AI Plan Recommender
Ranked, explainable plan matching across every line you are contracted to sell. One engine, reachable three ways: natively in Solved Enroll, as a REST API, and as an MCP server.
One engine, three ways to use it
The ranking an agent sees in Solved Enroll is the ranking the API returns. There is no separate model for partners, and no drift between what your platform shows and what our agents see.
- Native in Solved Enroll. Ranked results appear between quoting and enrollment, so the agent recommends and enrolls without changing screens.
- REST API. Post a structured client profile, get ranked results with their factors, server to server.
- MCP server. Expose recommendation as a tool an agent runtime or copilot can call mid-conversation, with the same inputs and the same scoring.
- Built for platforms. For teams that need plan intelligence without rebuilding underwriting rules, formulary math, and Medicare election logic from scratch.
Solved Enroll is in private beta. The platform has a 2027 public rollout, but recommender licensing is available before general availability. Start a licensing conversation or join the waitlist for platform access.
What the recommender weighs
The inputs are the same things an experienced agent weighs at the kitchen table, made explicit so they can be scored, shown, and reviewed.
Life underwriting fit
Medications, conditions, height and weight, tobacco use, age, and face amount run against encoded carrier rules. Products the profile knocks out are removed, and the expected rate class is part of the score rather than a footnote.
Medicare plan fit
Election eligibility, the named providers checked against each plan's network, and the client's medication list priced across each formulary and coverage phase at their own pharmacy. Network and drug cost are the two factors that move a Medicare ranking most.
Budget and stated needs
What the client said they can spend each month, and what they said they care about: keeping a specialist, a dental allowance, a hospital stay they are afraid of, a face amount that covers a funeral. Stated priorities are inputs, not decoration.
Explainable ranking
Every result carries the factors that lifted or lowered it, in language you can read out loud. An explanation an agent can defend to a client is the same explanation a compliance reviewer needs.
Ranked, explained, and reproducible
A recommendation that cannot be explained is a liability, and a recommendation that cannot be reproduced is worse. Both are design constraints here, not features bolted on afterward.
- A score and a why. Each plan returns a score with the factors behind it, so the ranking survives the client's questions.
- Deterministic. The same profile, weights, and rule set version return the same ranking, so a case can be reconstructed in an audit.
- Versioned inputs. Each response carries the rule set and plan year it ran against, because carrier rules and plan data change mid-season.
- Tunable per agency. Weight drug cost, network, benefit richness, or premium differently, and the weights used come back with the response.
How a recommendation runs
Three steps, whether the call comes from an agent inside Solved Enroll, your own CRM, or an agent runtime holding an MCP session.
-
1
Pass the client profile
Health details for life, or election status, providers, pharmacy, and medications for Medicare, plus the budget and the priorities the client stated. From Solved Enroll this is the record the agent already built; from the API it is one JSON body.
-
2
The engine scores and filters
Underwriting rules, election rules, network checks, and formulary pricing remove what the client cannot get. What remains is scored against the weighting in effect, with each factor's contribution recorded.
-
3
Ranked plans come back
A short ranked list with scores, factors, and the rule set version. Inside Solved Enroll the agent moves straight into compare and enrollment. Through the API, your product renders the shortlist however it likes.
Access and availability
How each way of reaching the engine stands today, while the platform itself is in private beta.
| Access mode | Stage | What it is for |
|---|---|---|
| Native in Solved Enroll | Live in beta | Agents get ranked results between quoting and enrollment, with compare and signature in the same session. |
| REST API | Licensing open | Server-to-server recommendation for CRMs, dialers, agency platforms, and custom quoting interfaces. |
| MCP server | In build | Tool-style access for AI agents and copilots that need plan matching inside a conversation. |
| Agency-tuned weighting | In build | Per-agency weight profiles applied to every ranking, returned with each response. |
| White-label and volume licensing | By agreement | IMO, carrier, and platform licensing with dedicated onboarding, scoped by line of business and state. |
Recommender licensing is available before the platform's 2027 general availability release.
Licensing
License it for your stack
For platforms that need plan intelligence without rebuilding underwriting rules, formulary math, and Medicare election logic. Licensing is available now, ahead of the 2027 platform rollout.
HTTP API
Server-to-server recommendation for CRMs, dialers, agency platforms, and custom quoting interfaces. JSON in, ranked results with factors out.
MCP server
Recommendation exposed as a tool, so an agent runtime or copilot can call it in context and read back an explanation it can repeat to a client.
Native platform access
The same engine already wired into intake, quoting, compare, and enrollment for your own agents.
Explore the platform AgreementVolume and white-label
IMO, carrier, and platform licensing with dedicated onboarding, scoped by line of business, state, and volume.
Talk to usPOST https://api.solvedenroll.com/v1/recommendations
Authorization: Bearer $SOLVED_ENROLL_KEY
Content-Type: application/json
{
"line": "medicare",
"plan_year": 2027,
"client": {
"zip": "89117",
"birth_date": "1959-03-08",
"election_period": "AEP",
"providers": [{"npi": "1234567893"}],
"pharmacy": {"ncpdp": "1234567"},
"medications": [
{"name": "apixaban", "days_supply": 30},
{"name": "metformin", "days_supply": 90}
],
"priorities": ["keep_cardiologist", "dental"]
},
"weights": { "drug_cost": 0.40, "network": 0.35, "benefits": 0.25 }
}
200 OK
{
"request_id": "rec_01JQ8ZK4T3XV",
"ruleset_version": "2027.02",
"deterministic": true,
"weights_applied": { "drug_cost": 0.40, ... },
"results": [
{
"rank": 1,
"score": 0.91,
"plan_id": "H1234-005-000",
"plan_type": "MAPD",
"factors": [
{ "name": "network",
"detail": "both named providers in network" },
{ "name": "drug_cost",
"detail": "both drugs on formulary, no step therapy" },
{ "name": "benefits",
"detail": "dental allowance meets stated priority" }
]
},
{ "rank": 2, "score": 0.78, "plan_id": "H5678-011-000", ... }
]
}
Field names above show the shape of a request and a ranked response and are illustrative; the published contract lives in the API reference and the documentation. Credentials are not self-serve: keys are issued through a licensing conversation, because access is scoped by line of business, state, carrier data, and volume.
What the engine ranks
Medicare
Election rules, provider networks, and formulary-level drug cost.
ExploreLife
Encoded carrier underwriting rules and build-to-budget matching.
ExploreAncillary
Dental, vision, hearing, and indemnity attached to the same record.
ExploreAgency and IMO
Weighting, hierarchies, and reporting for the whole downline.
ExploreSee the full platform, the quote-to-enroll workflow, pricing, or join the waitlist.
FAQs
Questions about the recommender
What does the AI Plan Recommender actually do?
It takes a client profile and returns a ranked list of plans or products, each with a score and the factors that produced it. For life, the ranking runs against encoded carrier underwriting rules. For Medicare, it runs against plan data, provider networks, and the client's real drug list. It does not invent plans or premiums; it orders the options the client can actually get.
Is it ranked or filtered?
Both, in that order of importance. Options the client cannot qualify for or elect are removed, and everything that survives is ranked rather than dumped into a grid. The conversation starts from a recommendation with reasoning attached instead of from a spreadsheet you have to interpret in front of the client.
Can a case be reconstructed later?
Yes. The engine is deterministic: the same profile, the same weighting, and the same rule set version return the same ranking. Each response carries its rule set version and request identifier, so a recommendation made during AEP can be reproduced months later in an audit or a carrier review.
Can an agency change how the ranking is weighted?
Yes. Weighting is tunable per agency, so an agency that leads with drug cost and an agency that leads with network breadth both get rankings that match how they sell. The weights used are returned with the response, which is what keeps a tuned ranking explainable.
Can I license the engine without using Solved Enroll?
Yes. The same engine is available as a REST API and as an MCP server, so a CRM, dialer, quoting interface, or AI agent runtime can call it without rebuilding underwriting and Medicare logic. Licensing is separate from platform access and is available before general availability.
How do I get credentials?
Through a licensing conversation, not a self-serve signup. Access depends on the lines of business, states, carrier data, and volume involved, so keys are issued as part of the licensing agreement. Contact us to start that conversation, and see the API reference and documentation for the published request and response shapes.
Something else? Contact us
Put ranked plan matching in your product.
License the engine as a REST API or an MCP server, or run it natively inside Solved Enroll.