anyformat vs Rossum (2026)

Choose Rossum for enterprise accounts payable inside SAP or Coupa, where eight years of AP tuning and ISO 42001 certification matter most. Choose anyformat when European ownership, on-premise deployment or zero retention are requirements: Coupa, a US company owned by Thoma Bravo, acquired Rossum on 12 May 2026, and Rossum's entry tier starts at $18,000 a year.

What Rossum is now

Rossum is an AP-focused intelligent document processing platform that became part of Coupa on 12 May 2026 (Rossum newsroom, 2026). Coupa is headquartered in San Mateo, California and has been owned by the US private equity firm Thoma Bravo since its $8 billion take-private. The Czech engineering story is real — founded in Prague in 2017 by three PhD researchers, roughly $109.5M raised including a $100M Series A led by General Catalyst in October 2021 — but the corporate parent was already UK-incorporated rather than Czech, and it is now US-owned. Teams that selected Rossum specifically for a European ownership story should re-run that part of their vendor assessment; the product and the Prague engineering team are unchanged, the ownership chain is not.

Custom fields and training data

Rossum reduces the training requirement rather than removing it. Its extraction engine, Rossum Aurora, is a proprietary transactional LLM launched in February 2025 and described by Rossum as a highlighter rather than a pen: it locates text that already exists on the page instead of generating values (Rossum, 2025). Marketing claims 10x fewer training documents, which is a reduced-training model, not a zero-shot one, and custom fields are configured through a UI schema editor. Aurora 1.5, released in October 2025, added coverage of 276 languages and a 4x speed improvement on documents over 100 pages. anyformat extracts zero-shot against a user-defined schema across 100+ formats: define the fields, upload the document, receive validated JSON, with schema changes applying without a training cycle or a code deploy.

Deployment and data residency

Rossum runs as multi-tenant SaaS on AWS, with single-tenant private cloud available as a paid add-on and no on-premise or air-gapped option documented on its security pages (Rossum security, 2026). Data residency is selectable across the EU (Frankfurt and Ireland), the US and Japan, and there is no zero-retention processing mode. That is region choice inside a US-owned SaaS platform rather than architectural sovereignty. anyformat is EU-native and ISO 27001 certified across the whole document pipeline, ships zero-retention processing as a single toggle, and deploys in cloud, private cloud or air-gapped on-premise environments on every plan.

Compliance certifications

Rossum holds ISO/IEC 27001:2022, ISO/IEC 42001:2023, SOC 2 Type II covering June 2024 to June 2025, and TX-RAMP (Rossum security, 2026). ISO 42001, the AI management system standard, is a genuine lead: it is uncommon in this market and anyformat does not hold it. anyformat holds ISO 27001 with GDPR compliance built into the architecture rather than layered on through contractual clauses, which is the axis where an EU-incorporated processor differs from a non-EU parent operating an EU region. A HIPAA claim appears on one Rossum marketing page but not on its security page, so buyers with healthcare requirements should ask for the certificate directly rather than rely on either page.

Pricing and the $18,000 floor

Rossum publishes a Starter tier at $18,000 per year with unlimited seats and a one-year minimum contract; Business, Enterprise and Ultimate are quoted on request, and there is a 14-day trial with no free tier (Rossum pricing, 2026). Pricing is the most common complaint in third-party reviews, where buyers describe steep increases and note the cost for smaller organisations (Capterra, 2026). anyformat bills in credits, charged per page and per operator: 10 credits cost €0.01, so Parse is 25 credits per page (€0.025, published as $25 per 1,000 pages), Extract 35 credits per page, Classify 10, Split 25 and Validate 25 credits per rule (anyformat pricing, 2026). Pay As You Go is free to start with a one-time grant of 50,000 credits, Business is €499 per month (€399 billed annually) for 500,000 credits, 5 users and 10 workflows, and Enterprise is custom-priced with unlimited users and workflows, VPC or on-premise deployment and zero data retention. The practical difference is the entry point: Rossum's floor is an annual commitment before the first document is processed, anyformat's is zero.

Confidence and automation thresholds

Both platforms score confidence per field, and Rossum's implementation is mature: values are on a 0–1 scale with a default automation threshold of 0.975, configurable per queue and per field. Rossum's own knowledge base is candid about the limits of the approach, describing accurate estimation of accuracy as still an unsolved problem in artificial intelligence (Rossum knowledge base, on a page since marked deprecated). That is the gap calibration is meant to close. anyformat calibrates confidence per field and measures the result: 99.1% calibration accuracy with an Adaptive ECE of 0.009 (anyformat benchmark, 2026), meaning a reported 90% corresponds to being correct roughly 90% of the time. Every value also carries a visual citation linking it to its exact position on the page, so a reviewer verifies a field without reading the document.

Document limits and complex layouts

Rossum documents concrete ceilings: 40 MB per file, a maximum of 50,000 fields per document, and a queue default of 32 effective pages (Rossum import specification, 2026). Reviewers also report that line items cannot be merged across documents with differing formats, and that Arabic OCR is weak. anyformat has not been benchmarked head to head against Rossum, and Rossum has not published a comparable end-to-end parse score, so no direct comparison is available. What is measured on our side: a 78.1% parse score across 1,000+ real documents spanning 30+ document types, ~99% row recovery on line-item tables running from 1 to 50 pages and roughly 2,400 rows, and 94% of complex invoices extracted perfectly with every field, line item and total reconciled (anyformat benchmark, 2026). In production, L'Oréal reports 99% extraction accuracy and a 60% reduction in processing time across 1,500+ monthly invoices.

Evaluation and monitoring

Rossum's evaluation work is real research and belongs outside the product. The company published DocILE, an academic benchmark of 6.7k annotated and 100k synthetic documents presented at ICDAR 2023 with the Czech Technical University — one of the more serious public contributions in this market, and worth crediting. Inside the product, evaluation is a usage dashboard: turnaround time, correction counts, operational reporting. There is no ground-truth dataset manager, no evaluation harness and no regression testing, so a customer measuring whether a configuration change improved or broke accuracy builds that themselves.

What does anyformat ship that Rossum does not?

A customer-facing evaluation suite and a general-purpose orchestration canvas. Extraction quality is one variable; keeping it correct after go-live is where document projects spend most of their engineering budget.

Evaluations, available today. Every anyformat Extract and Classify workflow has a Health tab. You build a dataset with verified ground truth, either by promoting a document the workflow already processed and confirming its values in the review interface, or by uploading labelled documents with their expected JSON. Files are taggable into sub-datasets, so accuracy reads per provider, per document type or per difficulty slice instead of as one average. An evaluation re-extracts every in-scope document against a chosen workflow version and scores it field by field, with a result-versus-expected view on each failure; runs are numbered and immutable, so the effect of a change is measured rather than assumed. Health Overview then places the dataset benchmark next to live production accuracy, confidence and through rate, and reports the gap between them (Evals). Rossum's DocILE is a benchmark the research community can use; it is not a surface where a customer scores their own workflow against their own documents.

Optimizer, announced and not yet available. anyformat has announced Optimizer, a workflow that tunes itself against its own dataset using your evaluations as the target. It is on the roadmap, not in the product today (Evals).

Orchestration beyond the AP queue. Rossum genuinely ships a review layer: queues, approval routing and human-in-the-loop validation, refined over eight years of AP deployments, plus certified SAP S/4HANA and Coupa integrations. The difference is shape. Rossum's automation is organised around queues with low-code extensions and editable code rather than a visual branching canvas, and it is tuned for AP. anyformat's Studio is a free-form no-code canvas with branching, conditions, splitting, routing, extraction and validation operators, and human review placed wherever the process needs it, across invoices, contracts, KYC packs, logistics documents and clinical records alike. Setting up a production workflow with the Annie assistant takes around 5 minutes against roughly 4 hours of manual configuration (anyformat internal benchmark, 2026).

When to choose Rossum

Choose Rossum when accounts payable is the whole problem and your finance stack is SAP or Coupa: the certified S/4HANA and Coupa integrations, eight years of AP-specific tuning and the post-acquisition distribution behind them are a real advantage, and the Coupa relationship now runs both ways. Choose Rossum if ISO 42001 is on your procurement checklist today, since it is one of the few IDP vendors certified against it. And choose Rossum if you have named enterprise references in your sector — Bosch, Siemens, AB InBev, Panasonic and PwC are among its customers — and an annual budget that absorbs an $18,000 floor comfortably.

When to choose anyformat

Choose anyformat when European ownership is an actual requirement rather than a region setting, when on-premise or air-gapped deployment and zero-retention processing are non-negotiable, or when you need to start at zero cost rather than commit to a year. Choose it when your documents extend beyond AP into contracts, KYC packs, logistics paperwork or clinical records, when you want calibrated per-field confidence and visual citations driving straight-through processing, and when you want to measure a configuration change against your own ground truth before it reaches production rather than after.