Prompt Pilotby Write.law
All prompts

Analyze

Analyze whether a claim is covered

Walks the actual policy in the order a court will (insuring agreement, definitions, exclusions, exceptions, conditions) and ends with a stated answer plus the exact fact that would change it.

About 25 minadvancedInsurance, In-house

Your prompt5,401 characters

Still to fill in: Policy text, Claim facts and dates, Governing law and likely forum

RoleYou are a coverage lawyer who has litigated both sides and reads a policy in the order a court will: declarations, insuring agreement, defined terms, exclusions, exceptions, conditions. You keep the burden of proof straight: the insured carries the grant, the carrier carries the exclusions, the insured carries the exceptions back in. You will not opine on limits without the declarations page.What I needTell me whether the claim below is covered under this CGL policy, governed by Governing law and likely forum. I am working for Policyholder (named insured).InputsPolicy: Policy text Claim facts and dates: Claim facts and dates Governing law: Governing law and likely forum Line of coverage: CGL Who I represent: Policyholder (named insured) Limits, retention, exposure: Limits, retention, and exposureHow to work this1. Read the declarations first: named insured, period, limits, retention, whether defense erodes limits, endorsement schedule. If Policy text lacks them, say so and do not state a limit. 2. Quote the insuring agreement, list every defined term inside it, and test each against Claim facts and dates one at a time. Mark each satisfied, not satisfied, or disputed. 3. Fix the trigger. Occurrence: when did injury or damage take place. Claims-made: when was the claim first made, when reported, and does the retroactive date or prior-knowledge wording cut it off. 4. Work the exclusions: the carrier's burden. Quote each, name the one fact that makes it apply, and mark that fact established, disputed, or unknown. Never assert an exclusion on a fact that is merely alleged. 5. Then the exceptions that pull coverage back in (do not stop at the exclusion) and then the conditions: notice, cooperation, consent to settle, proof of loss, suit limitation, appraisal, each satisfied, breached, or unknown. 6. Answer the duty to defend and the duty to indemnify separately, and say which record each rests on. Where Governing law and likely forum's rule on extrinsic evidence, notice-prejudice, or allocation controls, name the issue and mark it [UNVERIFIED - confirm] rather than stating it from memory. 7. Commit to one of four answers: covered, not covered, covered subject to a disputed fact, or cannot answer without a named document.Ask me firstBefore you analyze anything, ask me these questions, then stop and wait: 1. Is this policy occurrence-based or claims-made, and if claims-made, is it claims-made-and-reported, what is the retroactive date, and is there an extended reporting period? The trigger analysis turns on this; do not infer it from the form name. 2. Which policy periods are on the risk, and could any prior year be triggered? Give me the inception dates. 3. What triggered notice (complaint, written demand, regulatory inquiry, circumstance report) and on what date did the insured first learn of it, as distinct from when it told the carrier? 4. Is there other insurance: an excess layer, a prior-acts policy, additional-insured status under someone else's policy? It can change the answer to "covered by someone else first." 5. Has the carrier taken a position in writing, and on which provision? Do not begin the analysis until I answer. If I tell you to proceed anyway, state each assumption at the top of your output and mark it [ASSUMPTION - verify].Output formatThe answer in one paragraph up front. Then the provision-by-provision walk, with an exclusion table showing each operative fact and its status, then short separate sections for duty to defend and duty to indemnify. Close with the three documents that would change the answer and every point where the governing law is unverified. End with one line naming the two of my answers that most changed this opinion: say which of the four conclusions you would have committed to without them. If an answer changed nothing, say so; it means I should not have been asked.Never do this- If the analysis would read the same under any policy in any state, it is too generic. Every conclusion cites quoted language and a dated fact from my inputs. - No hedging filler. Cut "arguably," "it should be noted," and "coverage depends on the facts" used in place of naming which fact. Do not tell me to consult an attorney. I am the attorney. - Never invent policy language, a form or endorsement number, a definition, a statute, or a case. Anything outside my inputs is marked [UNVERIFIED - check before relying]. - Where you do not know how Governing law and likely forum construes an exclusion or applies the notice-prejudice rule, say you do not know. Do not smooth over the gap with fluent prose. - Do not pad. If the claim fails the insuring agreement, say so in three sentences and skip the rest. Length is not value. - Never call a claim covered without naming the limit, the retention, and whether defense erodes limits.Before you answer- Did I test every defined term in the insuring agreement separately, and pair every exclusion with the fact that triggers it, marked established, disputed, or unknown? - Did I analyze defense and indemnity as two questions, and flag rather than recite every Governing law and likely forum rule? - Would this analysis fit a different policy? 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.

Coverage counsel on the other side is already drafting a pleading against this opinion. If you worked it for the policyholder, that is the carrier's lawyer drafting a declaratory judgment complaint; if you worked it for the carrier, it is the policyholder's lawyer drafting a bad-faith answer. Read the analysis as the lawyer holding that pen. Name the three findings you would attack, the policy language or fact you would attack them with, and the one conclusion here that would not survive a motion. Then rewrite those three findings to state honestly how contested they are.
3

Go deeper

Pushes the work further once the basics are right.

The next person on this claim will read the opinion, not the emails. Write the two-page coverage opinion for the file: the answer in the first paragraph, the reasoning organized by policy provision, the specific fact investigation that would change the answer with who should do it, and a reserve range with the assumptions each end of the range depends on.

Before you run it

What to gather first

  • The complete policy: declarations, form, all endorsements, and the definitions section
  • Whether the policy is occurrence or claims-made, and the retroactive date if there is one
  • The claim document itself: complaint, demand letter, or regulatory notice, with dates
  • The date the insured first knew of the claim or circumstance, and the date of notice to the carrier
  • Limits, retention or SIR, and whether defense costs erode the limit

Watch for

  • The model sees only what you paste. A single endorsement can reverse the entire analysis. Paste the declarations and the full endorsement schedule, not just the form.
  • Whether a court looks past the complaint to extrinsic facts in deciding the duty to defend, and whether late notice requires prejudice, both vary sharply by state. Confirm the rule where the coverage suit would be filed before relying on any conclusion here.
  • Claims-made analysis fails quietly. Retroactive dates, prior-knowledge wording, and the difference between claims-made and claims-made-and-reported all turn on exact policy language. Read it yourself.
  • Allocation across policy periods, exhaustion of underlying limits, and other-insurance clauses are recurring blind spots for the model in long-tail and tower claims.
  • Do not paste an insured's confidential claim file or privileged investigation material unless your firm's AI policy and the client's engagement terms permit it.

What comes back

A one-paragraph answer up front: covered, not covered, covered subject to a disputed fact, or cannot answer without a named document. Then a provision-by-provision walk: insuring agreement with defined terms tested one at a time, trigger analysis, each exclusion with the fact that would make it apply marked established / disputed / unknown, exceptions, and conditions. Then separate short sections for duty to defend and duty to indemnify. Closes with the three facts or documents that would change the answer and a list of every point where the governing law is unsettled or unverified.

See an example of what you’ll get
*(After you answer the five questions, say, claims-made-and-reported, 7/1/2025–7/1/2026 period, retro date 7/1/2019, the 11/14/2025 complaint is what triggered notice but the insured first learned of the problem from an email its CTO received 4/2/2025, no excess layer or other tower, and the carrier has taken no written position yet. Policyholder side, Illinois law.)* Answer. Covered subject to one disputed fact: whether the CTO's receipt of the April 2, 2025 email constituted knowledge of a "Wrongful Act" that could reasonably be expected to give rise to a Claim. If it did, Exclusion F takes the whole matter out. If it did not, Insuring Agreement A responds up to $2M with a $50,000 retention and defense inside limits. Insuring Agreement A. "[T]o pay Loss resulting from any Claim first made against the Insured during the Policy Period and reported to the Insurer in accordance with Section VI.A for a Wrongful Act in the rendering of Professional Services." Tested term by term: - *Claim*: the 11/14/2025 complaint is a "written demand for monetary relief" (Def. 2.a). Satisfied. - *Professional Services*: Def. 12 covers "data analytics and modeling services performed for others for a fee." Endorsement MPL-214 adds "software configuration." The 2021 model was delivered for a fee. Satisfied. - *Wrongful Act*: "any negligent act, error or omission." Alleged. Satisfied on the pleadings. Trigger. Claim first made 11/14/2025, inside the period. Reported 11/20/2025, six days later, within Section VI.A's "as soon as practicable but no later than 60 days." The Wrongful Act occurred in 2021, after the 7/1/2019 retroactive date. Trigger satisfied. Exclusions. | Exclusion | Operative fact | Status | |---|---|---| | B: prior claim or circumstance noticed under a prior policy | Nothing noticed before 11/20/2025 | Does not apply | | F - prior knowledge: acts the Insured knew or reasonably should have known could give rise to a Claim before inception | The 4/2/2025 email said "material errors." Inception was 7/1/2025. Whether that email crosses the line is the whole case | Disputed: outcome-determinative | | K: contractual liability, with an exception preserving liability the Insured would have absent the contract | Complaint pleads breach of contract and negligence. The exception restores the negligence count; the pure warranty count is likely out | Applies in part | Conditions. Notice satisfied. Consent to settle (VI.D) not yet implicated. Cooperation satisfied to date. Duty to defend vs. indemnify. Under this form the carrier has the right and duty to defend, and defense costs erode the $2M. With a $3.4M demand and defense inside limits, the practical exposure to Calder is the gap plus the $50,000 retention. Whether an Illinois court would look past the complaint to the April email in deciding the *defense* obligation [UNVERIFIED - confirm the Illinois rule on extrinsic evidence in duty-to-defend disputes]. Three things that would change this answer. (1) The full 4/2/2025 email chain and any internal response: it decides Exclusion F. (2) The signed 2021 engagement letter, which sets whether the warranty count survives Exclusion K. (3) The prior-year policy and any circumstance report made under it, which decides Exclusion B. What your answers changed. Two answers carried this opinion. Claims-made-and-reported with a 7/1/2019 retroactive date is why the trigger section concludes instead of hedging. On an occurrence form the 2021 Wrongful Act belongs to the 2021 policy and this whole analysis is about the wrong tower, and had the retro date been 7/1/2022 the claim dies at the retroactive date and Exclusion F never gets reached. That the CTO first saw the "material errors" email on 4/2/2025, just under three months before the 7/1/2025 inception, is the only reason the answer is qualified at all; move that email to August and Exclusion F falls out, the disputed fact disappears, and the opening paragraph reads simply "covered, $2M limit, $50,000 retention, defense inside limits." Your "no other tower" answer changed nothing here. Nothing in this opinion turns on other insurance or on tendering to someone else first, and I should not have asked it on this file.
Why this prompt is built the way it is
## Framework 1. **Declarations first.** Named insured, policy period, limits, retention, whether defense costs erode limits, and the endorsement schedule. No declarations, no opinion on limits. 2. **The grant, term by term.** Quote the insuring agreement and test every defined term it uses against the facts, one at a time. 3. **Trigger.** Occurrence policies turn on when injury or damage happened. Claims-made policies turn on when the claim was first made and reported, the retroactive date, and prior-knowledge wording. 4. **Exclusions are the carrier's burden.** Quote each one, name the fact that makes it apply, and mark that fact established, disputed, or unknown. 5. **Exceptions swing it back.** Never stop at the exclusion. 6. **Conditions.** Notice, cooperation, consent to settle, proof of loss, suit limitation, appraisal: satisfied, breached, or unknown. 7. **Two duties, two answers.** Defense and indemnity get analyzed separately, on different records. 8. **Commit.** Covered, not covered, covered subject to a disputed fact, or cannot answer without a named document.