputes, not complex litigation. Sarah’s own hesitation is less about capability and more about what gets normalized: “Will a six-year-old today, ten or twenty years from now, think that what an LLM does is better than any human being?” That question, more than any accuracy statistic, is the one she keeps circling back to — alongside her conviction that empathy and eye contact are not features a model can replicate.
We’re grateful to Sarah for her candor and her ongoing research. Please take a moment to add your voice to her survey, The Future of the Room: AI and the Master Mediator.
By William Froehlich and Amy J. Schmitz, with the first draft created by Claude from the interview recording transcript.
