Decision engines are not execution engines.

1. Why we exist: the residual tax

To take on the residual labor tax enterprise software leaves on the desk.

EHRs, revenue-cycle platforms and CRMs are built to store records and to carry the decisions clinicians and payers make. On live desks, a material share of administrative volume still has no partnered API. Our working estimate is 15–30% of volume on mixed-payer worklists. It is not measured and not a universal constant; the 30-day ledger is how a desk finds its own number.

When an authorization request must go to a regional insurer with no clearinghouse connection, or a medical-necessity chart must be split and attached on a legacy portal, systems such as Epic or Waystar reach their structural boundary. We name them here as examples of record and decision platforms, not as partners or integrations.

We exist because knowing what should happen is only half the work. We do not ask hospitals to rip out their systems. We work underneath them, in the coordinator's browser, to finish the leftover work.

2. Why now: why scripts failed, and what changed

Why portal automation failed for 15 years, and why the floor is different now.

Automating web portals is not a new desire. Healthcare and enterprise operators have attempted it for two decades. It failed because the tools were structurally brittle.

Scripted automation failed (2010–2023)

• It relied on rigid HTML paths (XPaths) that broke whenever a portal changed its layout.
• Cloud scraper farms were blocked by bot mitigation (Akamai, Cloudflare and their equivalents), hospital VPNs and MFA. Those names describe the class of barrier, not a claimed bypass of any vendor.
• It sold unattended autonomy. That is the liability a desk owner cannot sign.

What changed (2024–2026)

Multimodal models

Models can ground against a live screen, so a moved field is found again by what it shows instead of failing on a dead path.

Fast chart-to-field staging

Structured extraction from a chart into form fields takes a few seconds on a typical case. That is timed per lane in the pilot, not promised as a constant.

Voice assist

Low-latency audio models over WebRTC can work through phone trees and hand off when a live representative answers. This is a change in the field; the Operator Extension itself does not call payers.

How DeskGate is built to be onboarded

• Session
  Client-side, in the coordinator's session. The Operator Extension runs inside the coordinator's already-authenticated browser. The MFA and VPN are the coordinator's, not a farm's. There is no embed and no script tag.
• Role
  Hands, not a second brain. Execution under the EHR or RCM you already use. No competing clinical decision engine.
• Gate
  A mandatory human gate. Staging is automatic; Submit stays with the coordinator. Liability is contained by design, not declared gone.

3. How we exist

The extension in the coordinator's session. Submit stays human.

Operator Extension 2.6.8 opens in the Chrome or Edge side panel beside the payer tab the coordinator already logged into. The work stays in that session. The coordinator reviews the staged fields and clicks Submit. Under that gate, four layers do the residual work. The old embed.js overlay is retired: a script on a page you own cannot read a third-party payer portal.

L1The session

The coordinator's own authenticated browser session: their login, their MFA, their VPN, and the payer tab they already have open. DeskGate adds no access path of its own and does not unlock portals.

L2The draft

Plans the order of the residual form inputs. Quotes chart facts already on screen into the draft, marks anything the chart does not say as [MISS], and stops. It makes no clinical or medical-necessity decision and writes no rationale of its own.

L3 and L4Screen grounding and in-page staging

Visual grounding binds each action to the live viewport. In-page staging types into empty fields and reads every box back. On a known portal the aim is most fields in a few seconds; coverage is measured per lane, not assumed.

4. Is our price fair?

Priced against recovered desk hours, not speculative seats.

Most enterprise software asks for a large commitment on a long procurement cycle before anyone has watched a live lane. DeskGate starts with a 30-day measurement on one named portal family: handling time and rework recorded against the desk's own baseline, in a ledger we publish blinded. There is no minutes SLA, no savings guarantee and no share of recovered revenue.

First 10 desks

30 days free on one named portal family, in exchange for one coordinator, a pre-period baseline and a blinded ledger DeskGate may publish. Practice-manager or billing-director sign-off.

After day 30

Stop, or subscribe. S-Practice is $490 a month or $5,400 prepaid for the year (1–3 seats, one named portal family). S-Desk is $1,950 a month or $21,000 prepaid for the year (unlimited coordinators on that desk, one named portal family). S-Agency is $4,900 a month or $54,000 prepaid for the year (up to three named portal families). Zero case metering on every plan.

Larger networks

Talk to us.

Worked example, to be tested in 30 days

Illustrative arithmetic, not a forecast and not a DeskGate result. The table models a 20-FTE billing desk. Every input is an assumption stated in the row, so a buyer can substitute their own loaded cost, residual hours and plan. We have not measured a live desk.

Substitute your FTE count, residual hours and loaded rate. If the 30-day ledger is thin, that is usually a parameter miss (cohort, exceptions, portal family), not a broken promise, because there is no SLA to break. Fairness here means measured access, not a guaranteed cut.

5. Value beyond minutes

Even if minute savings vary, the desk still buys structural control.

Handling minutes in the 30-day measurement are the first quantitative bar. Desks that stay usually stay because the same runtime also tightens five other failure modes.

01Denial avoidance

Completeness checks before Submit on required fields and attachments, so a missing item is caught before it becomes appeal rework.

02Audit provenance

The ledger keeps a SHA-256 of the payer reference. It is not a copy of the case and not the portal response.

03Staff retention

Removes the worst copy-and-paste load from the coordinator's day. Turnover in this work is high across the industry; the bet is keeping the people you already trained.

04Cash-flow velocity

Aims to shorten queue time from a multi-day backlog toward same-day submission on the piloted lane, which is how days in accounts receivable actually move.

05Continuity

Keeps the case's working context with the work item in the coordinator's session, so a complex prior authorization is less likely to drop at a shift change or during time off.

6. Will a desk actually onboard it?

It survives review because it answers the three reasons healthcare says no to AI.

Healthcare leaders reject most AI pitches for the same three reasons: clinical liability, rip-and-replace of the EHR or RCM, and patient data leaving the perimeter. DeskGate is built to answer each one.