Protecting the self-employed: a business of volume and of bespoke, at the same time.
Lilycare is a digital broker specializing in social protection for the self-employed: health insurance, income protection, retirement review, and retirement savings. A promise that sounds simple — finding, for each profile, the best-suited policy among dozens of partner insurers — but one that hides considerable operational complexity once you look under the hood.
A Lilycare client is never just a pricing file. It's a self-employed person or a small-business owner, with their own line of work, a specific social status, sometimes an associated company; it's a need that evolves with their career — from a first health plan as a freelancer through to a retirement review at the end of their working life; and it's a policy that, once signed, has to be tracked, renewed, documented, with its commission checked month after month against the insurer.
The challenge: beyond volume, a classification problem.
Before Standards, every income-protection quote required an advisor to log in, profile by profile, to the extranets of insurers as different from one another as April, ACPS, Ilona, Malakoff, MetLife, Welfaire, or Abeille — each with its own flow, its own fields, its own pricing rules.
But the difficulty isn't just the number of insurers: none of them speaks the same language. Every insurer splits professions according to its own nomenclature — some list around 300, others close to 1,000 — and the same profession can be classified, and therefore priced, completely differently from one insurer to the next. Add to that each pension fund's own rules and every policy parameter, and you get a sizeable decision tree to work through just to know whether a client is insurable, on which product, and under what conditions — before even comparing a single price. This is not a simple job-to-offer mapping exercise.
Once contracts were signed, Lilycare still had to cross-check, insurer by insurer, PDF commission statements against the real book of business to verify it was paid the right amount — a tedious reconciliation job, redone every month, on formats that have nothing in common from one insurer to the next.
« The operational burden of cross-checking commission statements every month was such that it ate into the energy we should have been putting toward our real goal: finding the best protection for every self-employed person. »
A network of business agents, one per insurer, one per task.
With Standards, Lilycare first untangled its operations from a patchwork of software tools, at best loosely connected, at worst duplicating each other (Pipedrive, Sellsy, Brevo, Notion). It then built a network of specialized agents, each modeled on a specific insurer or a specific task, all connected to the same data graph (contacts, companies, deals, documents). Rather than one generalist AI asked to do everything: narrow agents, each an expert in a single extranet or a single document type, and verifiable one by one.
These agents step in whenever the online enrollment journey can't find a price through the connected insurers' APIs. Data from API-connected online journeys, from AI agents, and from client portals all converge into Standards.
The AI agents can:
- Interpret the client's actual profession, map it to the nomenclature of each of the thirty partner insurers, determine which combinations are insurable, then run live pricing simulations to produce the required quotes.
- Compare the quotes obtained for the same profile using a weighted score (price, coverage, simplicity, etc.), to prepare the advisor for their presentation to the prospect.
- Score each lead by conversion probability and case complexity, and switch automatically between a self-serve journey and manual handling by an advisor as soon as the case calls for it.
- Read insurers' PDF commission statements across every different format, match the right contract in the database behind each line, handle splitting rules (monthly, quarterly, half-yearly, annual), and produce a structured reconciliation report ready to check in a few minutes.
- Attach every quote, every statement, every supporting document directly to the relevant contact or deal record, in the right place in the drive.
- Check the consistency of a data point before saving anything — an IBAN, an amount, an expiry date — and pause to ask for human validation whenever a field is ambiguous or sensitive.
- Log every exchange and every action in the notes tied to the contact or deal, so an advisor picking up a file gets the full context immediately.
- Generate targeted exports from strict, consistently applied business rules — marketing audiences, for instance, where every field follows a mapping validated once rather than improvised on each request.
What used to force an advisor to juggle ten browser tabs — and take days to arbitrate — is now a ready-to-use comparison file.
Documentary compliance, without re-entry.
A self-employed income-protection or health policy can't be signed without a complete file: a valid ID, proof of address, bank details, URSSAF affiliation certificate, CPAM social security certificate, latest tax notice, a signed entry-into-relationship document. Each of these documents has its own date, its own review status, and its own renewal deadline.
In Standards, every document carries its metadata directly on the contact's record — status (to validate, validated, rejected, expired), validity date, expiry date — with no separate re-entry into a tracking spreadsheet. An advisor opening a client's record sees at a glance what's missing or coming due, instead of reopening a physical file or an email attachment to check a date.
Making the career talk, not just the policy.
Standards goes beyond pricing and compliance: it helps carry out the retirement review. It gives agents the ability to read a career statement (CNAV, AGIRC-ARRCO) to automatically detect anomalies, estimate lost entitlements, and score how relevant a retirement-review case is.
Lilycare's path is that of a broker that grew fast, adding insurer partners and products, never able to afford bolting on a new piece of software for every new need. The answer wasn't to stack up tools, but to build, extranet by extranet and document by document, a family of narrow, verifiable agents while keeping humans as the decision-makers to stay compliant.
The result isn't a technology gadget: it's time given back to advisors, time they can spend on what Lilycare exists for — protecting the self-employed, who have no one else in France to do it for them!