Case study

Concentrate AI

How Concentrate AI launched a HIPAA-compliant AI gateway to 500K+ views

Every new AI provider a healthcare company wants to use means a contract, a security review and a BAA, and months of waiting. Concentrate AI replaces all of it with one agreement. We made the film that took it to general availability, with both co-founders on camera.

Views on X
Likes on X
320+
Bookmarks on X
161

Concentrate AI × EPTA

Concentrate AI buys inference for every major model through one API. For healthcare and life sciences it adds the part nobody else had: one universal BAA that covers HIPAA-eligible models across eight providers, with provider agreements, billing, spend controls and fallbacks handled for you.

Ari Jacoby and Todd Lieberman lead it from New York. It is SOC 2 Type 2 and HIPAA enabled, and its first customers include Pocket, Kinstead Health and Hummingbird Healthcare.

The brief

The buyer is an engineering leader at a company that handles patient data, and the pain is not technical. Their team can ship a feature in three weeks and then wait months for legal, security and compliance to approve the model behind it.

The film had to name that wait precisely enough that the buyer recognised their own backlog, then make one agreement feel like the obvious way out.

The film

Ari opens with a feature, a lab-result summarizer three weeks in the making and ready to ship, and then the question that stalls it: is this provider approved to handle PHI? Todd takes the process from there, a contract, a security review, a BAA, months on a calendar that fills and starts again with every new provider.

One BAA, every model. The film spends its first half on the problem so that line can do the selling on its own.

Then the turn: a single BAA, the models that were off limits now available, switching providers with one routing change, lower cost models in production. The close answers the compliance reader directly, broader access without less control, with ZDR, spend limits, PHI redaction and audit trails. Sixty seconds, start to finish.

Next launch

Planning a launch?

Tell us what is launching and when. Thirty minutes is usually enough to know whether we are the right team for it.