Why family law, and why now.

People ask me why we started with family law. Usually the subtext is: isn't that a small market? Isn't commercial litigation where the money is?

I understand the question. I just think it gets the logic backwards.

We didn't choose family law because it was easy. We chose it because it's the practice area where the evidence problem is hardest, the stakes are most personal, and the existing tools fit worst. If you can build evidence intelligence that works here, you've built something structurally defensible. Start where it's easy and you've built a feature someone else will replicate.

The evidence problem, concentrated

A financial remedies case typically spans hundreds of pages of disclosure. Bank statements, pension valuations, property surveys, company accounts, WhatsApp exports, handwritten notes. The evidence isn't neatly filed. It's scattered, inconsistent, and — often — contradictory.

The job of the solicitor is to find the contradictions. Did one party value the family home at £850,000 in their Form E and £650,000 in a later negotiation? Is there a pension that appears in a bank statement but never in the disclosure? These aren't abstract questions. They determine what a settlement looks like and whether a client is treated fairly.

Today, that work is done manually. Days of review, cross-referencing, and margin notes. It's expensive, it's slow, and it's exactly the kind of work where a tired human eye misses things.

Why generic legal AI doesn't fit

Most legal AI tools were built for contract review or general document search. They answer questions about text. Family law financial remedies needs something different: answers about cases — parties, assets, claims, and the relationships between them.

A generic tool can summarise a bank statement. It cannot tell you, with certainty, that two parties made different claims about the same asset — because it doesn't model parties, assets, or claims at all. It predicts what a plausible answer looks like. In a context where the solicitor carries direct professional liability for what goes to court, a plausible answer is not enough.

That's why we built Pivot around a claims-based ontology — a structured model of the domain itself. What constitutes a contradiction between two valuations. What counts as a gap in Form E disclosure. How Section 25 criteria map to evidence types. That ontology was built with practising family solicitors, and it's the part of the product that can't be replicated by pointing a large language model at a document bundle.

Why now

Three things have converged.

Disclosure has exploded. The evidence in a modern financial remedies case includes message threads, screenshots, and digital financial records that simply didn't exist when the disclosure process was designed. The volume is growing faster than any firm's capacity to review it manually.

The courts are under strain. The majority of family proceedings now involve at least one party without legal representation, navigating disclosure alone. Judges and practitioners are dealing with incomplete, poorly organised evidence on both sides of the "v". The system needs structure, not more documents.

AI accuracy is now a professional issue, not a technical one. Regulators and courts have made clear that lawyers are responsible for the outputs of the tools they use. Hallucinated citations have already led to sanctions. The question firms are asking has shifted from "can AI help?" to "can I stand behind what this tool tells me?" That shift favours deterministic architecture — systems where every output is either traceable to source evidence or doesn't exist.

Depth first, then breadth

Our thesis is simple: in legal AI, depth beats breadth. A tool that does everything shallowly will lose to a tool that models one domain properly — because the domain model is where trust lives.

Family law financial remedies is our starting point, not our ceiling. The infrastructure we've built — structured claims, provenance trails, contradiction detection — extends naturally to other evidence-heavy practice areas, and to the people navigating the system without representation at all. But it starts here, in the practice area where accuracy matters most and where I understood the problem from the inside before I ever wrote a line of the spec.

Why family law? Because it's the hardest place to earn trust. Why now? Because the profession has stopped asking whether to use AI, and started asking which AI it can defend.