Life
AMI Life (Protection) Agent Standard
Version: 0.2 (Draft for Comment)
Status: Working Draft
Steward: Marrow (onmarrow.com), on behalf of the Agent Mediated Insurance (AMI) working group
Line of business: UK individual life and health protection
Published: June 2026
This is a draft request for comment. It defines how an autonomous AI agent should quote, disclose, apply for and complete a UK individual protection policy against a regulated carrier, safely and auditably. Life is the most regulated and consequential line in this set, so this standard is deliberately the most conservative. Comments to standards@onmarrow.com.
Foundational dependency. This standard relies on the AMI Core Patterns Standard for all shared behavioural rules (grounding and no fabrication, assume-infer-confirm, the advice boundary, disclose-and-acknowledge, fair presentation, honest absence, data protection, vulnerability and Consumer Duty, handover machinery, and audit) and the canonical roles, and on the AMI Consent, Authority and Scopes Standard for authentication and transaction authority. Only the rules distinctive to life protection are stated in full below.
Scope (UK-first). v0.2 covers UK individual pure protection: level and decreasing term, whole-of-life protection, critical illness, and income protection, sold in a personal capacity. It excludes investment-linked, with-profits, savings and pension products, and any business protection, trust, or estate-planning arrangement, all of which carry advice and complexity beyond this version. Where in doubt, the agent treats the case as out of scope and hands over.
What is distinctive about life protection
Three things make life the hardest line, and shape this standard.
Life sets a particularly high bar for duty of care. A protection sale stands or falls on the accuracy of the customer's health and lifestyle answers. Under the Consumer Insurance (Disclosure and Representations) Act 2012, the customer's duty is to take reasonable care not to make a misrepresentation when answering the carrier's questions. A careless or deliberate misrepresentation can lead to a claim being reduced or refused and the policy avoided, at the worst possible moment, on a death or critical-illness claim. The agent's central job is to put the carrier's questions clearly, help the customer answer them fully and accurately, explain the consequences of getting them wrong, and capture an explicit attestation. This is the travel medical-screening problem, with far higher stakes and a statutory frame.
Advice stays with a human. Protection is frequently an advised sale, and the needs behind it (who depends on this person, for how long, against what) are exactly the kind of judgement that suitability rules exist to protect. In v0.2 the agent operates on a non-advised basis only, with a recorded demands-and-needs statement under ICOBS, and must not give a personal recommendation or assess suitability itself. Where advice is needed or asked for, it hands over to a human adviser.
Underwriting is often not straight-through. Unlike motor or home, a protection application is not an instant bind. A healthy applicant may be accepted at standard rates automatically, but many cases are rated, carry an exclusion, are postponed, are declined, or require medical evidence (a GP report, a nurse tele-interview, or a medical examination). The agent assembles and submits an application; the carrier underwrites it; an offer follows; the customer accepts. The agent must represent every outcome honestly and route the non-standard ones to the carrier's underwriting process and to a human where required.
Sources and provenance
- The Consumer Insurance (Disclosure and Representations) Act 2012 (CIDRA) sets the consumer's duty to take reasonable care not to misrepresent, and the carrier's remedies for careless or deliberate misrepresentation. It is the statutory basis for the disclosure rule.
- ICOBS (the FCA Insurance Conduct of Business Sourcebook) governs non-investment protection, including the demands-and-needs requirement and the advised-versus-non-advised distinction. FCA Consumer Duty applies throughout.
- Carrier and reinsurer underwriting practice establishes the health and lifestyle question set (medical history, BMI from height and weight, smoker and nicotine status, family history, alcohol, occupation, hazardous pursuits) and the evidence path (GP report, nurse tele-interview, medical examination report including blood pressure, urinalysis and a cotinine test).
- UK protection insurer and adviser question sets corroborate the application fields and the underwriting-decision taxonomy (standard, rated, exclusion, postponed, declined, evidence required).
1. Design principles
The shared design principles (carrier-agnostic by construction; no fabrication; assume-infer-confirm; disclose-and-acknowledge; observable compliance) are defined in the AMI Core Patterns Standard and apply in full. The principles distinctive to life:
Disclosure care under CIDRA. The agent presents the carrier's questions clearly, helps the customer take reasonable care to answer fully and accurately, and explains plainly that careless or deliberate misrepresentation can lead to a claim being reduced or refused and the policy avoided. Health and lifestyle answers are special-category data and are processed only with explicit consent.
Non-advised only, in v0.2. The agent provides information and facilitation, records a demands-and-needs statement, and does not give a personal recommendation or assess suitability. Advice is a human responsibility.
Apply, underwrite, offer, accept. A protection sale is a process, not an instant bind. The agent must represent rated, excluded, postponed, declined and evidence-required outcomes honestly, and never imply standard acceptance before underwriting has reached it.
2. Roles and terms
Roles and terms (Customer, Agent, Surface, Rail, Carrier, Material fact) are defined in the AMI Core Patterns Standard. Life-specific terms: Life assured, the person whose life or health is covered (there may be two on a joint policy); Applicant, the person applying and answering the questions (usually the life assured); Beneficiary, who receives the benefit; Underwriting decision, the carrier's terms for the risk.
3. Canonical life application request
The agent assembles a life_application_request. An indicative premium can be produced from a small subset (the indicative fields); a firm application requires the full set. Fields are marked required, enrichable (the rail can populate; the customer confirms), consented (special-category, only with explicit consent), or conditional.
3.1 Lives assured (one or two)
| Field | Type | Notes |
|---|---|---|
date_of_birth | required, indicative | Primary rating factor. |
gender | required, indicative | |
smoker_status | required, indicative | Nicotine use in the defined lookback, including cigarettes, cigars, pipe, vapes and nicotine replacement. A primary rating factor and a common misrepresentation risk. |
height / weight | required, indicative | Used for BMI. |
residency | required | UK residence; non-UK residence is commonly out of appetite. |
occupation | required | Risk classification; some occupations are rated or excluded. |
earnings | conditional | Required for income protection and for financial justification of the sum assured. |
3.2 Cover
| Field | Required | Notes |
|---|---|---|
product_type | required | Level term, decreasing term, whole-of-life protection, critical illness, income protection. |
sum_assured / benefit_amount | required | Lump sum, or monthly benefit for income protection. |
term | conditional | For term products. |
premium_basis | required | Guaranteed or reviewable. Material to the customer and frequently misunderstood. |
indexation | optional | Whether benefit and premium increase over time. |
additional_benefits | optional | Canonical list: waiver of premium, total permanent disability, terminal illness, critical illness rider, children's cover. |
basis_of_cover | conditional | Single life, joint life first death, or joint life second death. |
3.3 Health and lifestyle (special-category, consented)
| Field | Band | Notes |
|---|---|---|
medical_history | consented | Past and present conditions, investigations, treatment and medication. Answered to the carrier's specific questions, not as open narrative. |
family_history | consented | Serious illness in immediate family, in particular before a defined age. |
alcohol_use | consented | |
hazardous_pursuits | consented | Activities affecting mortality or morbidity risk. |
gp_details | consented | For a GP report where the carrier requires evidence. |
3.4 Suitability context and declarations
| Field | Required | Notes |
|---|---|---|
demands_and_needs | required | The recorded statement of what the customer is trying to achieve, captured non-advised. |
existing_cover_and_replacement | required | Whether this replaces existing protection; replacement triggers handover (Section 6). |
in_trust / beneficiaries | conditional | Trust arrangements are out of v0.2 scope and route to handover. |
disclosure_attestation | required | The customer's explicit attestation, after the consequences of misrepresentation have been explained, that the answers are accurate and complete to the best of their knowledge. |
4. Enrichment and evidence
Enrichment for life is thin and consent-led. The substance of the risk comes from the customer's disclosure and from carrier-side underwriting evidence, not from databases.
| Hook | Source | Populates / does | Basis |
|---|---|---|---|
occupation_pursuit_classification | Industry-standard occupation and pursuit reference data | Maps occupation and activities to the carrier's risk classes. | Reference data |
medical_evidence | The customer's GP, a nurse tele-interview, or a medical examination, each with explicit consent | Provides the carrier's underwriting evidence where the application is not straight-through. Carrier-side and human-administered; the agent arranges and consents, it does not adjudicate. | Consented medical evidence |
There is deliberately no central health-data lookup. The standard does not imply one, and relies on disclosure and carrier underwriting.
5. Agent behavioural rules (compliance semantics)
The general behavioural rules (no fabrication, information-not-advice, disclose-and-acknowledge, data protection, fair presentation, honest absence) are defined in Core Patterns Section 3 and apply in full. Life-specific rules:
- Conduct disclosure with care, under CIDRA. The agent presents the carrier's questions clearly, does not lead or abbreviate them, explains that careless or deliberate misrepresentation can lead to a claim being reduced or refused and the policy avoided, and captures the explicit disclosure attestation before any application is submitted.
- Special-category data. Health, family-history and lifestyle data is special-category and is processed only with explicit, recorded consent, and protected by the rail beyond the general PII rules.
- Non-advised only. The agent records a demands-and-needs statement and does not give a personal recommendation, rank products as suitable for the customer, or assess suitability. If the customer wants advice, the agent hands over.
- Represent underwriting outcomes honestly. The agent does not imply standard acceptance before underwriting reaches it, and presents rated, excluded, postponed, declined and evidence-required outcomes plainly, including what each means for the customer.
- Premium-basis clarity. The agent makes explicit whether the premium is guaranteed or reviewable, because the difference is material and frequently misunderstood.
6. Handover triggers
The shared handover machinery and common triggers are defined in Core Patterns Section 3.9. Life sets a deliberately low bar. Life-specific triggers, additional to the common ones:
- The customer asks for advice, or the case calls for a personal recommendation.
- The application is replacing existing protection.
- The underwriting outcome is rated, excluded, postponed, declined, or requires evidence (GP report, tele-interview, medical examination).
- The need is complex or outside scope: trusts, business protection, estate or tax planning, joint-life nuances, or a sum assured above the carrier's financial-underwriting or anti-money-laundering threshold.
- Health, family history or circumstances the canonical questions cannot represent.
- Any vulnerability indicator, which is treated with heightened sensitivity given the subject matter.
7. Audit requirements
Audit requirements are defined in Core Patterns Section 3.10 and apply in full. Life-specific evidentiary emphasis: the exact questions put to the customer and their recorded answers, the explanation of the consequences of misrepresentation and the customer's disclosure attestation, every consent for special-category data and for medical evidence, the demands-and-needs statement, and the underwriting decision and any evidence relied on. This record is the basis on which a disputed claim or a misrepresentation question can later be assessed fairly.
8. Versioning and conformance
This standard is versioned (v0.2) and is UK-first. It relies on the AMI Core Patterns Standard and the AMI Consent, Authority and Scopes Standard; conformance requires conformance to those at a declared version. A conformant agent implements Sections 3 through 7 in full, including the disclosure-care rule and the non-advised-only and handover requirements. A conformant carrier maps its protection products to the canonical application and response and exposes its underwriting decision and evidence requirements. Later versions may extend to advised sales, trusts and other jurisdictions; none are in v0.2.
9. Open questions for comment
- How a non-advised demands-and-needs statement should be represented and evidenced for an agent-conducted sale, and where the line to advice sits in practice.
- Whether, and how, an agent should ever conduct an advised protection sale, and what permission and human oversight that would require.
- The canonical health and lifestyle question model: whether the standard mandates a common question set or defers to each carrier's questions, and how answers map for straight-through underwriting.
- How the medical-evidence path (GP report, tele-interview, examination) is represented, consented and tracked through an agent mediated application.
- Financial-underwriting and anti-money-laundering thresholds above which an application must always involve a human.
Comments and signatories: standards@onmarrow.com.
Published as an open standard. Marrow stewards the drafting but does not assert proprietary control over the interface; the compliance, enrichment, and audit implementations behind it are separate. Carriers, agent platforms, and regulators are invited to adopt, critique, and co-author future versions.