01
Paid discovery
A short, paid discovery to agree on the first use case, the data flows, the BAA map, and the delivery budget. It ends with a go or no-go decision.
We put a secure foundation in place and get the first AI system into production with a team the company keeps. The code, data, and know-how stay with the company, which is what a buyer’s diligence team will check at exit.
Healthcare services companies, recently acquired, merged, or in the middle of a roll-up, arrive with a document-heavy workflow that still runs on spreadsheets, fax inboxes, and manual re-keying into the EHR or the billing system. A managed service provider runs IT. There is no engineering team. Leadership has a model on the shortlist and wants the company to own what gets built.
They need people to build the system and a clear answer on which business associate agreement covers each part.
Payer rules, contracts, rates, and workflows change every quarter. When a vendor builds the system and leaves, nobody inside the company can change it.
At exit, the buyer’s diligence team will ask who owns the code, the data, and the operating knowledge, and who can keep the system running.
Under HIPAA, the company keeps its duties for patient data, and the risk that comes with them, even when someone else writes the code.
Hiring an AI engineering team before the first system exists is more risk than most lower middle market companies can take on. The managed service provider runs IT. The company still needs someone to own the AI build.
Anthropic, OpenAI, AWS, and Microsoft all offer BAA-covered routes, but coverage varies by product, platform, and feature, and vendor documentation does not always agree. We confirm coverage for the exact route in writing before any patient data moves, and we develop on synthetic or properly de-identified data, so contractors never touch production PHI.
A short, paid discovery to agree on the first use case, the data flows, the BAA map, and the delivery budget. It ends with a go or no-go decision.
The security and compliance plan covers the reference architecture, the platform, data flows, and integrations, and the vendor and BAA decisions. When on-premises is the right call, we size the compute and inference hardware to the measured workload before anyone buys it, then source it and install it at the company’s site.
We lead the build with one or two contract engineers, or manage a long-term build-and-operate partner under the company’s ownership. Anything that touches patients or billing gets human review.
The company controls the accounts, data, code, and documentation throughout. The handover to the team it keeps, whether new hires or converted contractors, is planned when the work begins.
For operating partners: the second portfolio company starts from the first one’s playbook and reference architecture.
At Elevance Health, Christian Busch led product for the AI/ML platform built on FHIR R4 lifetime patient records for 70 million members, on a zero-trust architecture across AWS and GCP. The prior authorization engine he built there combined OCR document intake, FHIR data, and guideline-derived decision trees, and cut decisions from days to seconds.
At a U.S. medical-device manufacturer, he built the cybersecurity and IT compliance program, with the roadmap and budget for ISO 27001, HITRUST, and NIST 800-53, that carried the company through health-system security reviews. He holds the CISSP.
Christian also runs Inoculis, a fractional CIO and CISO practice for medical device, SaMD, IVD, and digital health companies.
If you are an operating partner or a portfolio CEO with an AI use case and no one to build it, send a short line on the company, the workflow, and the data involved.
contact@mvp.svBy scheduling a call, you agree that the information you submit will be used by Mulligan Valley Partners to manage your inquiry, schedule the meeting, and communicate with you about our services. Scheduling may be provided through a third-party scheduling provider, which may process your booking information under its own privacy practices.