Kratu AI: IntelligentAgents Across theVisit and RevenueCycle
Four specialized agents, one unified engine inside the chart and the claims table your team already works in. Kratu AI prepares the patient before the visit, documents the encounter, verifies payer rules, and resolves accounts receivable.
Unified Architecture
One AI Engine, Four Agents, One Patient Journey
Most healthcare AI arrives as disconnected point tools — a scribe here, a claim scrubber there, each with its own login and isolated data silo. Kratu AI operates as one continuous engine across four vital milestones in the exact same record.
Stage 01 • Before Visit
Visit Prep
Patients share their goals via secure SMS links with zero portal logins. Agendas land structured in the clinician's chart.
View Agent 01Stage 02 • During & Post Visit
Kratu Co-Pilot
Ambient clinical documentation drafts directly on your organization's custom templates with deep EHR care insights.
View Agent 02Stage 03 • Before Claim Filing
Payor Policy Intelligence
Complex payer manuals and billing grids are converted into live pre-submission claim validation rules with reason codes.
View Agent 03Stage 04 • Post-Submission
A/R Agent
Continuous revenue exposure scoring, root-cause denial intelligence, and fully drafted, evidentiary appeal packets.
View Agent 04
The handoffs are the point: what the patient said before the visit informs the documentation, which informs the billing rules, which shields the claim from denials.
Agent 01 • Pre-Visit Access
Visit Prep: AI Pre-Visit Questionnaire with Zero Portal Friction
Patients arrive unprepared because preparing is inconvenient: forgotten passwords, complex portal apps, or rushed waiting room clipboard forms that clinicians skim seconds before entering.
Visit Prep sends an encrypted, clinic-branded SMS link before the appointment. After rapid biometric or one-time passcode verification, patients articulate their chief complaints, questions, and medication changes in conversational English or Spanish.
The appointment opens on the patient's priorities instead of five minutes of awkward administrative catching up.
No App or Portal
Frictionless 1-click web experience via encrypted SMS. Nothing to download.
Identity-Verified
Name, DOB, and secure OTP token verification required before any PHI access.
Bilingual Native
Automatic Spanish/English conversational language switching.
Chart Integration
Confirmed agenda items attach directly to the scheduled encounter in Elixir.
Patient Mobile Check-in (SMS Secured)Verified (OTP)Agenda Transferred to Clinician EHR:
Agent 02 • Clinical Encounter
Kratu Co-Pilot: Ambient Documentation on Your Own Organization's Templates
Generic ambient scribes produce a rigid SOAP summary and stop. In specialized clinics — behavioral health intakes, addiction treatment plans, IOP reviews — clinicians end up spending their evenings retyping unstructured audio notes into mandatory forms.
Kratu Co-Pilot listens ambiently and populates the exact EHR templates your organization designed during onboarding. Simultaneously, an intelligent Insights panel compares conversational facts against the longitudinal chart.
Approving documentation brings orders and referrals forward for review — but nothing is transmitted without clinician sign-off.
Custom Templates
Conforms to intake assessments, group therapies, and ASAM criteria.
Care Gap Radar
Surfaces missed screenings, HEDIS gaps, and undocumented comorbidities.
Form Dictation Scribe
Speak naturally once; AI routes details into separate specific fields.
Zero Order Auto-Send
Clinicians maintain 100% control over meds, referrals, and diagnoses.
Kratu Co-Pilot Clinical WorkspaceTemplate Matched- SOAP Note Draft
- Patient Insights (4)
- Care Plan
- Form Scribe
Subjective
Patient reports bilateral knee tenderness following moderate exertion. Denies acute trauma. Adherent with home stretching regimen.
Objective
Active ROM limited to 110 degrees on right flexion. No erythema or effusion observed upon manual palpation.
Assessment & Plan
1. Right knee osteoarthritic flare. Prescribe NSAID regimen.
2. Referral to physical medicine evaluation.
EHR Insight: Patient due for annual diabetic foot screening (Quality Measure CMS122).Add to Orders
Agent 03 • Pre-Claim Intelligence
Payor Policy Intelligence: Transforming 200-Page PDFs into Active Billing Rules
The rules governing claim adjudication already exist — buried inside 200-page policy PDFs, Excel reimbursement grids, and clinical coverage bulletins that no human coder has time to read on every claim.
Payor Policy Intelligence reads and indexes payer documentation into live deterministic validation rules. It recommends compliant CPT/HCPCS codes with citation links, scoring each draft claim prior to 837 transmission.
Billing errors are not a diligence failure — they are a document format failure solved by AI.
Multi-Format Parser
Ingests PDF guidelines, Excel fee schedules, and Word policy updates.
Evidence Backlinks
Every recommended code displays the specific clause, page, and policy year.
Pre-Submission Scoring
Distinguishes deterministic structural edits from probabilistic risk signals.
Explicit Action Fixes
Provides exact corrective field entries before clearinghouse submission.
Pre-Claim Scoring • BCBS Commercial Episode94 / 100 Clean Pass94Score
CMS-1500 Ready for 837P Transmission
Evaluated against BCBS Provider Manual v2026.3, Section 14.8
- CPT 90834 requires Modifier 95 for Telehealth encounter — Verified on Line 24d.
- Primary Diagnosis F33.1 supports Medical Necessity threshold under policy §4.2.
- Frequency Limit: Prior service on 10/12 exceeds 1 visit/7 days without prior authorization note.
Recommended Resolution: Attach Auth #77291B from EHR encounter tab before transmitting to clearinghouse.
Agent 04 • Post-Claim Recovery
A/R Agent: Scoring Accounts and Claims by Real Financial Exposure
Traditional billing teams prioritize aging claims by day count alone (30, 60, 90+ days), because older reports are the only thing legacy EHRs can sort by. As a result, a $5,000 recoverable balance and a $20 balance get identical priority.
The A/R Agent scores every account at creation using historical payment propensity and contract analytics. When denials occur, root causes, payer overturn odds, and ready-to-mail appeal letters arrive bundled together.
Your billing staff spends their day recovering money that is genuinely collectable instead of drowning in spreadsheets.
Account Risk Score
Predicts patient and payer collection propensity 0–100 before claims open.
Root Cause Clarity
Translates obscure ANSI/CARC/RARC denial codes into plain remediation actions.
Drafted Appeals
Assembles clinical excerpts and policy citations into complete appeal packets.
Forecasted Pay Window
Provides statistically backed expected payment days for CFO cash forecasting.
A/R Worklist Intelligence • High Exposure QueueScored by Risk| Claim ID | Payer | Balance | Risk Score | Status |
|---|---|---|---|---|
| #CLM-8831 | Aetna Choice | $4,850.00 | 89 / High | CO-197 Denied |
| #CLM-8849 | UnitedHealth | $1,220.00 | 45 / Med | Pended 18 Days |
| #CLM-8890 | Medicare Part B | $3,140.00 | 82 / High | CO-16 Info Req |
AI-Generated Appeal Packet Available
Re: Denial CO-197 (Pre-certification missing). Root cause identified: Hospital transfer auth was granted under parent NPI. Formal reconsideration letter with admission timestamps generated and ready for billing approval.
Platform Cohesion
Why One AI Engine Beats Four Disconnected Point Tools
Buying four point solutions means four separate vendor contracts, four software logins, four complex API syncs, and four systems that each see only an incomplete fragment of the patient record.
One Single Record, Never Reconciled
Every agent reads and writes directly to the primary Elixir patient chart and claims ledger. Zero night batch syncs, zero discrepancies between third-party databases.
Context Connects Across Handoffs
What the patient shared in pre-visit check-in flows seamlessly into the clinician's draft. The clinical notes directly validate the billing rules scored prior to submission.
Zero Friction Learning Curve
Kratu AI surfaces directly within the native EHR chart, care episode, and claims grid your staff already uses every day. No new tabs, no extra logins, no workflow disruptions.
Single Accountability Partner
One trusted healthcare technology vendor responsible for both the clinical documentation and the revenue cycle outcomes — eliminating cross-vendor blame.
Responsible Governance
How Kratu AI Keeps Clinicians and Billers in Total Control
Every agent prepares work and hands the final decision to an authorized person. That is not a temporary limitation — it is our foundational system architecture that guarantees complete auditability when payers or regulators inspect a decision.
| Agent | What AI Prepares | Where the Human Decides |
|---|---|---|
| Visit Prep | Synthesizes open-ended patient remarks into structured, clinical agenda items prior to check-in. | Never diagnoses or prioritizes triage severity. The patient confirms their agenda, and the provider reviews before opening the chart. |
| Kratu Co-Pilot | Drafts documentation on configured organizational templates and highlights potential care gaps. | The clinician must review, edit, and sign notes. Approval alone never triggers automated medication orders or specialty referrals. |
| Payor Policy Intel | Extracts rules from payer policies, suggests code sets with confidence scores, and flags claim risk factors. | Uncertain or low-confidence values are held for manual coding review. No code is committed and no claim filed without user confirmation. |
| A/R Agent | Ranks accounts by risk, diagnoses denial root causes, and compiles evidentiary appeal packets. | Risk ranks the team's worklist but never executes automated adjustments or writes off balances without billing manager sign-off. |
Protected Health Information (PHI) is never used to train public or cross-tenant foundational models.HIPAA • SOC 2 Type II • HITRUST Certified
Frequently Asked Questions About Kratu AI
What clinical leaders, revenue executives, and healthcare CIOs ask most often when evaluating Kratu AI.
Kratu AI is the native AI engine embedded across Elixir EHR and RCM. It operates as four specialized agents: Visit Prep before the visit, Kratu Co-Pilot during and after documentation, Payor Policy Intelligence before filing claims, and the A/R Agent for post-submission denial resolution.
No. Kratu AI is built into Elixir and appears inside the chart, care episode and claims screens your staff already use, under the same login and the same Salesforce permissions. There is no separate application or extra tab to open.
No. The four agents share one engine and one patient record, but each can be introduced on its own. You can start with the agent that addresses your most pressing problem, such as Visit Prep for intake or the A/R Agent for denials, and add the others when you are ready.
No. Every agent prepares work and hands the final decision to an authorized person. Clinicians review, edit and sign notes; no code is committed and no claim is filed without user confirmation; low-confidence values are held for manual review; and the A/R Agent never adjusts or writes off a balance without billing manager sign-off.
Yes. Payor Policy Intelligence shows the policy clause, page and policy year behind each recommendation, A/R Agent appeal packets cite the clinical excerpts and policy sections they rely on, and every final decision is made and recorded by a named user. Auditors see both the reasoning and the person accountable for it.
No. Protected health information is not used to train public or cross-tenant foundation models. Our Responsible AI page explains how patient data is handled across Elixir's AI features.
Next-Generation Healthcare Intelligence
See All Four Agents Running on Your Own Workflows
Walk through the visit, the clinical documentation, the claim scoring, and the denial follow-up using your organization's custom templates, payer mix, and appointment types.