The half of EMR documentation that scribes do not touch

It splits into two layers. The first is capturing the encounter and turning it into a draft note, which ambient AI scribes do from the audio of the visit. The second is entering structured data into the EMR itself: patient demographics, insurance, problem lists, codes, orders, and flowsheet fields, often across several screens. The first layer is well covered by scribe products. The second is the part that is still hand-keyed in most clinics, and it is what Mediar automates.

No. A scribe listens to the conversation and produces text. Mediar takes structured data, whether it came from a scribe's note, an intake form, a fax, or a PDF, and enters it into the EMR by driving the application the way a person would. The two are complementary: a scribe can hand off its output and Mediar puts it where it belongs in Epic or Cerner.

Yes. On a native Windows session Mediar reads the EMR through the Windows UI Automation tree, the same interface screen readers use. Inside Citrix the ICA protocol ships pixels but not that tree, so Mediar detects the Citrix session (wfica32.exe, CDViewer) and builds a pseudo accessibility tree with on-device segmentation, then drives Epic Hyperspace or Hyperdrive keyboard-first using Epic's own hotkeys. It also runs on Cerner, eClinicalWorks, Veridigm, Greenway, and Meditech.

Vision and OCR can run fully on-device with self-hosted models (Qwen2-VL, OmniParser), so screenshots of PHI never leave the endpoint. The agent acts under the clinician's existing session, inheriting the same access controls and audit trails the EMR already enforces, and every action is logged. Mediar is SOC 2 Type II certified and HIPAA compliant, and can deploy on-prem or inside your security perimeter.

No. That is the whole point of the accessibility-API approach. Mediar reads what the application already exposes to assistive technology and types into it directly, so it works on screens that have no public API at all. There is nothing to wait on from the vendor and no interface contract to maintain.

Mediar caches the layout of stable screens and only re-reads a screen when its visual hash changes, so a template tweak or a moved field does not silently break the workflow. Each field write is confirmed with a before-and-after screenshot diff; if the screen did not change the way it should have, the agent retries once and then escalates rather than charging ahead.

Traditional RPA binds to selectors or pixel coordinates, which break constantly on EMR screens and especially fail inside Citrix where there is no DOM and no UIA tree to bind to. Mediar learns a workflow by watching it once, then executes through accessibility APIs with a verify-after-act loop, so it self-heals instead of needing a developer every time a label moves. It also bills at runtime ($0.75 per minute) with no per-seat licensing.

Runtime is billed at $0.75 per minute of workflow execution, with no per-seat licensing. There is a $10,000 turn-key program fee that converts to credits with a bonus, so it functions as prepaid usage. For a clinic re-keying patient intake all day, the math is usually dominated by the staff hours it gives back.