01 · What DinoDial does

AI agents that run revenue cycle work from start to finish.

DinoDial builds AI agents for healthcare revenue cycle teams. One agent takes a case, such as a prior authorization, an unpaid claim or a denial, and works it until it is resolved. It calls payers, works payer portals, reads policies and letters, and updates the EHR over the days or weeks the case stays open. A supervisor checks every step, and billing staff approve anything sent to a payer.

Long-horizon

A case stays open for days or weeks and takes dozens of steps. The agent keeps the full history and acts on each step when it falls due.

Multimodal

The same agent works across phone calls, payer portals, scanned documents and the EHR, the systems a billing team uses every day.

Supervised

Every action is checked against fixed rules before it runs, and people approve what goes to a payer.

What the agents cover in the revenue cycle

Eligibility and benefitsCoverage, network status and patient cost before the visit

Prior authorizationRequirements, evidence, submission and follow-upExample in this briefing

Authorization upkeepReschedules, order changes, expiry and units used

Claim pre-checkAuthorization, codes, units and dates matched before billing

Claim statusPended and unpaid claims followed up by portal and phone

Denials and appealsReason codes read, records gathered, appeals drafted

Patient balancesBalance calls, payment plans and inbound billing questions

The problem

Volume keeps rising, and billing teams cannot hire fast enough to keep up.

Every authorization and every claim follow-up takes staff time on portals and phones. As volume grows, the work that does not get done on time turns into backlog: requests submitted late, pended claims nobody checks, denials that miss their appeal window. Adding people is slow and costly, and the queue keeps growing in the meantime.

52.8MMedicare Advantage prior authorization decisions in 2024, up from about 50M in 2023

39prior authorizations per physician per week, about 13 hours of staff and physician time

43%of providers say they are understaffed

41%of providers now see denial rates of 10% or more

02 · Example workflow: prior authorization

How the agent prepares and submits a prior authorization.

A prior authorization is the insurer's approval that a planned procedure is covered. Without it, the claim is denied. This example follows one spine fusion request from the signed surgical order to a complete submission on the reviewer's portal, within one working day. Along the way the agent calls the payer, reads a 37-page policy, checks the chart against every criterion, and holds the case when one has no evidence. Each mark below is a step the agent logged. Select one to see what happened.

Signed order to submitted request · Day 1

Goal:

After submission

The agent checks the request on day 3, day 5 and day 7, calls the payer when the portal does not say what is missing, sends records only after approval, and reads the decision when it arrives.

The rest of the cycle

Eligibility, auth upkeep, claim pre-checks, claim status, denials and patient balance calls run on the same agents, supervisor and approval gates.

03 · Product

What your billing team sees.

Staff work from two screens. The approval screen holds everything a specialist needs to sign off a submission. The case view shows where each request stands and what the agent is doing on it right now.

DINODIALApprovals 4CasesCallsPayersAuthorization specialist

M. Alvarez · Lumbar fusion L4-L5Submit to the plan's delegated spine reviewer · prepared by the agent at 14:23

7 of 7 criteria met

Policy criteria · reviewer lumbar fusion policy, current version

Neurogenic claudication or radicular painProgress note · p.2

Functional impairmentPT evaluation · p.1

3+ months supervised conservative carePT notes · injections

Imaging confirms stenosisMRI report · p.1

Grade 1 listhesis or 4 mm translationFlex-ext X-ray

Single-level fusionSurgical order

Tobacco-free 6 weeks before surgeryAddendum · gap resolved

Codes

226332285322840M43.16M48.062

Code sets current · NPI check digit valid · member ID format valid

Attachments · 9

Order · Progress notes · PT evaluation · PT notes · Injection reports · MRI report · X-ray report · Addendum · Request form

Approve and submitReturn to agent

Approval screen. Every criterion links to its source page. Nothing is submitted until a specialist approves.

DINODIALApprovalsCasesCallsPayersAuthorization specialist

M. Alvarez · Lumbar fusion L4-L5Delegated spine reviewer · submitted day 1, 14:49

Call in progress

RequirementsPolicyEvidenceApprovalSubmittedFollow-upDecision

Activity

D1 14:49Submitted on reviewer portal

D3 09:00Day-3 status check due

D3 09:01Portal: pended, no reason given

D3 09:04Call to reviewer started

SupervisorStage: follow-upAllowed tools: portal.status · call.reviewer · ehr.readBlocked: submit, code change

Reviewer provider line24:18 · 19 min on hold

REPThanks for holding. How can I help?

AGENTI'm calling about the lumbar fusion request for M. Alvarez, case ending 4471. What does the clinical reviewer need to finish it?

REPShe's asking for the PT discharge summary. The evaluation is there, the discharge note isn't.

Captured from this call

Missing itemPT discharge summary

Call referenceStored with transcript

Next stepLocate record, then send for approval

Case view. The stage tracker, the live call and what the supervisor allows at this stage, on one screen.

Product screens shown with an illustrative case. No patient data.

04 · Manual vs DinoDial

What it takes to run 100 requests like this one a week.

From the signed order to the payer's decision, the spine fusion request in the example takes about 2 h 45 min of staff time by hand: the steps shown above plus the follow-up, with three payer calls in total. The DinoDial agent takes about 1 h 36 min on the same request, 55 minutes of it on hold, and needs a person only for two approvals.

ManualDinoDial

Time per request

2 h 45 min on portals, phone and forms

1 h 36 min, 55 of them on hold

Staff time per request

2 h 45 min

~6 min, two approvals

Staff hours for 100 a week

275 hours, about 7 full-time staff

~10 hours. Agent time runs in parallel, with no extra staff

A payer status call

25 min of staff time, most of it on hold

The agent waits on hold and runs hundreds of calls at once

Volume grows 50%

Hire and train 3 to 4 more staff

Same team. Agents start on the extra cases the same day

Follow-up on open requests

When someone reaches the case in the queue

On day 3, day 5 and day 7, every case

Quality checks

A small sample of calls

Every call and action

Staff hours per week as volume growsManual · 2 h 45 min eachDinoDial · ~6 min of approvals each

Times per request are from the example. Approval time of about 3 minutes each is our estimate for a prepared packet. Full-time staff counted at 40 hours a week.

05 · Technology: how the agents stay predictable

Every call and every action runs through a supervisor.

The models listen, speak and read. The supervisor decides what they may do next. Calls run in fixed stages, and every portal, document and EHR action is proposed by an agent and verified before it runs.

Facts carry their source

Every value taken from a chart or letter links to its document and page. A value without a source is not used.

Codes checked by rules

CPT and ICD-10-CM against current code sets, NPI check digit, member ID format, dates against the auth window.

Calls stay in their stage

The agent states only facts in the case record. Answers come back as structured fields with the call reference.

People approve what is sent

Submissions, records, code changes and appeals wait for billing staff. Peer-to-peer stays with the physician.

0Guardrail violations

across 1M+ production calls

100%Calls evaluated

every call scored, no sampling

~650 msResponse latency

live calls, supervisor included

~100%Tool-call reliability

stage-bound verification