Anonymized production case study
Their largest client went live in two weeks. Billing never stopped.
Medex agents took production work from two clinical systems through coding and claim preparation into the billing company’s existing PM—with no migration and no revenue lost during the transition.
- 2 weeks
- to go live
- $0
- revenue lost
- ≈5,000
- claims in a later week
A larger account—without pausing the operation behind it.
The billing company brought Medex into its largest client launch. Two weeks later, agents were doing defined production work inside the client’s existing systems. Claims kept moving and the transition caused no lost revenue.
- Largest provider client
- Launched
- Claims in a measured five-day cohort
- 2,290 billed
- Later weekly production volume
- ≈5,000 claims
This was not a field-to-field integration.
The account brought activity from two clinical source systems into a separate billing PM. Moving known values between them was the easy part. The operating burden came from deciding what the destination values should be, applying the client’s SOP, and taking the next action when the source was incomplete or the patient’s current PM state changed the answer.
Transports values that already exist
A conventional interface can copy a known patient, code, or payer field from A to B. It does not review a note, choose codes, resolve competing coverage, interpret an eligibility response, or decide how an existing claim should change.
Make the bounded decisions between systems
Agents read unstructured context, inspect the patient and claim already in the PM, apply account-specific rules, take the next action, and abstain to a human when the evidence falls outside the approved scope.
The agent used the existing systems like an operator.
- 01Interpreted the note
Read unstructured clinical documentation, selected CPT and ICD-10 codes, sequenced diagnoses, and applied the customer’s supporting coding logic.
- 02Resolved claim context
Chose the correct rendering provider, facility, place of service, payer/default, fee schedule, and billing status from source context plus the patient’s existing PM state.
- 03Resolved coverage
Matched or created the patient, chose the policy and member ID to verify, submitted real-time eligibility, interpreted active/inactive and Medicare/Medicaid/Advantage evidence, then added, suppressed, ordered, or defaulted policies according to the SOP.
- 04Constructed every charge line
Set CPT and ICD-10 relationships, modifiers, units, prices, and diagnosis pointers; applied conditional rules; and blocked duplicate same-date charges rather than blindly appending rows.
- 05Acted inside the PM
Created or updated the patient and claim, selected the resulting status, read the writeback, preserved an audit trail, and routed low-confidence or conflicting evidence to a person.
Observed insurance actions: in a bounded production batch, Medex submitted 423 real-time eligibility checks. The workflow made insurance changes for 374 patients, added 203 policies across 195 patients, and reordered or converted coverage for 179. Those are stateful decisions and actions—not records merely transported between systems.
A measured five-day claim cohort included 2,290 claims in billed state. As the account grew, a later weekly production cohort approached 5,000 claims. The customer kept its source systems and destination PM throughout.
Estimated time savings
601 staff-hours savedper 5,000-claim week
This counts decision-bearing work—not straight data transfer and not Medex checking its own work:
- 333.3 hrCoding decisions4 minutes per claim × 5,000 claims
- 125 hrClaim-level rules1.5 minutes per claim × 5,000 claims
- 125 hrCharge-line rules1.5 minutes per claim × 5,000 claims
- 17.5 hrEligibility + payer state349 projected actions × 3 minutes
333.3 + 125 + 125 + 17.5 = 600.8 hours → rounded to 601
Eligibility is incidence-weighted rather than applied to every claim. The measured cohort contained 2,433 source claims and 1,814 unique patients; 170 of those patients had a matching documented production eligibility record. Holding that observed incidence constant projects 349 eligibility actions per 5,000 claims. The three-minute allowance covers selecting the policy, submitting and interpreting the response, and updating payer state.
Methodology and claim definitions
The customer and provider identities are withheld. Production counts come from bounded operating cohorts, while the time-savings figure is an explicit estimate.
- Two weeks is the elapsed time from scoped implementation to production go-live for this customer workflow.
- $0 revenue lost means billing continued through the transition; no revenue interruption was attributed to the Medex launch.
- 2,433 source claims were present in the measured five-day cohort; 2,290 showed the billed state. The cohort included 1,814 unique patients, and a later weekly cohort approached 5,000 claims.
- 170 patients in that measured cohort matched a documented production eligibility record. The estimate holds that observed incidence constant; it does not assume eligibility for every claim.
- 601 staff-hours is an operating estimate: 7 minutes per claim for coding and decision-bearing billing actions, plus 3 minutes for each incidence-weighted eligibility/payer-state action. Straight data transfer and Medex QA/readback are excluded. It is not a measured customer time study or guarantee.
Results depend on source quality, system access, workflow rules, and the agreed automation scope.