> ## Documentation Index
> Fetch the complete documentation index at: https://docs.runspark.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Underwriting Assistant
> Real-time carrier fit recommendations during live calls, grounded in actual carrier underwriting guides
The Underwriting Assistant listens to your live calls and tells you which carriers are likely to take the case — while the client is still on the phone. As the client shares their age, health history, and coverage goals, the assistant checks those facts against a library of rules built from real carrier underwriting guides and shows a fit status for each carrier and product.
It lives in the **Phone** tab as a panel alongside your live call. There's nothing to turn on — once the client starts sharing their situation, recommendations appear.
The Underwriting Assistant is **decision support, not an underwriting decision**. It never tells you a client "qualifies" — carriers make the final call, and you always have the final say on what to recommend. Treat every status as a starting point for the conversation, not a guarantee.
## How It Works
During a live call, the assistant reads the conversation transcript and extracts **only facts the client actually states**. It never guesses or infers — if the client hasn't said it, the assistant doesn't know it.
A deterministic rules engine — no AI guessing — compares the stated facts against a rule library built from real carrier underwriting guides. Spark maintains the library.
Each carrier/product shows one of four statuses, updated as new facts come up in conversation. Products whose maximum face amount is below what the client wants are filtered out entirely.
Every recommendation includes the reasons behind it, the questions you still need to ask, any red flags, and citations to the actual carrier guide — page number and excerpt — so you can verify the rule yourself.
## What It Extracts
The assistant picks up facts as the client states them:
| Fact | Examples |
| ----------------- | -------------------------------------------------------------------------------------------------------------------------------------------- |
| Age | Stated age or date of birth |
| Tobacco use | Smoker, non-smoker, cessation timeline |
| Build | Height and weight |
| State | Where the client lives |
| Coverage amount | The face amount they want |
| Health conditions | Type 2 diabetes, heart conditions, hypertension, cancer history, stroke, COPD, sleep apnea, kidney/liver disease, mental health, DUI history |
| Medications | Anything the client mentions taking |
## Status Meanings
| Status | What it means |
| ------------------ | --------------------------------------------------------------------------------------------- |
| **Likely fit** | The stated facts fall inside the carrier's acceptable band |
| **Rated** | The case fits, but expect a table rating or loading — includes a hint about the likely rating |
| **Needs info** | A key fact hasn't come up yet — the panel shows you which questions to ask |
| **Likely decline** | A stated fact falls outside the carrier's published criteria |
A "Likely decline" on one carrier often means a "Likely fit" on another — that's the point. The panel is there to help you place the case, not to disqualify the client.
## Build and Coverage Checks
* **Build/BMI** is checked against standard height/weight tables. If the client is slightly over a carrier's max, the assistant flags it rather than declining — small tolerances above the limit are worth a closer look, not an automatic no.
* **Coverage amount matters.** If the client wants $50,000 and a product caps at $25,000, that product is filtered out of the recommendations so you're only looking at carriers that can actually write the case.
## Getting the Most Out of It
The assistant only knows what the client says, so the quality of your recommendations tracks the quality of your questions.
* **Ask health questions naturally.** Work age, tobacco, build, conditions, and medications into the conversation — every fact the client states sharpens the recommendations.
* **Use the Needs Info list as your question checklist.** When a carrier shows **Needs info**, the panel tells you exactly which facts are missing. Ask those questions and watch the status resolve in real time.
* **Check the citations before you commit.** Each recommendation links to the carrier guide page and excerpt it's based on. A ten-second glance confirms the rule before you pitch the product.
The earlier in the call the client states their basics — age, tobacco, height/weight, desired coverage — the earlier you'll have a working shortlist of carriers, which means you can spend the rest of the call selling instead of fact-finding.
## FAQ
No. Fact extraction reads the transcript, but the fit decision itself comes from a deterministic rules engine — the same facts always produce the same statuses. There's no AI guessing at carrier criteria, and every rule traces back to a real carrier underwriting guide.
A key fact for that carrier's rules hasn't come up in conversation yet. The panel lists the specific questions to ask — once the client answers, the status resolves to Likely fit, Rated, or Likely decline.
From real carrier underwriting guides. Spark maintains the rule library, and every recommendation cites the guide it's based on — page number and excerpt included — so you can verify it yourself.
No — it means one stated fact falls outside that specific carrier's published criteria. Check the other carriers in the panel; a decline at one carrier is often a fit at another, and the carrier always makes the final decision anyway.
## Related
* [Power Dialer](/phone/dialer) — where the Underwriting Assistant panel lives during calls
* [Sales Log](/insurance/sales-log) — log the sale once you've placed the case
* [Commission Tracking](/insurance/commission-tracking) — see what the carrier you picked pays you