The fertility clinic operating system

AI belongs inside the record. Not bolted onto the outside of it.

Solux holds the funnel, the chart, the embryology lab and the ledger in one database. That is the only reason its AI can answer anything worth asking — and the reason you run more cycles on the team you already have.

IVF cycle · generated from one date Day 1 → Day 34

The AI wrote this cycle. A nurse used to. Protocol-aware, generated from one date, and the patient reads the identical schedule in her portal. This is what AI in fertility should look like — it removed a job nobody wanted, and it didn't need a chat window to do it.

Funnel, chart, lab, ledger
One database
Every AI question
Logged verbatim
Time to go live
60 days
Status
Live in a clinic today

AI, done the way it should be

Everyone in this market shipped a chatbot. Almost nobody shipped the part that matters.

A chat window is the easy half. The hard half is a system where the model can actually see the clinical record, is structurally prevented from seeing more than the person asking is allowed to see, and leaves a trail a regulator would accept. That isn't a feature you add later. It's a decision you make about the database on day one — and we made it.

AI bolted on
  • Sits on top of a system it cannot see into
  • Answers from a stale export or a help-centre article
  • Gives the same answer to a receptionist and a medical director
  • No record of what it was asked or what it returned
  • Confidently wrong, with nothing to check it against
  • Your patients quietly improve somebody else's model
AI inside the record
  • Reads the live clinical database the app itself runs on
  • Runs under the asking person's permissions — it cannot return a row they couldn't open themselves
  • Read-only, row-capped, time-limited. Blocked from writing anything, ever
  • Every question and every generated query written to an audit table
  • Predictions stored next to the actual outcome, so accuracy is measurable rather than asserted
  • Your data is never used to train a model. Configuration is prompt and knowledge, not fine-tuning

What actually happens

Someone asks a question. Here is the whole trace.

Not a black box, and not a demo trick. A question in plain language becomes a read-only query, scoped to the person asking, capped, timed, and written to a log alongside who asked and what came back.

AI you can audit is the only kind a clinic should run. If a vendor can't show you this screen, they haven't built it.

ai_query_audit — live
Asked by
Medical director · signed in
Question
"Which stimulating patients have AMH below 1.0 and no consent on file?"
Generated
SELECT display_name, amh, consent_status FROMWHERE status = 'stimulating' AND amh < 1.0
Scope
Asker's own permissions. Cannot reach a row she couldn't open herself
Guardrails
Read-only · row cap · query timeout · writes blocked
Returned
6 rows · 340 ms
Written to
ai_query_audit — user, clinic, query text, status, rows, duration
Every question your clinic asks leaves a record. That is the difference between AI you can deploy in healthcare and AI you can't.

Nine surfaces, not one chat window

The AI isn't a feature on the sidebar. It's load-bearing.

Most of the work it does, nobody has to ask for. It runs where the clinic was already losing hours — and where a mistake used to stay invisible until it wasn't.

01 · Clinical

Cycle calendars

The full protocol — birth control, stim, monitoring, trigger, retrieval, transfer — generated from a single date, and refused outright when scheduling rules or provider availability would make it wrong.

02 · Intake

Document extraction

Outside records, referrals and lab reports read on upload, fields extracted and mapped into the chart — flagged for a human when confidence drops. Never silently.

03 · Operations

Alert detection

Equipment maintenance, tank alarms, failed QC, expiring credentials, stock-outs, aged receivables. Found by the system, filed into one inbox with an owner and a resolution trail.

04 · Clinical

Readiness scoring

Every patient scored across clinical, laboratory, consent, financial and operational readiness — so a director opens four charts instead of sixty.

05 · Governance

Predictions, scored

Each prediction is stored with its inputs, its confidence and its model version — then the actual outcome is written back beside it. Accuracy becomes something you measure, not something we claim.

06 · Governance

Insights that wait

Population-level findings are proposed, not enacted. Each carries a confidence level and sits in review until a named clinician signs it. The human stays in the loop by construction.

07 · Embryology

Viability scoring

Time-lapse incubation data and AI viability scores attached to the individual embryo — in the same record as the grade, the PGT result and the straw position.

08 · Language

Bilingual by construction

Full English and Spanish across the app, the portal and the AI. Machine-critical fields are deliberately excluded from translation, so a language toggle can never change a medication name.

09 · Identity

Your assistant, not ours

Name, persona, tone and clinic knowledge configured during onboarding. It introduces itself as your clinic's assistant. And to be explicit: that is prompt and knowledge configuration, not a model trained on your patients.

Why one database matters

No AI can help you across a seam it can't see.

No fertility clinic loses a cycle because someone is bad at their job. Cycles are lost in the handoffs — where a record is carried from one system to another by a human who is already behind. Bolt an assistant onto any one of those systems and it is blind to the other three.

Seam 01

Lead → chart

Your CRM knows she enquired. Your EHR knows she exists. Neither knows she went quiet nine days ago and is about to book somewhere else.

Seam 02

Chart → lab

The clinical record says retrieval happened. The embryology record — the one that actually matters in five years — lives in a spreadsheet on someone's desktop.

Seam 03

Nurse → patient

The cycle calendar is built by hand, then retyped into an email. Two versions of the truth, and the patient is holding one of them.

Seam 04

Clinic → ledger

Month end is a reconciliation exercise between systems that were never meant to agree, run by the person you can least afford to lose.

What changes

Four people in the room. Four different versions of the same relief.

Owner / operator

"We go from 60 cycles a month to 120 without hiring three more coordinators — and I stop paying for four systems that don't talk to each other."
Scale without headcount

Medical director

"I confirm 60 stimulating patients are tracking correctly without opening 60 charts. The system tells me which four need me."
Cohort view, not chart-by-chart

Lab director

"Every straw is findable down to the physical slot, every handling step has a second name on it, and I stop keeping the real record in a spreadsheet."
The lab record becomes the record

Finance

"Money is collected before the cycle starts, and the clinical record and the ledger are the same record. There is nothing to reconcile."
One ledger, no month-end archaeology

What you're actually buying

You're not replacing software. You're replacing a stack.

Here is what a clinic normally buys, integrates and maintains separately. In Solux it is one system, one login, one record — and there are no integration fees between the parts, because there are no parts.

  1. 01Fertility EHRChart, trended labs, protocols, monitoring, encounter notes with sign-off, amendments and addenda.Enterprise EHR licence
  2. 02Embryology lab systemRetrieval outcomes, per-embryo grading, PGT, cryo inventory to tank / canister / cane / straw, transfers, thaw records, tissue labelling.Standalone lab system
  3. 03CRM and patient funnelEight-stage pipeline, timestamped stage changes, stale-lead detection, conversion reporting that is a query, not a spreadsheet.CRM + integration build
  4. 04Billing and revenue cyclePackages, fee schedules, CPT and ICD, invoices, payments, statements, AR aging, claims, remittances, denial analytics, eligibility, prior authorisation.PM / RCM system
  5. 05Patient portalThe same schedule the clinic sees. Documents, messaging, consents, education — in English or Spanish.Portal add-on
  6. 06The AI layer, all nine surfacesAssistant, cycle generation, document extraction, alert detection, readiness scoring, predictions, insights, viability scoring, bilingual output.BI tooling + an analyst
  7. 07Lab QC and equipmentRegister, calibration and PM due dates, alarms with an acknowledgement trail, routine QC with supervisory review, proficiency testing.QMS software
  8. 08Inventory with lot-to-case traceabilityAnswers both recall questions: which cases used this lot, and which lots touched this case.Inventory system
  9. 09Workforce complianceCredentials and competencies with expiry dates, and the report you assemble the night before an inspection.A spreadsheet and some risk
  10. 10Integrations already builtWitnessing, time-lapse incubation, PACS imaging, HL7 lab interfaces, payments, e-fax, label printing.An engineer, per interface
  11. 11Your own isolated instanceOwn database, own storage, own domain. Isolation is the architecture, not a filter on a shared table.Not sold anywhere else
  12. 12Tamper-evident audit chainEvery record change hash-chained, plus a technical due-diligence pack you can hand to your own counsel.Security consulting

Time to value

The honest answer everywhere else is six to twelve months.

Ours is weeks — because there is nothing left to build. Everything above already exists and already runs. The only work is typing in your prices, your protocols, your consents and your staff.

Traditional EHR installSolux
Instance stood up Months of tenant provisioning, contracts and sequencing Days — cloned and deployed from one versioned template
Configuration A bespoke consulting engagement, priced by the hour A 40-item checklist, screen by screen, run by us
Data migration A custom ETL project Staged import with a review step before anything lands in a chart
AI capability A roadmap item, or a chat widget over a knowledge base Nine surfaces live on day one, because they were built into the data model
Integrations Phase two, quoted separately, per interface Already built. Switched on when your credentials arrive

Why this won't hurt you

Every clinic has a graveyard of half-adopted systems. Here's why this one lands.

The real fear is never whether the software has the features. It's whether the vendor survives, whether the data survives, and whether the install takes the clinic down with it. Those are fair questions and they have concrete answers.

Proof 01

It runs a real clinic

Solux is the productised version of a system already running a working fertility clinic — not a prototype and not a pilot deck. The features you'll see exist because a clinic needed them on a Tuesday.

Proof 02

Your instance is physically yours

Separate database, separate storage, separate domain. There is no query anyone can write that reaches your patients from another clinic, because there is no shared table to write it against.

Proof 03

The audit log is tamper-evident

Every record change writes a hash-chained entry. A verification routine walks the chain and returns either OK or the exact row where it broke. Most systems tell you what happened; this one can prove nobody edited the answer.

Proof 04

Isolation is tested, not asserted

An automated probe counts how many rows from other clinics are visible to a signed-in user. Anything other than zero fails the build. A second probe blocks any view that could bypass row-level security.

Proof 05

Go-live is a gate, not a feeling

A live screen tracks 40 configuration items across identity, scheduling, clinical, embryology, billing, CRM, payments, lab QC, inventory and integrations — each marked blocker, important or optional. All blockers green, you go live.

Proof 06

Diligence is already written

There is a technical due-diligence document with every critical and high finding remediated and verified. Ask for it on the first call and hand it to your own security people before you sign anything.

The offer

The 60-Day Cycle Capacity Install

For self-pay fertility clinics scaling past 50 cycles a month. Your own isolated fertility EHR, CRM, embryology lab system and billing ledger — live in 60 days, on the headcount you have today.

  • Your own private instance, provisioned and deployed
  • All twelve modules and all nine AI surfaces on from day one
  • Structured data migration with a review step
  • Configuration run against the 40-item go-live checklist
  • Your clinic's assistant configured to your voice and knowledge
  • Staff training by role, not one long generic session
  • Integrations connected as your vendor credentials land
  • A named person who answers the phone after go-live

Go-live guarantee

If we miss the agreed go-live date for reasons on our side, the implementation is free.

Vendor continuity

A 12-month dedicated support commitment, written into the agreement rather than promised on a call.

Data exit guarantee

Full export in an open format, on demand, no fee and no notice period. Your patient data is yours, and is never used to train a model.

Pricing is tailored to your clinic and shared on the call.

The arithmetic

One recovered cycle a month pays for the year. Everything after that is margin.

Price this against a cycle, not against software. At a typical self-pay IVF price, a single cycle you would otherwise have lost — to a lead that went quiet, a calendar that slipped, a patient who wasn't financially cleared in time — covers the platform for months.

On the call we run this with your numbers: your monthly enquiries, your conversion rate, your cycle price. Not ours. If the arithmetic doesn't clear on your own figures, we'll tell you and you shouldn't buy it.

Next step

Bring your hardest question. Ask the system directly.

Forty-five minutes. We ask about your cycles, your systems and your handoffs first, then show only the parts that answer what you told us — and at some point we hand you the keyboard and let you ask the AI anything about a live clinic. That part is not scripted.