How does an insurer put insurance comparison and arrangement into its own AI?
Comparison and arrangement puts the whole sale inside an insurer's or a distributor's own AI assistant. The tools call a partner core system, from one free sentence to an issued contract. In emable's pilot run on 26 August 2026, the person was asked one question. Six required fields were seeded or extracted against a partner demo environment. The comparison returned 26 offers from 6 insurers at 136 to 1,136 CZK, in 26.4 seconds of machine time, and a contract was issued in that demo environment. Status is pilot and payment is not claimed.
Updated . Figures keep their period, source and reliability grade.
Key figures
| Metric | Value | Period | Source | Grade |
|---|---|---|---|---|
| Questions asked of the person | 1 questions | 26 August 2026 | emable pilot run record, partner core system demo environment | A |
| Required fields seeded or extracted | 6 fields | 26 August 2026 | emable pilot run record, partner core system demo environment | A |
| Offers compared | 26 offers | 26 August 2026 | emable pilot run record, partner core system demo environment | A |
| Insurers compared | 6 insurers, not named publicly | 26 August 2026 | emable pilot run record, partner core system demo environment | A |
| Offer price range | 136 to 1,136 CZK | 26 August 2026 | emable pilot run record, partner core system demo environment | A |
| Machine time across tool calls | 26.4 seconds | 26 August 2026 | emable pilot run record, partner core system demo environment | A |
The answer
On 26 August 2026, emable's pilot showed MCP tools comparing and arranging travel insurance inside a conversation in 26.4 seconds of machine time. An insurer, a broker pool or a bank puts these tools into its own AI assistant. The assistant compares and arranges insurance against the carrier core system behind it, from one free sentence to an issued contract. The person was asked exactly one question. Six required fields were seeded or extracted. The comparison returned 26 offers from 6 insurers at 136 to 1,136 CZK, and a contract was issued at one insurer. The run was a pilot against a partner core system demo environment. Status: pilot.
The numbers
Every figure behind this answer sits in the key figures table above, with its value, unit, period, source and reliability grade: A official statistics, B reconciled from published sources, C modelled estimate, D indicative.
How the number is built
The tools follow the order of the sale. Start from a free sentence, set client, calculate, select offer, arrange, transaction status. Each is one MCP tool call, and the AI client chains them inside the conversation. Machine time is the sum of those calls; it excludes the person's time to read and answer.
Six fields are the minimum the partner's calculation needs. In the run they were seeded or extracted, and the person was asked one question. The comparison returned 26 offers from 6 insurers. The AI client selected one, called arrange, and the demo environment issued a contract number.
What the run does not show. Payment is not claimed: the payment leg in the pilot runs beside the insurer's own collection and is not a production payment gateway. The contract was issued in the partner's demo environment, not in production. Insurers are not named on public pages. For property insurance the calculation and live offers were verified, and the arrangement was intentionally not completed.
The tools run through the Emable Finance Connector at https://api.emable.ai/mcp with OAuth. Every call is a run in the shared lifecycle: queued, running, succeeded, failed, refused, awaiting_approval. Trials need a login and no card.
What it means for your team
For a broker pool or an adviser network, comparison stays inside your own assistant and so does the adviser. Arrange is a separate tool call, so a human or your policy decides when a quote becomes a contract. The awaiting_approval state exists for that decision.
For an insurer or a core system vendor, we deliver this as software over your own engine. Your calculation engine does the pricing. Our tools ask for the fields in the right order, inside your assistant, your adviser desktop or your client app.
For an AI product lead, 26.4 seconds measures the tool chain, not the end client's experience. The end client read one question inside your assistant. Behind it sits one demo environment, one line of business verified end to end, and a pilot label that stays until production evidence exists.
Go deeper
- MCP connector docs: connect Claude, Codex or another AI client to the tools.
- Build with emable: REST, CLI and MCP on one run lifecycle.
- Insurance assistant coverage: sourced answers on the products being compared.
- What is embedded intelligence for financial distribution?: where the tools sit in the platform.
Related questions
- Was a payment made?
- No payment is claimed. The payment leg in the pilot runs beside the insurer's own collection and is not a production payment gateway.
- Which insurers were compared?
- Six insurers returned offers in the demo environment. Insurers are not named on public pages.
- Does it work for lines other than travel?
- For property insurance the calculation and live offers were verified and the arrangement was intentionally not completed. Other lines carry no public claim.
- How do we connect our own AI client?
- Through the Emable Finance Connector at https://api.emable.ai/mcp with OAuth. Every tool call is a run in the shared lifecycle: queued, running, succeeded, failed, refused, awaiting_approval. Trials need a login and no card.
Methodology and limitations
All figures come from one pilot run on 26 August 2026 against the partner's demo environment, recorded across the tool calls start from a free sentence, set client, calculate, select offer, arrange and transaction status. Machine time is the sum of those calls and excludes the person's time. The contract was issued in the demo environment, payment is not claimed and insurers are not named.