Medius’ AI advantage is that AI is embedded into finance workflows versus being layered on top.
With its agentic, event-driven core, Medius captures the actual events that take place in the platform.
This event-driven architecture means that AI agents can take direct action, not just surface recommendations.
Over the last 10 years, Medius has accumulated 2.4 billion+ invoice field data points, with 17%+ of that dataset (393 million+ fields) derived from real-world human corrections. This combination of proprietary training data, real workflow outcomes, plus an agent-ready infrastructure is structurally impossible to replicate.
Medius’ platform enables outcomes like higher touchless rates, faster processing, and less reliance on external variables (like supplier adoption), which is where many competing platforms still struggle.
10+ yrs
Data science and AI development
2.4B+
Human validated invoice data points
~1000x
Faster than off-the-shelf LLM
96.3%
Touchless processing rate
12,000
Agent-handled emails / month
Invoice received
PDF attachment via email
Audit trail
Invoice ready for connection
Cross-referencing required for PO #8892
Match successful.
No manual deviations.
Deviation detected
Invoice amount exceeds PO #4451 by 15%
Routed for human approval with guidance.
Batch ready for review
Signals and anomalies detected
Batch optimized. Risk flagged and logged.
Supplier statement received
EOM statement – Blue Steel Inc.
Statement processed. Exceptions flagged.
Data extraction and validation often require the most manual work when processing an invoice. Medius Capture has been refining its AI and machine learning algorithms since 2016, learning from hundreds of millions of invoices. The result: a 96.3% touchless processing rate for PO invoices and 99.5% for non-PO invoices powered by a proprietary multi-stage pipeline of Siamese CNNs, tree-based ensembles, and Markov models that run ~1000x faster than large language model (LLM) alternatives.
Medius AP Automation features SmartFlow, a powerful proprietary CNN that auto-fills coding, tax, and approver values for non-PO invoices with 95%+ precision after just two invoices.
Automatically match invoice line details against purchase orders, goods receipts and/or contracts and resolve discrepancies in invoices, POs, and receivables. Avoid fraudulent invoices, pay the correct amounts and safeguard your assets using N-way matching to validate quantities, prices and taxes.
Medius Copilot autonomously guides approvers through the invoice process, empowering them to make confident, accurate invoice decisions for increased on-time payments while reducing the burden on AP teams. Medius Copilot delivers useful information about an invoice through pre-defined prompts for popular invoice processing queries. Approvers can also enter free-text questions and Medius Copilot will deliver highly accurate and informed responses.
Medius Fraud & Risk Detection intelligently identifies and surfaces potential bad actors and errors to protect your business against fraud and risk, with complete oversight through the Fire Station console and reporting capabilities. AI-powered behavioral monitoring and anomaly detection reduce payment fraud and error rates by 90–95%, tracking every instance of potential fraud across the invoice lifecycle and surfacing mitigations before they impact cash.
Building a payment batch means hunting the same signals every time: comments buried in the invoice, cash discount deadlines, a strategic payee list kept in someone's spreadsheet. The Payment Optimization Agent gathers them and surfaces each one as a recommendation inside the batch, with its reason and its named source. It recommends, you review, and nothing is added without your click.
AI automatically matches supplier statements and invoices and spots missing or mismatched invoices to help prevent late payments, duplicate payments, over payments or fraudulent payments. Deliver a level of accuracy manual reconciliation methods can’t touch with AI that automatically extracts the data you need from supplier statements and eliminates the need for manual checks and reviews.
Give vendors peace of mind with autonomous invoice and payment status updates right to their inbox. Supplier Conversations responds to supplier email inquiries—classifying intent, retrieving context, and responding autonomously within defined guardrails.
PO invoices - Touchless processing rate
PO invoices - Total invoice processing time
1 Ardent Partners' Accounts Payable Metrics that Matter in 2025
2 Medius Research AP Automation Benchmark Report 2025
Medius has more of the AP and payments lifecycle automated with AI than any competitor.
AI is embedded across the lifecycle doing the work.
| Capability | Medius | Basware | Tipalti | Esker |
|---|---|---|---|---|
| AI invoice capture & coding | SmartFlow auto-fill (95% precision); 99% header/line accuracy | SmartPDF + SmartCoding agents | Invoice Capture Agent + GenAI GL coding | Predictive coding + intelligent extraction + smart batch split |
| PO / 2–3-way matching with AI | Auto-match including multi-system POs | Smart Matching across systems & imperfect figures | PO Matching Agent (contextual) | Rules + AI matching with centralized exception handling |
| Conversational AI / approver copilot | Medius Copilot (OpenAI-based) for approvers | AP Business Agent + AP Data Agent shipping; AP Pro Agent announced 2026 | Tipalti AI Assistant (multi-model) | Esker Synergy Assistant (Azure ChatGPT) |
| Supplier inbox / inquiry AI | Supplier Conversations: ~94% inbox time reduction | Supplier Agent — voice + summary, planned 2026 | AI Assistant can answer status questions; no inbox auto-reply named | Triage + sentiment + GenAI replies (95% accuracy) |
| Statement reconciliation AI | AI auto-matches statements; flags missing/mismatched invoices; tied to Supplier Conversations | Statement Matching (acquired AP Matching) — claims up to 90% manual work removed | Payment & transaction reconciliation; no published supplier-statement product | Synergy AI extracts statement data and reconciles to invoices |
| Fraud & risk detection AI | AP fraud & risk AI: anomalies, supplier bank-detail changes, 4-eye policy enforcement; 90-95% reduction | Trust-by-design controls + autonomy gates; no published fraud-specific agent | Tipalti Detect: payee screening, duplicate bill + anomaly flags | Anomaly detection across AP; no published fraud-specific agent |
| Payment automation & optimization AI | Payment Optimization Agent: recommends what to pay and when, capturing discounts and flagging supplier risk, each with its reason and source | Basware Discount + Discount Management; no published optimization agent | Mature global payments + early-pay / dynamic discount capture | Esker Pay with dynamic discounting + LSQ supply chain finance; no published optimization agent |
Medius has more of the AP and payments lifecycle automated with AI than any competitor.
AI is embedded across the lifecycle doing the work.
Following the independent evaluation from leading analyst firm Ardent Partners, Medius has been named both a Leader and an Elite Performer for AI and Innovation the 2026 AP Automation and Payments Technology Advisor.
This recognition highlights the work of the Medius product team to deliver real value through the power of AI that empowers finance teams of the future. Ardent Partners notes that Medius is ahead of the curve, with advanced AI capabilities already in use today and delivering measurable impact across AP and payments.
Medius makes AP autonomous – using AI and ML built over 25+ years of domain expertise. The architecture, built AI-native since 2016, is designed to understand context, make decisions, and take action within clear controls.
Agents do the work.
Routine work gets completed without manual effort. Touchless processing grows from AP to cover the entire spend lifecycle.
Agents apply the policy.
Invoices, payments, supplier data, and contract terms continuously checked against finance-defined rules.
Agents support the call.
When to approve. When to pay. Which discounts to capture. The Payment Optimization Agent brings all of it into the batch, with the reason and the source attached.
