A person decides the match. Always.

Matching is the part of your agency that cannot be rushed and cannot be guessed. So we do not automate the decision. We remove the work of assembling it, and hand your coordinator a recommendation that already shows its reasoning.

Book a 30-minute call How the gate works

What it replaces

Not judgement. The hours before the judgement.

TodayHow it goesA coordinator opens both files, reads them end to end, cross-references states and timelines, checks the clinic, remembers what the carrier said on a call in March, and forms a view.2 to 4 HOURS A PAIR

With the engineHow it goesThe comparison is already assembled, ranked and annotated. The coordinator reads a brief, disagrees with it where they disagree, and decides.MINUTES, THEN JUDGEMENT

The hours figure is an estimate, not a measurement. The audit measures yours, and if it turns out to be twenty minutes we will tell you the engine is not worth building.

The matching engine

It recommends. It never decides.

We build the algorithm and the agentic system that reads what you already hold: the carrier’s application and profile, the intended parents’ profile, and the history around both. It analyses the pair, ranks the candidates, and hands your coordinator a recommendation that arrives with its reasoning attached.

Not a score with no explanation. A named set of signals, what agreed, what did not, and what somebody should ask about before the introduction call.

What it reasons over

State and legal fitBoth parties in workable jurisdictions, counsel already engaged

Timeline alignmentTransfer readiness within the same window, no six-month gap

Clinic proximityMonitoring reachable without weekly air travel

Expectations on commsContact frequency and channel preferences compatible

Journey historyPrior carries, outcomes, and what the carrier says she wants next

Compensation fitExpectations and budget inside the same range

Your criteria, not ours. The signals and their weighting are yours to set, and you sign them off before anything runs.

Match recommendationAwaiting coordinator review

CARRIER · TX · 2 prior, both term

PARENTS · CA · clinic in Dallas

Why this pair

+ Both jurisdictions workable, counsel engaged on each side

+ Transfer windows overlap in the same quarter

+ Clinic is a 40 minute drive, not a flight

+ Both asked for weekly contact, same channel

Ask before the intro call

− Carrier has not carried for out-of-state parents before

− One medication needs clinic sign-off

Decision: a person. This screen cannot approve a match.

Illustrative record. Names and identifiers are blacked out because in a real one they never leave your systems.

Why it never decides

The screen cannot approve a match.

This is a product constraint, not a policy we could quietly relax later. There is no auto-advance, no silent approval path, and no configuration that turns one on.

A match is two families and a pregnancy. The cost of getting it wrong is not a bad quarter, and no model should carry that.

You approve the signals and their weighting in writing before anything runs.
Every recommendation is logged with the inputs that produced it, so a decision can be reconstructed months later.
Nothing is auto-rejected. A low-ranked pair still reaches a person.
We review the distribution with you for patterns nobody intended.

Before you ask

Three questions we always get.

Will it match better than my coordinator?

Wrong question, and we would not claim it. It assembles faster and forgets nothing. Your coordinator brings the judgement, and now brings it to a complete picture instead of a half-read file.

What if we disagree with a recommendation?

Then you are using it correctly. Disagreement is logged too, and it is the most useful signal we get for tuning the weighting.

Whose data trains it?

Yours, for you. Nothing built for you is reused for anyone else, and nothing is trained on your data for use elsewhere.

Book a 30-minute call

Matching is stage five of six. The audit tells you whether it is your bottleneck.