By Joti Basi
AI is already part of most collaborative cases, whether or not anyone has said so out loud. Clients arrive with AI-drafted proposals, budgets and "legal research." Colleagues use AI tools to draft letters and summarize documents, often in free consumer apps running on default settings. And bar associations and courts are steadily issuing guidance on how professionals should use these tools and when they should disclose it.
For Collaborative professionals, the question is no longer whether AI will touch our cases. It's whether the team will decide together how it's used, or leave each professional to work it out alone.
What AI is, and what it isn't
Generative AI works by predicting likely next words based on patterns in the text it was trained on. That's why its output reads so smoothly. It's also why that output can be wrong in ways that are hard to spot. AI has no built-in fact-checking and no legal or clinical judgment. It is a powerful drafting assistant, not a source of truth. Fluency is not accuracy.
Why a Collaborative team raises the stakes
In Collaborative Practice, a family works with several professionals at once: each spouse's attorney, often a coach, a neutral financial professional, sometimes a child specialist. Each brings their own tools and habits. If one team member pastes a financial statement into a free chatbot, the whole family's privacy is affected, even if everyone else is careful. A team is only as protected as its least careful tool.
That's why AI use on a Collaborative case works best as a team decision, made at the start and written down, rather than a private choice each professional makes on their own.
Know where the information goes
Free consumer tools and business tools are not the same. Under default settings, consumer tools may use what you type to improve their models, may keep conversations and have them reviewed, and usually don't offer you any confidentiality agreement. Business and enterprise tools typically exclude your data from training under contract, define how long data is kept, and offer a data processing agreement and administrative controls.
A simple rule of thumb: don't put anything into an AI tool that you wouldn't send to a stranger in an unencrypted email, unless the tool's terms say otherwise.
Before adopting any tool for client work, ask the vendor plain questions. Is my data used to train your models? How long is it kept, and can I delete it? Will you sign a confidentiality or data processing agreement? Where is it stored, and who can access it? Can I export my data if I leave? And never let an AI note-taker record a joint session unless everyone in the room has agreed.
Verify before you rely
The best-known warning came in Mata v. Avianca (S.D.N.Y. 2023), where attorneys were sanctioned after filing a brief citing cases ChatGPT had invented. In a divorce, the error is more likely to be a number than a case: a balance carried forward from the wrong month, a support figure calculated on the wrong income, a date that slipped.
So recalculate support and property figures yourself. Compare any AI summary against the original documents. Be wary when a confident tone covers an uncertain point. And treat every AI output as a first draft, never as finished work.
Risks that matter more in family matters
Some AI risks are sharper when a family is in transition.
Bias and neutrality. AI can frame issues in ways that favor one spouse or reflect assumptions about gender, culture or income. That matters for every professional, and especially for neutrals, whose value depends on both spouses trusting them.
Power imbalances. Polished, AI-generated settlement options can look finished before a quieter spouse has fully said what they need. Make sure options grow out of the family's own conversation.
Safety screening. Screening for domestic violence and coercive control depends on trained human judgment. AI should never assess safety or decide whether the collaborative process is right for a family.
Children's information. Parenting discussions involve sensitive details about children who can't speak for themselves in the room. That information deserves the most caution of all.
Where AI fits, and where it doesn't
A practical way for a team to sort AI uses is green, yellow and red.
- Green: drafting templates and checklists, plain-language explanations of the process for clients, and administrative tasks.
- Yellow, with care and in vetted tools only: summarizing documents, drafting agreement language for professional review, brainstorming options, and preparing meeting agendas.
- Red: entering identifiable client information into consumer tools, relying on unverified AI legal research, any safety or risk assessment, letting AI make substantive decisions, and recording sessions without consent.
A one-page team agreement
Most Collaborative teams already agree on roles, communication and fees at the start of a case. Add AI to that conversation and capture it on a single page:
- Inventory. Which AI tools does each professional use?
- Information. What may never go into an AI tool in this case?
- Approval. Which tools has the team agreed are acceptable, with proper agreements in place?
- Review. Who checks AI-assisted work before it's shared with the team or the clients?
- Disclosure. How will the clients be told, and how can either of them opt out?
Add one neutral question to each client's intake as well: "Have you used any AI tools to research or prepare for your divorce?" The answer often explains expectations the team would otherwise never see.
Even a one-page agreement shows reasonable care and keeps everyone consistent. It doesn't require special software. It requires the same thing Collaborative Practice always has: agreeing, out loud and in writing, on how information moves.
The bottom line
Protect confidentiality first. Verify everything. Put the team's approach in writing. And remember that ethical duties don't transfer to software: AI drafts, and the professionals decide and confirm.
For guidance that applies to lawyers, the American Bar Association's Formal Opinion 512 on generative AI tools(July 2024) is a good starting point; the ABA's announcement of the opinion summarizes it in plain terms. IACP's Ethics and Standards and the Uniform Collaborative Law Act resources remain the foundation for how collaborative teams protect confidentiality and full disclosure.
To learn more about the global community working to help families, visit IACP's About Us page or find a Collaborative Professional near you.
About the author
Joti (Sukhjot) Basi is a divorce mediator with Crossroads Neutral LLC and a collaborative professional in King County, Washington. After 20+ years building enterprise software, Joti founded VennBoard, a shared workspace for collaborative divorce teams.
