
AI Prompts and Templates for Personal Injury Teams
Why one-off AI prompts give PI firms uneven results, and how saved prompts and firm templates for chronologies and demands fix it.
When five people at a firm each type their own request into a chatbot, they get five different chronologies. One lists visits by provider, one by date, one skips billing, and one invents a header nobody asked for. The attorney then spends review time reformatting instead of checking facts.
The fix is to decide the format once and reuse it. This post covers which PI tasks benefit most from a saved prompt or template, what a good one contains, and how to keep review in the loop.
Why one-off prompts give uneven output
- Each person words the request differently, and small wording changes change the output.
- A general chatbot has no fixed output format, so tables, headings and level of detail drift between runs.
- Without the source file loaded, the tool fills gaps from general knowledge, which is how invented facts get into drafts.
PI tasks worth a saved prompt or template
Medical chronology
Fix the columns once: date of service, provider and facility, complaints, findings, treatment or plan, and a citation to the page. Decide whether billing goes in the same table or a separate one.
Treatment summary by provider
Group by provider, list first and last visit dates, total visits, diagnoses and the current treatment status.
Issue review
Ask the same questions on every file: prior conditions, gaps in treatment, inconsistent histories, and statements tying symptoms to the incident. Each answer should cite the page.
Demand letter first draft
Start from the firm's own demand structure: facts of the incident, liability, injuries and treatment, specials, and the demand section the attorney writes. The template should leave the valuation and the number to the attorney.
Deposition transcript summary
A page-and-line summary by topic is a common custom prompt. In LawUnleashed this is something a firm sets up as its own saved prompt.
What a good PI prompt includes
- The output format, written out (columns, headings, order).
- An instruction to cite the document and page for every fact.
- An instruction to say "not found in the file" instead of guessing.
- What to leave for the attorney, such as valuation or legal conclusions.
How this works in LawUnleashed
LawUnleashed includes premium PI templates and lets a firm create its own templates and saved prompts and share them with the team. Prompts run against the documents uploaded to the project, and generated text carries citations back to the document and page, which open beside the draft for checking. Drafts export to Word.
Review still applies
A template makes output consistent; the facts in it still need checking. ABA Formal Opinion 512 (July 29, 2024) says lawyers using generative AI must understand the tool's limits, review its output, and supervise staff who use it. A fixed format makes that review faster because the reviewer knows where everything is.
Start a free trial with one user and projects of up to 1,000 pages, no credit card.
Sources
- ABA Formal Opinion 512, Generative Artificial Intelligence Tools (July 29, 2024): https://www.americanbar.org/content/dam/aba/administrative/professional_responsibility/ethics-opinions/aba-formal-opinion-512.pdf
- LawUnleashed templates, prompts and pricing: https://www.lawunleashed.ai/
Related: What AI can and can't do for a PI firm and Medical record review for PI firms.
