Renee Jackson is the founder of Dyspute.ai, a company built around what she believes is the first true AI mediator. She came up with the idea in March 2024, launched a beta in March 2025, and relaunched a fully rebuilt version this past January. She is also, not incidentally, a former college basketball player, a lawyer with in-house experience at several companies, and someone who has clearly spent a long time thinking hard about the gap between how much mediation could help ordinary people and how little of it they can financially access.
That gap is the whole reason Dyspute.ai exists. “Mediation is very effective, and I’ve used it very effectively over my career,” Renee told us, “but it’s really expensive and time-consuming.” Those everyday disputes, such as a withheld security deposit, a refund that never came, or a landlord-tenant dispute, rarely get anywhere near a mediator, because the cost of mediating often exceeds the value of the claim itself. Her founding question was simple: Can AI effectively mediate common two-party disputes, the high-volume, low-dollar-value cases that never make it into a mediation room today?
Meet Adri
The AI mediator at the center of the platform has a name: Adri, short for ADR plus the letter I. Users can even choose Adri’s avatar, a woman, a man, a different race, gender-neutral, whatever makes them comfortable. The process starts with a private intake chat, separate for each party, where Adri asks questions tailored to the dispute: what happened, what are the key dates, what documents matter, and what would a resolution look like. Renee was precise about what Adri is not doing during this stage. “She’s not doing any legal analysis” or estimating how your case would play out in a particular jurisdiction. The bot just focuses on whether there’s any common ground that can be found for a settlement, pulling out each party’s underlying interests and goals, and looking for overlap in what a workable settlement might look like.
A Mediator’s Proposal, Right Out of the Gate
The most distinctive design choice in the platform is what happens next. Rather than waiting through the hours of positioning and gamesmanship, it often takes a traditional mediation to help the parties get to a number both can live with. Adri makes the first (non-binding) settlement proposal immediately. “We bring the mediator’s proposal to the beginning of the process,” Renee explained, “really to cut through a lot of the inefficiencies that you will find in a traditional mediation… who’s going to make the first proposal, what should it be, and there’s that gamesmanship.”
Parties vote to accept or reject that first proposal. This is a deliberate departure from how mediators typically run blind mediator’s proposals, as the vote itself is visible to the other side, even though the private feedback each party gives Adri is not. “We were very transparent about that,” Renee said. From there, parties can keep voting on Adri’s revised proposals, make direct counteroffers, or modify what’s on the table, entirely at their own pace. “We did try to mimic a lot of human mediation, including the caucusing,” she told us, “but we’ve put our spin on other parts of it as well, just to make the process a little more efficient.”
Human in the Loop, By Design and By Feedback
Dyspute.ai’s original beta was fully automated, with no human involved at all. Renee didn’t sugarcoat what came back from that testing. “A big chunk of the beta testing feedback was like, this is cool, I would feel more comfortable if there was a human somewhere involved in the process.” So, the team rebuilt the platform from scratch, and the January relaunch is now sold to mediation providers, including solo neutrals and mediation organizations who invite parties in and then decide, case by case, whether and when to step into an AI-mediated process.
That mediator’s visibility is deep. “They can see literally everything,” Renee said of the human overseer’s dashboard: the intake chats, where a party stalled, how each side voted on proposals, the private feedback given to Adri, shared documents, and whether a settlement agreement has been signed. Notifications flag exactly what a busy provider needs to know. This includes when a vote comes came in, or whether a party hasn’t engaged in three days. “If you’re managing hundreds of cases,” Renee said, “there are certain cases that are just going along smoothly… just leave it, let it be. But if there are cases that are stalling, or maybe the parties have questions… they can step in to see if they can get it the rest of the way.”
Guardrails Against Hallucination and Bias
When Amy raised the well-known struggles of earlier AI-mediator experiments, including hallucinated “facts” that were never part of the record, Renee had a detailed answer ready. Dyspute.ai currently runs on two large language models (down from three at one point), and Adri is instructed to work only from the case file. “She is not going to go out to the internet,” Renee said. “That’s really what prevents the hallucinations from happening — she’s instructed, this is your world of information. Get creative within this world, but don’t go outside and seek external sources.”
Adri is also trained on the ethical standards a human mediator would follow. Renee named the ABA, JAMS, and ICODR standards specifically and noted that Adri is instructed to disregard any demographic information she might pick up, an explicit attempt to reduce the bias that both humans and language models bring to the table. Every input a user submits, and every proposal Adri generates, is scored in real time against abusive language and proposal quality rubrics Renee and her co-founder built. “We’re kind of dialing the temperature down as far as emotion is concerned,” she said, “because… that often is what stalls negotiations, is someone gets dug in on their position.” User data isn’t used to train the underlying models, and Renee, a lawyer herself, was blunt about her standard for the whole build: “I don’t want to just put some ChatGPT wrapper AI slop out there. We really take our role there seriously.”
Where This Is Headed
Dyspute.ai shifted this year from selling directly to consumers to selling to mediation providers, and Renee’s pilot programs are just beginning to generate real case data, which she hopes to have in meaningful volume by fall. Her longer-term targets are courts and state agencies, including small claims courts and bodies like the EEOC, but she’s realistic about the order of operations: “They’re not going to be the guinea pigs,” she said of courts, so the pilots come first, with the goal of eventually showing that the platform works as well as, or better than, a human mediator.
She’s equally candid about what she won’t build yet, even though the demand is obvious. Providers have asked for something closer to an evaluative case score: a rough sense of how strong a claim is in a given jurisdiction. Renee is wary of that kind of certainty.” You wouldn’t want a party relying on, oh yeah, you’re guaranteed to win this,” she said, especially given how relentlessly upbeat large language models tend to be about anyone’s case. “That’s the hard thing… I worry about building that with a hundred percent accuracy.”
Asked directly whether AI is coming for mediators’ jobs, Renee didn’t hedge: “A resounding no.” The cases she wants Adri to handle are ones that mostly go unmediated today, not the complex, multi-party disputes that call for human judgment. She’s watched mediators at conferences like the ODR Forum and the International Academy of Mediators embrace these tools rather than resist them, and her read is straightforward: mediators who learn to use AI well will be the ones who thrive. What AI still can’t do, in her view, is the harder emotional work — reading the room, sensing when two parties are dug in, finding the creative nudge that gets them unstuck. “There will always be things that AI is not good at, or not as good at as humans,” she said, “and that’s where humans will play the largest role — to complement the AI.”
We’re grateful to Renee for walking us through Adri in such detail, and we’ll be watching closely as her pilot data comes in this fall. Information on Dyspute.ai accompanies this column for any mediators who want to learn more.
N.B. Claude developed a draft of this column, which the authors heavily edited.
