One AI engine for your entire mid-cycle — note to claim to appeal.
Most hospitals stitch together a tool for CDI, another for coding, another for queries, another for denials. Medically captures clinical truth once — reading the whole record, citing the evidence, and flagging findings in both directions with a defensibility score on every one — and serves every downstream outcome from it.
Built on four non-negotiables.
The whole record, every time
Notes, labs, medications, vitals, orders, flowsheets, microbiology, imaging and pathology results, prior codes — across 100% of charts, not a sample. Read like a clinician would.
Evidence behind every output
Every finding, code, and query links to the verbatim chart text that triggered it — located in the record, verifiable in one click, defensible in an audit.
Downgrades get equal billing
Unsupported diagnoses and DRG downgrades surface with the same prominence as upgrades. Every finding carries a defensibility score, with the arithmetic shown.
Bounded, human-in-the-loop
Never-fabricate guarantees, anti-upcoding guardrails, reason-coded decisions, and a person approving every change that touches a patient's record.
Pre-bill CDI review
Catch the documented-but-uncoded diagnoses before the claim goes out — and the documented-but-unsupported ones, too. Each finding is shown with the clinical evidence behind it, its DRG and dollar impact, and a defensibility score, so your specialists verify in a click instead of hunting through the chart.
- Every inpatient discharge reviewed pre-bill, inside the bill-hold window — no DNFB impact
- Worklists ranked by dollar impact, denial risk, and bill-hold deadline, with auto-assignment to your reviewers
- Concurrent, too: findings recompute as new results arrive mid-stay, with an expected working DRG
- MS-DRG shift, APR-DRG severity and risk of mortality (SOI/ROM), and dollars — per finding
- Accept, reject, or query with a required reason code; every decision lands in the audit ledger
Findings that go both ways — with a defensibility score on every one
Revenue-first AI only ever moves the DRG up. Medically surfaces unsupported diagnoses and downgrades with the same prominence as upgrades, evaluates each finding against the clinical criteria the specific payer actually applies, and scores how likely the diagnosis is to survive clinical-validation review — with the arithmetic shown, not a black box.
- Two-way findings: upgrades, unsupported diagnoses, and DRG downgrades in one list, ranked the same way
- Defensibility score (0–100) computed deterministically from stored, located facts — the AI proposes and cites; it never produces the score
- Payer-specific clinical criteria — Sepsis-2 or Sepsis-3, ASPEN or GLIM malnutrition, KDIGO acute kidney injury, acute respiratory failure thresholds — versioned per payer
- Append-only audit ledger: disposition, reason code, evidence, criteria applied, model and prompt versions, on every finding
- OIG-ready audit export, plus precision and acceptance rates by finding type in the compliance console
Model rationale is shown separately and labeled as not used in the score. A payer selector re-runs the pass against a different payer's criteria.
Two-way findings
Upgrade, unsupported, and downgrade findings live in one list with equal prominence. Downside exposure is tracked in dollars, just like upside.
Defensibility score
Criteria met, evidence strength, consistency, and counter-evidence combine into a 0–100 score. Same inputs, same number, every time — and the breakdown is stored with it.
Payer criteria registry
Each hospital chooses which definitions apply per payer. Findings are evaluated against what that payer will accept, not a one-size algorithm.
Audit ledger & OIG export
Every disposition is appended, never overwritten. Export evidence, criteria, score breakdown, reason codes, and model versions in one permission-gated file.
Computer-assisted coding
Your coders stop reading entire charts to find the one line that matters. Medically auto-suggests ICD-10 and CPT codes with a confidence score, the supporting evidence, and the DRG math — they confirm, adjust, and move on.
- DRG-aware suggestions that optimize accuracy, not just speed — MS-DRG and APR-DRG (SOI/ROM) impact per code
- Confidence scores and one-click evidence on every code
- Present-on-admission indicators, CC/MCC capture, and quality-sensitive codes flagged for review
- Compliance checks built in — including down-coding when warranted
Physician queries that get answered
Compliant, pre-drafted queries reach the physician inside their workflow — not days after discharge. The clinical indicators and evidence are already attached, so the answer takes seconds.
- Non-leading templates that follow ACDIS/AHIMA compliant-query practice — multiple choice, clinical indicators attached, "unable to determine" always offered
- Real-time CAPD nudges at the point of documentation
- Full tracking — pending, answered, agree/disagree, turnaround — per physician and per query type
- Every query is tied to its finding, so the outcome flows back into the audit ledger
Labs show lactate 3.1, HR 112, temp 38.6°C. Based on your clinical judgment, please clarify the condition being treated:
Payer-aware denial prevention
Everyone optimizes the chart. Almost no one models how the specific payer will adjudicate it. Medically scores each claim against that plan's rules and tells your team exactly which evidence is missing — before submission, when you can still fix it.
- Checks medical necessity, level of care (including the two-midnight rule), and payer policy in real time
- Denial-risk prediction that learns from your own remittance (835) and denial history, by payer
- Flags the exact gap, not just a risk score — and holds or strengthens the claim before it goes out
- Closed loop — appeal outcomes teach the system upstream
Denials & appeals that finish the job
Most tools stop at a letter you copy and paste. Medically reads the denial, assembles the evidence from the original pre-bill finding, drafts the appeal with the payer's own policy cited, submits it electronically, tracks the deadline, and reports overturns by payer.
- Denial-letter intake reads faxes and low-quality scans, extracting the rationale, the disputed codes, and the deadline
- One finding, whole lifecycle: a persistent finding ID means the appeal auto-assembles the evidence you cited pre-bill
- Appeal letters drafted with citations to clinical evidence, coding guidance, and payer policy — edited and approved by your team
- Electronic submission through your clearinghouse, with status tracking, deadline alerts, and peer-to-peer prep sheets
- Overturn analytics by payer, denial type, and DRG — and outcomes feed back so diagnoses that keep getting denied are flagged for stronger documentation before billing
Outpatient, professional-fee & risk adjustment
The same engine reads outpatient and clinic encounters: HCC capture with MEAT evidence for risk adjustment, E/M and ED facility leveling, and missing or mis-leveled charges — with conservative, audit-aware guardrails, because E/M and ED leveling are active OIG audit areas.
- HCC / RAF capture with MEAT (Monitor, Evaluate, Assess, Treat) evidence on every suggested condition
- E/M and ED facility leveling with the documentation elements shown — and a lower level when the record supports it
- Missed or mis-leveled charges (CPT/HCPCS/APC), including infusion and injection hierarchies, prioritized by encounter
- Quality and registry capture from the same clinical truth — no second review pass
The Medical Scribe
Documentation shouldn't be the reason your physicians stay late. They dictate or type the way they think; Medically turns it into a clean, structured note — SOAP, H&P, progress, consult, or discharge — in seconds, ready for review.
- Speech-to-text built for clinical language
- Multiple note templates — pick one, generate, edit
- Context-aware nudges for diagnosis completeness and specificity inside the note, so gaps close before a query is needed
- Cleaner notes upstream mean fewer queries and denials downstream
Subjective
68 y/o F presents with 3 days of progressive dyspnea and bilateral lower-extremity edema. Known CHF (EF 30%).
Objective
BP 152/94, HR 96, SpO₂ 88% RA. BNP 1,840. CXR: bilateral pulmonary edema.
Assessment
Acute on chronic systolic heart failure exacerbation.
Plan
IV furosemide, strict I/O, daily weights, telemetry.
Leadership analytics
See realized revenue — billed and paid, not projected — traced from finding to claim to remittance. Every number links back to the chart that produced it. The report your board asks for, ready before they ask.
- Realized-revenue ledger: each finding tracked surfaced → accepted → coded → billed → paid, including dollars later denied
- CMI, CC/MCC capture rate, acceptance rate, query response, and denial rate — by facility, service line, payer, and reviewer
- Precision and acceptance by finding type, plus defensibility distributions — the compliance view of the same data
- Quality flags — PSI/HAC, POA, HCC capture — as they happen
Ask Medically
Every chart already holds the answer — Ask Medically lets your team get to it in plain language. Ask a question about the record in front of you and get a response tied to the exact chart text behind it. No hunting, no leaving your workflow, and nothing you have to take on faith.
- Ask in plain language — no query syntax, no dashboards to learn
- Every answer cites the exact labs, vitals, and note text behind it
- De-identified before analysis — the chart is redacted first, every time
- Scoped to one chart, inside your account — nothing leaves your environment
The record supports a query. Two SIRS criteria are met — HR 112 and temp 38.6 °C — with lactate 3.1. Sepsis and its acuity aren't documented yet.
To assign A41.9, the provider needs to document “sepsis” and its acuity. The evidence above supports a compliant, non-leading query — Medically can draft it.
Redact
Direct identifiers are minimized and masked before the chart ever reaches the model.
Ground
The model reads only that de-identified chart — nothing else in your environment.
Answer
You get a plain-language answer with the exact chart evidence cited beneath it.
Decide
Processing is transient — not retained beyond the request, never used to train AI. Your team makes the call.
What your team can ask
- “Is there evidence to support acute respiratory failure?”
- “Which Sepsis-3 criteria are met for this payer?”
- “Is the heart failure documented to the right specificity?”
- “What’s the best-supported principal diagnosis?”
- “Does the evidence justify this MCC?”
- “Is anything here at risk of down-coding?”
- “Summarize this patient’s hospital course.”
- “What am I being queried about, and why?”
- “What evidence supports this diagnosis?”
- “What’s missing to defend this DRG on audit?”
- “Why is the defensibility score 31 on this finding?”
- “Which findings still need a human sign-off?”
Illustrative examples. Every answer is grounded in the specific chart in front of you and cited to its source evidence — nothing is asserted that can’t be traced to the record.
Live in weeks — with proof before you go live.
A read-only HL7/FHIR connection, no rip-and-replace, and a free retrospective look-back on charts you've already billed — so you see the exact findings, cited to the record, before the first live discharge.
Connect
Read-only connection to your EHR through pre-built HL7 v2 and FHIR R4 connectors (or a secure flat-file feed). A signed BAA is in place before any PHI moves.
Look-back
We run 90 days of already-billed charts and show every finding your team — and any existing vendor — missed, in both directions, each one cited and scored. No obligation.
Go-live
Live pre-bill review inside your bill-hold window. Your reviewers work the ranked worklist in the workflow they already know; nothing changes for physicians until you choose.
Expand
Add coding, denial prevention, appeals, outpatient and risk adjustment, or the Scribe — on your timeline, on the same engine and the same evidence.
Target go-live: six weeks or less from a signed BAA. Implementation is scoped and written down before kickoff, so your IT team knows exactly what is being asked of them.
Enterprise-grade security. Defensible by design.
HIPAA compliant · SOC 2 Type II on our roadmap
PHI handled to the standards your security and compliance teams require. BAA available on request.
Read-only by default
Medically reads the record; humans approve every change. Nothing is written back without sign-off.
Audit ledger & OIG-ready export
Every finding carries its evidence, criteria, score breakdown, disposition, reason code, and model version — exportable on demand.
Anti-upcoding guardrails
Two-way findings, bias and drift monitoring, never-fabricate guarantees, and lower codes when the evidence warrants.
Works inside the EHR you already run.
Pre-built HL7 v2 and FHIR R4 connectors mean a read-only connection and a go-live measured in weeks — not a rip-and-replace. Findings are delivered in your CDI and coding workflow; write-back happens only with a person's sign-off.
See the whole platform on your charts.
Drop your work email and we'll walk your team through the platform on a 30-minute demo — no slideware.
Start with a $0 pilot — pay only on revenue we recover. See pricing →
Have questions first? Contact us →
Product screens on this page are illustrative. Patient details are fictional; values show the shape of the output, not live data.