Separated with Children and a Challenging Ex? Discover AI Solutions!

I. The Founder
Sol Kennedy often requested his assistant to read the messages from his ex-wife. Following their separation in 2020, he described their communications as “challenging.” Emails—or a flood of them—would arrive, containing discussions about their two children intertwined with unrelated emotional hits, making it difficult for him to respond. At that time, Kennedy, a seasoned tech entrepreneur and investor in Silicon Valley, was undergoing therapy. However, outside of his weekly sessions, he sought real-time support.
After the divorce, they shifted their communications to a platform called OurFamilyWizard, utilized by countless parents in the United States and beyond to exchange messages, share calendars, and track expenses. (OFW maintains a time-stamped, court-admissible record of all communications.) Kennedy opted for an additional feature called ToneMeter, which OFW promoted as “emotional spellcheck.” As users drafted messages, its software performed a simple sentiment analysis, flagging language that might be “concerning,” “aggressive,” “upsetting,” “demeaning,” etc. However, Kennedy noted a significant issue: His co-parent didn’t seem to be using her ToneMeter.
A tech enthusiast, Kennedy had been experimenting with ChatGPT to “cocreate” bedtime stories with his kids. He decided to seek its advice for interactions with his ex. He was impressed—and he wasn’t alone. Across Reddit and various online forums, individuals dealing with challenging ex-partners, family members, and colleagues were expressing astonishment at the seemingly valuable guidance and emotional validation a chatbot could offer. Here was a machine capable of assuring you, without any hidden agenda, that you weren’t the irrational one. Here was a counselor ready to support you 24/7 as you navigated through myriad challenges. “A scalable solution” to enhance therapy, as Kennedy describes it. Finally.
Yet, out of the box, ChatGPT was overly verbose for Kennedy’s requirements, in his opinion—and excessively apologetic. When presented with difficult messages, it often suggested lengthy replies filled with “I’m sorry, please forgive me, I’ll do better.” Lacking a self, it also lacked self-esteem.
Kennedy envisioned a chatbot with “spine” and believed that if he created one, many other co-parents would find it valuable as well. He believed AI could assist them throughout the communication process: filtering emotionally charged language from incoming messages and summarizing just the essentials. It could propose suitable responses and guide users towards “a better way,” as Kennedy puts it. Thus, he founded a company and began developing an app named BestInterest, named after the legal standard often used in custody decisions—the “best interest” of the child or children. He planned to utilize existing OpenAI models and infuse them with the strength he desired through his prompts.
Estranged partners often engage in intense conflicts for various reasons. For many, perhaps even most, things tend to settle down after sufficient time has passed, rendering a tool like BestInterest potentially less useful in the long run. However, when specific personalities—often described as “high-conflict,” “narcissistic,” “controlling,” or “toxic”—are involved, the disputes regarding the children may persist, at least from one side. Kennedy wanted his chatbot to confront these individuals, so he turned to a figure they might despise the most: Ramani Durvasula, a clinical psychologist based in Los Angeles who specializes in the impact of narcissism on relationships.