unthinkable
Health-tech · AI Product Engineering

Help hospitals deliver care without making patients wait.

For companies building software for hospitals

We help you build the product that makes it possible: AI that books, explains, summarises and follows up, with clinicians deciding what matters.

The hospital work a product can take on

~23%of appointments are missed: empty slots a waitlist feature can refillDantas et al., 2018, 105 studies
58%of insured US adults had trouble using their insurance in a year: questions a billing explainer can answerKFF, 2023
13 hoursa week of physician and staff time goes on 39 prior authorizations, the work a prior-auth product preparesAMA prior authorization physician survey, 2025
What we build

What we build for health-tech companies.

Five kinds of health product we build again and again. For each: what changes for the people who use it, what AI does (and what people still decide), and how we help you build it.

Corporate & consumer wellness

Employees use the health benefits they already have, without forms or phone calls.

BUILD AI IN

An assistant books consults and check-ups, explains what the plan covers, files reimbursements and follows up after a screening; clinicians handle anything clinical.

BUILD WITH AI

Built on your platform's benefits, claims and provider data. You own every line.

Shipped forVisit Health, Playing Forward, ASR Market Ventures

Records & imaging (EHR, PHR, PACS)

Clinicians and patients find the record or image they need without asking twice.

BUILD AI IN

Agents gather records from other systems, summarise the history for the clinician and tell the patient once a result is signed off. Nothing enters the chart without review.

BUILD WITH AI

HL7, FHIR and DICOM interfaces, tested against real message formats; every access logged.

Shipped forAuthentic4D, TeleRare Health, Guardian Health Services

Clinic & practice management

Patients book in a message and wait at home, not in the waiting room.

BUILD AI IN

Booking, reminders, waitlist offers, intake and delay alerts; visit notes drafted for the doctor to sign.

BUILD WITH AI

Connected to the clinic's records and payment systems; specs and tests keep each release safe.

Shipped forDope Doctors, Novadontics, Eye-Q Vision

Care coordination & remote monitoring

The right clinician knows in time, and families know what happens next.

BUILD AI IN

Readings and alerts routed to the right person, follow-ups chased, families sent status updates. Clinicians decide every clinical step.

BUILD WITH AI

Device and records integrations, alerting built to be reliable, an audit trail on every action.

Shipped forAliveCor, Mayamd, Elite Care Technologies, Mount Sinai

Revenue cycle management

Claims go out right the first time, and patients get a bill they understand.

BUILD AI IN

Charges checked against payer rules, denials sorted with the reason and the next step, bills explained to patients. Staff decide anything financial.

BUILD WITH AI

Payer rules run as code, not a model's guess; reads billing systems and insurer statements.

Shipped forSafer Health Solutions

Running a hospital or clinic? Hospitals & clinics →

Your customers' journey

The same patient journey, as your product sees it.

Seven steps where patients wait in a hospital. At each: what your product does, what it has to connect to, and what a hospital buyer will check before signing.

Today: a wait before every stepWith AI at work: straight through

1 · Booking

Your productBooks in a message, sends two-way reminders and offers each cancelled slot to the next eligible patient on the waitlist.

Connects toThe EHR's scheduling API: Epic, Oracle Health or athenahealth.

Hospital buyers checkThat the clinic's priority rules, in code, pick who gets a slot, never the model.

Clinic & practice management ↓

2 · Arriving

Your productRuns check-in as a chat the day before: payer eligibility and copay, history and medicine changes, then marks the visit pre-registered.

Connects toPayer eligibility APIs and the EHR's registration and arrival status.

Hospital buyers checkThat copay comes from the payer's response and no answer is filed until the care team confirms it.

Clinic & practice management ↓

3 · Seeing the doctor

Your productRecalculates appointment times from the live schedule, messages every patient affected, adds their priorities to the pre-visit summary and drafts the note.

Connects toThe clinic schedule and the EHR's notes, through FHIR where the EHR supports it.

Hospital buyers checkThat nothing is filed in the chart until the doctor signs.

Clinic & practice management ↓

4 · Tests and results

Your productChecks the payer's approval rules for the scan, books it with the patient, and sends a message when the signed report is filed.

Connects toLab and imaging systems over HL7 v2 and FHIR, with DICOM for the images.

Hospital buyers checkThat the product never reads out or interprets a result; a clinician does.

Records & imaging (EHR, PHR, PACS) ↓

5 · Going home

Your productTracks what each discharge is waiting on, chases pharmacy and transport, and sends the family the pick-up time.

Connects toDischarge orders and pharmacy status in Epic or Oracle Health, and the transport booking service.

Hospital buyers checkThat only a physician's order discharges a patient, and the hospital's checklist is never changed by the model.

Care coordination & remote monitoring ↓

6 · The bill

Your productMatches every bill line to the explanation of benefits, explains it in plain words, and offers payment plans or assistance by policy.

Connects toThe billing system, EOB and remittance data, and clearinghouses.

Hospital buyers checkThat amounts come from the billing system, never the model, and only a person adjusts a balance.

Revenue cycle management ↓

7 · Emergency

Your productGives verified family contacts live status from orders, results and the imaging queue, status only.

Connects toThe ED tracking board and the authorized contacts on the patient's record.

Hospital buyers checkThat no result or clinical opinion is shared; those come from the care team.

Care coordination & remote monitoring ↓

What the product does at each moment (text version)

Booking

  • Slot cancelled (From the scheduling system) → Pick from the waitlist (Clinic priority rules, insurance and referral checks) → Talk to the patient (Understands free-text replies) → Rebook and release (EHR scheduling API) → Clinical priority stays with staff (Urgent cases are ordered by clinicians). Guardrail: If the patient describes new or worsening symptoms, the agent stops and gives urgent-care instructions.

Arriving

  • Day before the visit (Scheduled message) → Eligibility check (Payer response, copay by rule) → Collect intake in plain words (History and medication changes) → Care team confirms changes (Nothing enters the chart unreviewed) → Pre-register (EHR arrival status). Guardrail: Red-flag symptoms mentioned during intake end the chat with emergency instructions and alert the clinic.

Seeing the doctor

  • Clinic runs behind (From the live schedule) → Recalculate times (Scheduling rules, per patient) → Talk to the patient (New time, and what they want to discuss) → Update the visit (Priorities added to the pre-visit summary) → Doctor leads the visit (Clinical questions stay with the doctor). Guardrail: If the patient describes urgent symptoms, the agent stops and gives urgent-care instructions.

Tests and results

  • Scan ordered (From the visit) → Check approval rules (Payer requirements by scan type) → Book with the patient (Slots, preparation, questions) → Report filed (Radiology) → Care team shares the result (A clinician explains it, never the agent). Guardrail: The agent never reads out or interprets a result; a clinician does.

Going home

  • Doctor signs the discharge (Medical fitness is a clinical decision) → Track what's left (Medicines, transport, paperwork) → Talk to patient and family (Times, pick-up, questions) → Nurse hands over (Medicines explained in person). Guardrail: The agent never tells a patient they can go home; only a physician's discharge order does.

The bill

  • Match bill to EOB (Line by line, flags mismatches) → Explain in plain words (From the matched figures only) → Offer options by policy (Plans and assistance eligibility) → Counselor decides (Adjustments, waivers, charity care). Guardrail: The agent never waives or reduces a balance; any adjustment is a person's decision.

Emergency

  • Family member asks (Verified contact on the patient's record) → Read status (Orders, results, imaging queue) → Explain next steps (Status only, no results) → Clinicians share results (In person). Guardrail: The agent never shares test results or clinical opinions; those come from the care team.

Try it with your own case →

Why products stall

Why health products stall, and how we help you avoid it.

BUILD AI IN
“Payer rules keep changing.”
Prior authorization criteria and claim edits differ by payer and plan. We keep them as versioned rules in code, not in prompts, so every change is reviewed, tested and traceable.
“A payer or auditor won't accept it.”
Every request and appeal cites the exact note, lab or letter it relies on. A clinician attests and a biller approves before anything is sent, and each step is logged.
“The AI will get clinical details wrong.”
Rules and code decide thresholds and eligibility. Drafts are scored with your clinicians for omissions and errors, rerun on every model or prompt change, and nothing is filed until a clinician signs.
BUILD WITH AI
“EHR integration will take forever.”
We build on FHIR R4 and SMART on FHIR where the EHR supports it, HL7 v2 where it does not, and test against Epic and Oracle Health sandboxes before any live data.
“HIPAA rules out language models.”
Patient data flows only to services covered by a BAA, including the model provider, and stays in US infrastructure. We map every hop, minimise what reaches the model and log each access.
“We can't hire a health-tech team fast enough.”
A senior lead and a complete team from week one, with building blocks already proven in production: an integration layer, review queues, audit logs and evaluation sets.
You inherit

Building blocks already built.

Already built
FHIR and HL7 integration layerin production
PHI redaction and minimisationin production
Clinician review and sign-off queuein production
Audit log of every agent actionin production
Evaluation harness with clinician-graded test setsin production
Role-based access for clinical and admin staffin production
Your health-tech productstarts on these, not from scratch

The team around it

Product designerShapes how people and agents work together
AI engineerAgents, prompts, evaluations
ArchitectRecords, integrations, security
DevelopersBuild it with AI, keep it maintainable
TestersSpecs and tests that prove it works
Domain expertKnows your industry's rules and edge cases
Security reviews

Will your product pass a hospital's security review?

Hospital buyers send security and compliance questionnaires before any pilot. These are the answers we build in from the first release.

BUILD AI IN
An audit trail a reviewer can replay
Every agent action is logged with the record it read, the rule that applied and the person who approved it.
Clinical output stays a draft
Notes, summaries and appeals wait for a clinician's or biller's sign-off, and the clinician-graded test sets are rerun on every model or prompt change.
Access by role
Clinical and admin users see only what their role allows, and each view is logged.
BUILD WITH AI
A BAA for every hop
Each service that touches PHI, the model provider included, sits under a business associate agreement; only the minimum data reaches the model.
Evidence, not a badge
No official HIPAA certification exists, so we document the data flows, controls and agreements your buyer's reviewers ask for.
Tested before live data
Built against Epic and Oracle Health sandboxes first, with the hospital's integration approval planned from the start.
Integrations

The integrations hospital buyers ask for.

A hospital will ask what your product connects to before it asks about features. These are the connections we build.

Epic and Oracle Health

FHIR R4 and SMART on FHIR launch where the EHR supports it, HL7 v2 feeds where it does not; athenahealth and other EHRs through their APIs.

Payers and clearinghouses

Eligibility checks, payer portals and claims through clearinghouses such as Office Ally, with payer rules versioned in code.

Lab, imaging and exchange

HL7 links to lab analysers, DICOM to PACS, and Redox for pulling records from other EHRs.

Devices and wearables

Apple Watch, Wear OS, HealthKit and vitals trackers, as in our remote-monitoring work.

You own the result: code, prompts, evaluation sets and data sit in your accounts from the first line, so your product's IP stays yours.

Proof

Health products we have shipped, and what they did.

Mount Sinai, New York47%faster STEMI response

A real-time coordination app: ECGs shared from the ambulance, GPS tracking, the team alerted at once. Heart-attack (STEMI) patients needed the cardiac team ready before they arrived.

Read the case study →
A chain of clinics across the UK47%better patient engagement and efficiency, 30+ clinics live

A multi-clinic system for booking, records, referrals, lab and imaging, HL7-compliant, re-using about 30% of the old code. After an earlier outsourcing attempt failed, its clinics still needed one practice-management system.

Read the case study →
The largest US network of eye-care clinics60%higher productivity for eye-care teams, 60+ clinics

A custom EHR for six roles: scheduling, call centre, billing, insurance claims and a patient portal, HIPAA-compliant. Its third-party EHR no longer fitted the patient and admin cycle, and cost more with every new clinic.

Read the case study →
A US non-profit digital health organisation3managed care plans billed from one platform (Kaiser, Health Net, PHC)

One platform for intake, care and billing: care plans drafted by AI for care managers to review, pre-billing checks for each payer, claims sent through Office Ally. Care and billing ran on spreadsheets, and each managed care plan's rules caused high claim rejections.

Read the case study →
Elite Care, US transitional careUp to 50%faster feature releases

A transitional care platform that pulls records from the EHR through Redox, with alerts for tasks due and care notes, HIPAA-compliant. Patients leaving hospital were readmitted because hospitals and primary care weren't coordinating.

Read the case study →
AliveCor, California30+heart abnormalities the app reports, up from 5

Rebuilt Android and iOS apps with AliveCor's own algorithm integrated, plus a health diary fed by Apple Watch and Wear OS. Its EKG app needed more detailed reports, modern code and remote monitoring fit for FDA and HIPAA requirements.

Read the case study →

Results as reported in our client case studies.

Which part of care would your product take on first? Name the moment where patients or clinicians wait most: booking, results, follow-up or billing. After a short discovery you get a plan with dates for the first release, measured against that wait.

Tap one: Ask Unthinkable starts your plan from there. Or say it in your own words ↓

How to choose

How to choose who builds health-tech products.

Four things a health-tech founder or CTO should check, and how we meet them.

Has shipped health products

Wellness, records, clinic, care-coordination and billing products, for companies such as AliveCor, Visit Health and Mayamd, and for Mount Sinai.

Often elsewhereStrong engineers who are new to clinical workflows and health data.

Knows the hospital side

We also help hospitals and clinics run day to day, so we know the workflows, approvals and reviews your product has to fit.

Often elsewhereHealthcare-only shops know the rules; few have built AI that acts.

Agents inside the product

In clinic software, for example: booking, reminders, waitlist offers, intake and delay alerts, with visit notes drafted for the doctor to sign.

Often elsewhereA chat widget bolted onto the product.

Feature twenty as fast as feature one

We set the architecture before the prototype and keep AI writing inside it; payer rules stay versioned in code, so a rule change is a tested release.

Often elsewhereA fast demo that stalls once integrations and compliance arrive.

The full comparison: us, other software companies, in-house teams and consultants →

Ways to work

Start the way that suits you.

Time & materials

A senior AI team billed for time, within an agreed monthly budget. Even on time and materials the work is tied to your business objective: every plan starts from it, and every report shows the number next to the hours.

One fixed outcome

One agreed outcome at a fixed price, with acceptance criteria, including the measure, written up front.

Dedicated team

A complete senior AI team, monthly, working as part of yours and reviewed each quarter on what shipped and what moved.

Linked to the result

Where the gain can be measured from your own systems: a reduced implementation fee plus a percentage of the expected gain, for a set period.

How we work and price →

Straight answers

Questions health-tech teams ask.

How do I choose a development partner for a health-tech product?

Judge them on what your hospital buyers will judge you on. Ask for an EHR integration they have shipped, over FHIR, HL7 v2 or a service such as Redox, and how they test it before it touches live data. Ask how a clinical draft stays a draft until a clinician signs, and how they rerun their evaluation sets when a model or prompt changes. Ask what evidence they can hand a hospital's security reviewer, from the audit trail to the agreements for each service that touches patient data, and whether the architecture will still let you release quickly once integrations and compliance pile up. Finally, check that code, prompts and data sit in your own accounts. Unthinkable has shipped health products such as Mount Sinai's STEMI coordination app, a practice-management system for a UK clinic chain and AliveCor's EKG apps, and it also helps hospitals and clinics run, so it knows the questions your buyers will ask.

I am a doctor with a product idea. Can you build it without me hiring a tech team?

Yes. You bring the clinical problem; we bring the complete product team: product designer, AI engineer, architect, developers, testers and a senior lead. We start with a short paid discovery that turns your idea into a scoped first release, a clickable design and a fixed price. You own the code, IP and data from the first line, and we stay until the product works in production with real clinicians.

Can you work with Epic?

Yes. We integrate through Epic's FHIR APIs and SMART on FHIR launch, and use HL7 v2 interfaces where a hospital's setup needs them. We build and test in Epic's sandbox first, then work with the hospital's integration team for production access. Timelines depend on the health system's own review process, so we plan for that from the start rather than treating it as a final step.

Is it HIPAA compliant to use large language models with patient data?

It can be, if every service that touches PHI signs a BAA and the data flow is controlled. We use model providers that offer a BAA, send only the minimum data needed, redact where possible, keep logs of every access and host within your cloud account when required. HIPAA compliance is a property of the whole system and your processes, so we document the chain for your compliance review.

How do you make sure a clinical AI feature does not give wrong answers?

We do not rely on the model alone. Rules and code handle anything with a fixed answer, such as thresholds or eligibility. The model drafts, cites the record it used and states its confidence, and a clinician approves anything that affects care. Before release we test against cases your clinicians write and grade, and we rerun those tests every time the model or prompts change.

How much does it cost to build a healthcare AI product?

It depends mostly on integrations and compliance scope: EHR integration and clinical evaluation are often the largest items, not the AI itself. A first release is usually one fixed outcome, and you get a scoped estimate after a short discovery. Unthinkable reuses tested building blocks for FHIR, audit and review queues, so your budget goes to what is specific to your product.

Can you build a prior authorization or revenue cycle product?

Yes. These are good fits because the work mixes judgement and rules. An agent gathers chart evidence and drafts the request; payer rules, code checks and submission logic run as code; a coordinator approves before anything is sent. Every step is logged so you can show payers and auditors what was submitted and why. We integrate with EHRs, clearinghouses and payer portals where APIs allow.

Can our product meet GDPR and UK rules as well as HIPAA?

Yes, if it is designed for each market from the start. For the UK and Europe that means a lawful basis for processing health data, hosting in the right region, data processing agreements with every provider and a data protection impact assessment before launch. Unthinkable designs the data flows so one product can meet HIPAA and GDPR together, and documents them for your compliance team and data protection officer.

Ask Unthinkable

Which wait will your product remove first?

Tell us about your product and the moment you want to change. Our AI assistant, Ask Unthinkable, helps you define your requirements, then connects you with a senior AI engineer.

Send us a note instead

Required

Thanks. A senior AI engineer will reply to you.
u
Ask UnthinkableClarity first, then the right team
Your conversation

An AI agent. This prototype uses prepared answers and sends nothing.