Agentic RCM · The Plutus Health Operating Model

25 AI Agents. Working your Revenue Cycle, End-to-End.

Plutus Health doesn’t layer AI on top of traditional RCM. OlympusAI is built as an agentic revenue cycle operating system in which specialized AI agents execute workflows across eligibility, coding, claims, denials, A/R, and collections in real time.

Every workflow operates against defined SLAs, confidence thresholds, payer logic, and measurable KPIs, with 60% Agentic AI and 40% specialist judgment.

AI agents in production

RCM workflow stages

$1.7B

annual claims processed

9000+

providers in production

What is Agentic RCM?

Three Properties Define an Agent. All 25 of Ours Have Them

The industry uses "agentic" to mean many things. We use it to mean three specific properties, and we hold our agents to all three before they go to production. If an agent doesn't pass these tests, we don't ship it

Autonomous workflow execution

Each agent owns a specific RCM workflow end-to-end, not a single task.
The Eligibility Agent doesn't just check a payer; it runs the full eligibility workflow, including secondary discovery, benefits parsing, and exception escalation.
Agents complete the work without prompting or supervision under normal conditions.

Confidence-gated human-in-loop

Every agent has measurable confidence thresholds.
Above threshold, the agent acts. At the threshold, it suggests and a specialist confirms.
Below threshold, it escalates with full context.
Thresholds are calibrated continuously against actual production outcomes, not set once and forgotten. This is why our coding agent ships at 99.2%+ accuracy and our clean-claim agent ships at 97%+.

KPI-bound to commercial outcomes

Each agent has a single measurable KPI tied to the commercial model.
Eligibility Agent → eligibility-related denial rate < 1%. Coding Agent → accuracy > 99.2%.
Cash Flow Forecasting Agent → forecast accuracy ±3% MoM.
KPIs report into the weekly client CFO pack. If an agent's KPI degrades, we fix the agent, not the client.

The 25 Agents, Mapped to Workflow

Every Stage of the Revenue Cycle. One or More Agents on Each.

Below: the full Plutus Health agent architecture, organized by the 7 canonical RCM workflow stages. For each agent: what it does autonomously, what triggers human-in-loop escalation, and the single KPI it's held to.

Stage 1

Patient Access

3 agents

Front-end intake, eligibility, and benefits verification, before services are rendered.

Eligibility & Benefits Verification Agent

Autonomous

Real-time and batch verification across all payers; flags coverage gaps, inactive plans, missing authorizations.

Human-in-loop

Coverage decisions on edge cases (carve-outs, tiered structures) escalate to specialist.

KPI

Eligibility-related denial rate < 1%

Insurance Discovery Agent

Autonomous

Searches multiple data sources for secondary or hidden coverage before patient moves to self-pay.

Human-in-loop

Confirmed coverage routes to billing specialist for verification.

KPI

2–5% revenue recovered from undiscovered payers

Benefits Exception Review Agent

Autonomous

Reviews complex benefit documents and payer call transcripts to render coverage determinations on edge cases.

Human-in-loop

Final adjudication routes to payer-specific SME.

KPI

Edge-case write-off rate reduced 60%+

Stage 2

Clinical & Coding

4 agents

Documentation integrity, charge capture, and specialty-aware coding before claim submission.

CDI Agent (Clinical Documentation Integrity)

Autonomous

Reviews documentation for specificity, medical necessity, and code support; queues queries to clinicians.

Human-in-loop

Clinician confirms or rejects suggested query.

KPI

Query resolution turnaround < 48 hrs

Charge Capture Agent

Autonomous

Validates encounter charges against documentation and contracted fee schedule.

Human-in-loop

Variance > threshold flags to revenue integrity reviewer.

KPI

Charge capture rate > 99%

Medical Coding Agent

Autonomous

AAPC-trained coding agent with specialty-specific rule sets (ABA, ASC, FSER, psych, SUD, surgical).

Human-in-loop

Confidence below threshold routes to certified coder.

KPI

Coding accuracy > 99.2%

Claim Scrubbing Agent

Autonomous

Pre-submission scrubbing with payer-specific edits and policy alignment; 200+ payer rules.

Human-in-loop

Edit conflicts flag to claim specialist.

KPI

Clean-claim rate > 97%

Stage 3

Prior Auth, Credentialing & Submission

4 agents

Prior authorization, provider enrollment, claim submission, and status follow-up.

Prior Authorization Agent

Autonomous

Auto-routes auth requests, polls payer status, generates appeal-ready packets with medical necessity evidence.

Human-in-loop

Adverse determinations route to clinical reviewer for appeal.

KPI

Median auth TAT < 48 hrs

Credentialing Agent

Autonomous

Provider enrollment monitoring, re-credentialing reminders, payer portal updates.

Human-in-loop

Application errors and exceptions escalate to credentialing specialist.

KPI

Credentialing-related cash gap < 5 days

Claim Status Agent

Autonomous

Polls payer portals and IVRs for claim status, flags stale claims for follow-up.

Human-in-loop

Stale claims > 30 days flag to AR specialist.

KPI

Zero claims aging silently past 60 days

Correspondence Review Agent

Autonomous

Reads payer correspondence (paper + EDI), routes to appropriate workqueues.

Human-in-loop

Ambiguous correspondence flagged for human review.

KPI

Correspondence backlog < 24 hrs

Stage 4

Denial & Appeals

4 agents

Denial root-cause analysis, appeal generation, and IDR / NSA arbitration.

Denial Analysis Agent

Autonomous

Categorizes denials by root cause and payer; surfaces systemic patterns and feeds upstream prevention.

Human-in-loop

Root-cause patterns reviewed weekly with operations leadership.

KPI

Denial rate reduced ≥2 pts quarter-over-quarter

Appeal Agent

Autonomous

Drafts payer-specific appeal letters with auto-assembled evidence packs.

Human-in-loop

Appeal specialist reviews and submits.

KPI

Appeal overturn rate > 65%

IDR / NSA Likelihood Agent

Autonomous

Predicts which OON claims qualify for federal IDR before submission.

Human-in-loop

Eligible claims route to IDR specialist team.

KPI

IDR-eligible volume capture > 90%

Correspondence Review Agent

Autonomous

Auto-generates IDR submission packages with QPA analysis and case-strength scoring.

Human-in-loop

IDR specialist confirms package and submits.

KPI

IDR prep time compressed from days to minutes

Stage 5

Payment & Collections

4 agents

ERA posting, underpayment recovery, patient AR, and outbound voice automation.

Payment Posting Agent

Autonomous

Auto-posts ERAs, reconciles variances, flags underpayments to contract terms.

Human-in-loop

Variances > threshold route to underpayment specialist.

KPI

Posting backlog < 24 hrs

Patient AR & Collections Agent

Autonomous

Patient-pay segmentation, statement cadence, propensity scoring, payment plan offers.

Human-in-loop

Disputed balances escalate to patient billing specialist.

KPI

Patient-pay yield > industry median

Voice AI — Payer Agent

Autonomous

Calls payer IVRs at scale; navigates trees, extracts claim status data, runs eligibility inquiries 24/7.

Human-in-loop

Failed calls or unusual responses route to AR specialist.

KPI

Payer follow-up coverage 100% within SLA

Correspondence Review Agent

Autonomous

Outbound patient calls for verification, statement reminders, payment plan setup.

Human-in-loop

Sensitive conversations escalate to live agent.

KPI

Patient outreach scaled without headcount

Stage 6

Operations & Intelligence

2 agents

Contract review, payer policy intelligence, and operational risk monitoring.

Contract Review Agent

Autonomous

Reads payer contracts; models reimbursement rates; flags termination and renewal triggers.

Human-in-loop

Material changes route to contracts SME for negotiation.

KPI

Zero missed renewal windows

Payer Intelligence Engine (PIE) Agent

Autonomous

Tracks payer policy changes, fee-schedule updates, denial-pattern shifts in real time.

Human-in-loop

Material policy changes alerted to operations leadership.

KPI

≥95% of material payer changes flagged ≥5 business days before effective date

Stage 7

Analytics & Commercial

4 agents

Process excellence sampling, cash flow forecasting, natural-language analytics.

Process Excellence Sampling Agent

Autonomous

Statistical sampling QA across coding, posting, and denial workqueues.

Human-in-loop

Quality variances trigger root-cause review.

KPI

Quality variance < 2% across all workqueues

Process Excellence Full Review Agent

Autonomous

Deep-dive audit on flagged accounts and high-value claims.

Human-in-loop

Audit findings route to operations remediation.

KPI

100% high-value claim coverage

Cash Flow Forecasting Agent

Autonomous

Forecasts collections by payer, line of business, and aging cohort.

Human-in-loop

Variance from forecast triggers CFO review.

KPI

Forecast accuracy within ±3% MoM

Analytics Agent

Autonomous

Natural-language query layer over the full RCM data lake, ask in plain English, get an answer.

Human-in-loop

Surfaces insight in real-time chat to operations and clients.

KPI

≥95% of natural-language queries answered in <15 seconds

30 minutes · No commitment

Ready to See Where Your Revenue Is Leaking?

Our 30-day diagnostic baselines your current state against MGMA / HFMA medians, quantifies the leakage, and shows you the to-be model — before you sign anything.

Get Your Revenue Diagnostic →Download Benchmark Report