Technology’s Effect on Advocacy
A maps interrogatories and document requests to the elements of each cause of action. discovery propounder that A deposition notice drafter that knows a notice to a party from a § 2020.310 records subpoena. A meet-and-confer drafter with a dispute-typed casebook of prior firm exchanges. A “ Matter Gap Scanner” that sweeps every active matter on a schedule — cloud documents, email threads, call transcripts — and surfaces gaps between what we know and what our case management system’s data model records and holds. A records sync that pulls medical records and bills from our records vendor and files them into the matter folder, classified and deduped. A final distribution generator that turns "the check cleared" into a signature-ready distribution statement in minutes; costs ledger pulled, interest on advanced costs computed, fee agreement applied, every lien netted at its negotiated reduction, math shown, all delivered for one-click e-signature by email or text, whichever the client prefers. None of these are demos; we use them all daily. None are sold or licensed from a vendor. All live in code I can read, understand, and own.
For those hesitant to leverage AI, the obvious critique writes itself: a lazy lawyer playing engineer with client data. I understand the architecture I deploy, how these systems talk to each other, how authentication works, where confidential client information lives, how protected health information gets de-identified before reaching a model, what travels encrypted, and what the audit trail shows when something breaks. Before anything touches a real matter, I ask explicitly for the trade-offs: security exposure, privacy implications, failure modes, worst case. I own the ultimate decisions. The AI does not decide, I do, and it is my bar license that takes the hit if I am wrong.
What We Have Built
Here is some of what runs in our production. The catalog matters more than any one tool:
A pleading paper and letterhead builder that produces conforming California pleading paper on demand — caption, attorney block, locked footer. The original problem? Solved. A complaint drafter with a taxonomy spanning personal injury, premises liability, wrongful death, civil rights, school district, and digital tort cases, anchoring each draft to firm exemplars. A demand letter pipeline that ingests records, strips protected health information and personal identifiers before anything reaches the model, drafts in firm voice against our verified citation registry, checks every cite against source documents, and files the finished product to the matter's folder. An exhibit packager that produces a Bates- stamped, tabbed exhibit binder mirroring the demand's enclosures list.
What Isn't Working Yet
One of the model’s weaknesses is its argumentative voice. Our firm has a particular cadence in plaintiff advocacy, and even with strong exemplar retrieval, the model drifts toward the mealy-mouthed register of every other AI-generated letter. The model hallucinates citations when it cannot find what it needs in the source documents. We built a cite- verifier that catches most of these, but there is still no
69
Consumer Attorneys of California
FORUM September/October 2026
Made with FlippingBook - Online catalogs