All case studies

Healthcare · A large healthcare organization

Healthcare: documentation help with weights the hospital owns

The situation

Clinicians were spending a large share of the working day on documentation — visit summaries, patient-message drafts, prior-authorization write-ups — and the organization wanted AI to carry the first draft. Off-the-shelf models helped, but they wrote in a generic voice, missed the organization’s terminology, and carried a per-token bill that grew with every clinic that joined the rollout.

There was also a governance question underneath the cost question: if the model learned from the organization’s documents, who owned what it became?

What moved to Run BiOS

The organization fine-tuned a model on its own documentation patterns, running on dedicated GPUs through Run BiOS and billed per second of compute. The trained weights are the organization’s property — exported, held, and theirs to run anywhere.

Day-to-day drafting now runs on that fine-tuned model at published per-token rates, with the platform scaling behind it. The result is documentation help that writes the way the organization writes, at a cost structure that survives a full rollout.

Managed fine-tuningDedicated GPUsCustomer-owned weightsServerless inference

The outcome

  • AI spend reduced significantly against the general-purpose rollout it replaced
  • Clinician documentation time reduced significantly — drafts arrive in the organization’s own terminology
  • The weights stay with the organization: no lock-in on the asset its data created

Before

General-purpose model, generic voice, per-token bill growing with every clinic

With Run BiOS

Fine-tuned model in the organization’s own terminology, owned outright

Illustrative, not measured.

Facing the same bill?

Estimate your workload on the rate card, or talk to us about the deployment pattern that fits.

More case studies