Prompt Pilotby Write.law
All prompts

Workflow

Build a claim chronology for coverage litigation

Turns a claim file into a dated, sourced chronology where every entry carries an evidentiary tag (proved, testimony, inferred, disputed, or gap) so you can see which dates you can actually establish and which ones you only believe.

About 25 minintermediateInsurance, Litigation

Your prompt5,180 characters

Still to fill in: Claim file documents, Policy periods on the risk, Coverage issues the chronology must serve, Governing law and forum

RoleYou are a coverage litigator who builds the chronology before writing anything, because in coverage cases the dispositive facts are almost always dates: when the damage occurred, when the insured knew, when notice went out, when the policy incepted. You cite the source for every entry, you refuse to merge two events into one line to make the story read better, and you would rather show a gap than fill one.What I needBuild a claim chronology from the file below for a coverage dispute under Governing law and forum. It will be used for Summary judgment briefing.InputsClaim file documents: Claim file documents Coverage issues it must serve: Coverage issues the chronology must serve Testimony bearing on dates: Testimony that fixes or contradicts dates Policy periods on the risk: Policy periods on the risk Governing law and forum: Governing law and forum Use: Summary judgment briefingHow to work this1. Lay the anchor rows first: every inception and expiration in Policy periods on the risk, any retroactive date, any extended reporting period. These are rows, not headings; every other date is read against them. 2. One event per row. Never merge events to make the narrative cleaner, and never move one to where it reads better. 3. Tag every row with exactly one of: [PROVED - document], [TESTIMONY - witness], [INFERRED - state the inference], [DISPUTED - both versions cited], [GAP - no record]. An untagged row is a defect. 4. Keep date precision honest. If the source says "November 2025" or "right after Thanksgiving," the date column says that. Never convert an imprecise date to a day; mark precision in its own column. 5. Assign every row to a track: loss, the insured's knowledge, notice and claim handling, or policy period. 6. Where two sources give different dates for one event, do not reconcile them. Enter both rows, cite both, and repeat the conflict in a date-disputes section. 7. List every gap: an interval with no entry, an event with no document, a date resting only on inference, and name the document or custodian that would close it. 8. Take each issue in Coverage issues the chronology must serve in turn and name the rows that decide it. Where the consequence of a date depends on Governing law and forum law (notice and prejudice, trigger, allocation), mark it [UNVERIFIED - confirm the rule] rather than state it.Close with these four sections, every time, without being askedAssumptions I made. Every factual and evidentiary assumption behind the table, marked [verify] or [safe]. Say so if you assumed a document is the earliest of its kind, that one date approximates another, or that two policy years share a form. Where this is weakest. The two or three rows tagged more strongly than the source supports, or inferences that read like facts. Name the row. What only you can decide. Options with tradeoffs, never a bare flag. At minimum: build this as work product, which keeps the inferences but cannot go to a mediator, or as a produced exhibit, which freezes your date positions and can impeach your own witness; show both versions of a disputed date, honest and survives cross but concedes the other side's date is arguable, or adopt your client's version, stronger on paper but an impeachment exhibit; and close the gaps in discovery now or leave them and argue the other side never asked. What would make this materially better. Ranked by impact: the custodian file, declarations page, or invoice that would fix the most contested date.Output formatA table with columns for date, precision, event stated neutrally, track, source with page or bates cite, and tag. Then date disputes, then gaps with the document that would close each, then a consequences section tying each coverage issue to its deciding rows. Then the four closers.Never do this- If the chronology would fit any claim in any state, it is too generic. Every row cites a document or transcript from my inputs. - No hedging filler. Cut "arguably," "it should be noted," and "shortly thereafter" used where a date belongs. Do not tell me to consult an attorney. I am the attorney. - Never supply a date, bates number, document, or witness statement that is not in my inputs, and never convert "around" or "sometime after" into a day. Anything uncertain is marked [UNVERIFIED - check the source]. - Where you do not know whether an interval is late under Governing law and forum law, say you do not know rather than smooth over it. - Do not pad. If the file supports nine entries, the chronology has nine rows. Length is not value. - Never reconcile conflicting dates into one row, and never write a narrative in place of the table.Before you answer- Does every row carry a source and exactly one evidentiary tag, and do the policy inception and expiration dates appear as rows? - Did I preserve every imprecise date as imprecise? - Are both versions of every conflicting date in the table? - Would this chronology fit a different claim file? It should not.

Adds driver's-seat tunes: options instead of answers, questions before work, every citation flagged. Your values come with it.

2

Pressure-test it

Makes the AI switch hats and attack its own answer.

Produce this chronology and opposing counsel gets it: the same documents, plus your exhibit, plus every tag you assigned. Read it back as the lawyer who intends to use your own entries against you. Which three entries are tagged more strongly than the source supports? Which inference would you make the preparer defend on cross? Which gap helps the other side more than it helps mine? Then re-tag the entries that cannot survive that reading.
3

Go deeper

Pushes the work further once the basics are right.

Which dates can you actually prove, and which are only asserted? Turn the gaps and date disputes into a discovery plan: the document requests and custodians that would close each gap, a deposition outline for the one witness who can fix the most contested date, and a short note on which gaps are worth closing given what each date actually decides.

Before you run it

What to gather first

  • The full claim file (correspondence, log notes, reports, emails, invoices) with page or bates references
  • Every policy period potentially on the risk, with inception and expiration dates
  • Which coverage issues the chronology has to serve: late notice, trigger, prior knowledge, exhaustion
  • Deposition or statement testimony that fixes or contradicts a date
  • What the chronology is for: brief, mediation, deposition, or internal analysis

Watch for

  • The model will smooth a chronology into a story if you let it. Check every [INFERRED] tag against the underlying document before you rely on an entry.
  • Whether late notice defeats coverage, and whether the carrier must show prejudice, varies by state and by whether the policy is occurrence or claims-made. Never take the consequence of a date from this output without confirming the rule.
  • A chronology you produce becomes an admission of your date positions and can be used to impeach your own witnesses. Decide early whether this is work product or an exhibit.
  • Imprecise dates are where cases are lost. "Late summer" and "around Thanksgiving" must stay imprecise in the table, not be converted to a day.
  • Do not paste an unproduced claim file or deposition testimony under a protective order into any tool your firm's AI policy does not cover.

What comes back

A chronology table with columns for date, date precision, event stated neutrally, track, source with page or bates reference, and evidentiary tag. Policy inceptions, expirations, and retroactive dates appear as their own rows. Then a date-disputes section listing every event with two competing dates, both sources shown and unreconciled. Then a gaps section naming each interval or event with no record and what document would close it. Then a consequences section that takes each coverage issue in turn and names the specific rows that decide it, plus what is missing to decide it now. Then the four closing sections.

See an example of what you’ll get
Chronology: Sabine Millwork LLC v. Gulf States Indemnity Co. | Date | Precision | Event | Track | Source | Tag | |---|---|---|---|---|---| | 7/1/2023 | Exact | GSI CGL policy year 1 incepts | Policy | Dec page GSI-000004 | [PROVED - document] | | 7/1/2024 | Exact | Policy year 2 incepts; renewal application signed by Prentiss 6/12/2024 answering "no" to known losses | Policy / Knowledge | GSI-000019, GSI-000027 | [PROVED - document] | | Late summer 2025 | Imprecise | Prentiss first observes ceiling staining | Knowledge | Prentiss dep. 47:12-19 | [TESTIMONY - witness] | | 11/18/2025 | Exact | Prentiss emails property manager: "the ceiling is staining again" | Knowledge | SAB-00407 | [PROVED - document] | | "Right after Thanksgiving" 2025 | Imprecise | Duplessis says he reported the condition to the property manager | Knowledge / Notice | Duplessis recorded stmt. 3/9/2026 | [TESTIMONY - witness] | | 11/2025 through 2/2/2026 | Range | No entry of any kind in the claim file | Notice | none | [GAP - no record] | | 2/2/2026 | Exact | GSI logs first notice of loss | Notice | GSI-000112 | [PROVED - document] | | 2/18/2026 | Exact | GSI engineer inspects; report attributes intrusion to flashing installed 2023 | Loss | GSI-000341 | [PROVED - document] | | 2023 construction | Imprecise | Flashing installed | Loss | Engineer report GSI-000341 at 6 | [INFERRED - the report says "original construction," and the certificate of occupancy is dated 9/2023; the installation date itself is not documented] | | 3/4/2026 | Exact | Sworn proof of loss submitted | Notice | GSI-000388 | [PROVED - document] | Date disputes. 1. *When the insured first knew.* Prentiss says late summer 2025 (dep. 47:12-19). The email at SAB-00407 uses the word "again," implying an earlier occurrence that no document captures. Both entries stand; neither is reconciled. 2. *When notice was given.* Duplessis places a report "right after Thanksgiving" 2025 to the property manager. The carrier's first log entry is 2/2/2026. Whether a report to a property manager is notice to the carrier is a legal question, not a date question. Gaps. (1) Nothing in the file between 11/18/2025 and 2/2/2026. The property manager's file, not yet obtained, would close it. (2) No document fixes the flashing installation date. (3) No writing corroborates the Duplessis report; the property manager's phone log and email would. What each date decides. - *Late notice.* The interval that matters runs from the first knowledge date to 2/2/2026, between roughly 76 days and 5 months, depending on which knowledge date holds. Whether that interval defeats coverage, and whether prejudice is required, is [UNVERIFIED - confirm the Louisiana rule for occurrence policies]. - *Trigger.* If damage began at original construction in 2023 and continued, both policy years are implicated. The chronology cannot resolve the allocation; the engineer's opinion on the onset date can. - *Prior knowledge at renewal.* Turns entirely on whether "again" in SAB-00407 refers to something before 6/12/2024. Nothing in the file answers that. --- Assumptions I made. That GSI-000112 is the earliest claim-file entry and no pre-FNOL agent contact exists [verify - request the agent's file]. That the certificate of occupancy date approximates the flashing installation [verify - this is an inference, not a fact]. That both policy years are on the same form [verify - only year 1's declarations were provided]. Where this is weakest. The 2023 installation row. It is tagged [INFERRED] and reads like a fact; it rests on "original construction" plus a certificate of occupancy. If the trigger analysis leans on it, it will not hold. The "again" inference in SAB-00407 is the second weakest: it is suggestive and proves nothing. What only you can decide. Whether to build this as work product or as a produced exhibit. Work product lets you keep the inferences and the argument, but you cannot hand it to the mediator; a produced chronology freezes your date positions and can impeach Prentiss on cross. Also yours: show both versions of the notice date, which is honest and survives cross but concedes the carrier's date is arguable, or adopt Duplessis's date, which is stronger on paper and hands the carrier an impeachment exhibit. And whether to depose the property manager now to close the November gap, or leave it and argue the carrier never asked. What would make this materially better. (1) The property manager's file and phone log: it decides the notice date. (2) The year 2 declarations page. (3) Any repair or maintenance invoice from 2024 that would show whether "again" refers to something before renewal.
Why this prompt is built the way it is
## Framework 1. **Anchor rows first.** Every policy inception and expiration, any retroactive date, any extended reporting period. Every other date is read against these. 2. **One event, one row.** Never merge events to smooth the narrative. 3. **Tag every row.** [PROVED - document], [TESTIMONY - witness], [INFERRED - reasoning stated], [DISPUTED - both versions cited], [GAP - no record]. An untagged row is a defect. 4. **Keep date precision honest.** "November 2025" stays "November 2025." Never pick a day. 5. **Four tracks.** Loss or occurrence, the insured's knowledge, notice and claim handling, policy periods. Mark each row. 6. **Do not reconcile conflicts.** Two documents, two rows, both cited, then a date-disputes list. 7. **Tie dates to consequences.** For each coverage issue, name the rows that decide it and what is missing to decide it now.