# BLP Digital

> Runs the recurring document processes around an ERP system, namely incoming invoices, order confirmations, delivery notes and master data upkeep, and writes the result back into the ERP, which remains the system of record.

- Vendor: BLP Digital AG, Zurich
- Canonical URL: https://www.convios.com/en/toolbox/blp-digital
- Language version: https://www.convios.com/de/werkzeugkasten/blp-digital
- Area: Operations & finance · Cluster: document processes around the ERP
- Role: Off-the-shelf product · Origin: AI-native
- As of: 2026-07-31 · Reviewed: 2026-07-31 · Author: Dr. Oliver Gausmann, Convios GmbH
- Toolbox: https://www.convios.com/en/toolbox — Markdown: https://www.convios.com/en/toolbox.md

## Verdict

Efficiency only: The tool lowers the cost per document. It lowers it at a point no customer ever sees and where no vendor sets itself apart from its competitors, because incoming invoices and delivery-note checks work the same way across the field. Above a certain document volume that advantage turns into a precondition, because competitors are doing the same arithmetic. Bought stays bought: whatever is gained here, anyone else can order too. So it is efficiency in a single case, and the single case is a large one.

## Suitability by company size

- Solo: not suitable — There is no list price, no self-serve access and no trial, the way in runs through an appointment. Anyone without an ERP carrying meaningful document volume has nothing to automate that would justify a rollout of roughly three months.
- Mid-market: suitable with caveats — The published references are mid-market, among them Witzig, Nussbaum and Pfeifer. The way in is still a project rather than a subscription: the vendor cites a production-ready proof of value in under a week and a full implementation in about three months. Without someone who knows and can name your own exception cases, the automation stops at precisely those cases.
- Enterprise: suitable — According to the vendor, Migros rolled the tool out across all its cooperatives within a year alongside the S/4 migration, and the integrations page lists more than 30 ERP systems plus the vendor's own SAP package built on standard BAPIs. Vendor assessment will still be work: storage location, processing agreement, certifications and audit-log scope are not publicly substantiated and belong in the contract in full.

## Vendor staying power

Funded: The company has existed since 2019, is based in Zurich and is entered in the commercial register of the canton of Zurich. The 50 million dollars from Goldman Sachs Alternatives in April 2026 are explicitly a secondary transaction, so the money went to existing shareholders rather than into the company. The vendor describes itself as profitable; no published accounts back that up.

- Legal entity and commercial register: BLP Digital AG, Schützengasse 16, 8001 Zurich, company number CHE-295.990.745, legal form stock corporation (Aktiengesellschaft), status registered, most recent register entry 12 May 2026 (source: https://zh.chregister.ch/cr-portal/auszug/auszug.xhtml?uid=CHE-295.990.745, as of 2026-07-31)
- Founding, founders and headcount: Founded in 2019 by Tim Beck, Sven Beck, Sabrina Schenardi and Thore Harmuth; grown from 12 to more than 150 people in five years by the company's own account (source: https://blp.digital/company/about-us, as of 2026-07-31)
- Investment: USD 50 million from Growth Equity at Goldman Sachs Alternatives, announced on 22 April 2026 and expressly described as a secondary transaction in the investor's release (source: https://am.gs.com/en-us/advisors/news/press-release/2026/goldman-sachs-alternatives-invests-50m-blp-digital, as of 2026-07-31)
- Customer count over time: more than 450 corporate customers in over 40 countries (release of 22 April 2026), more than 550 customers and over 1,300 projects (release of 29 July 2026) (source: https://blp.digital/550-customers-milestone, as of 2026-07-31)
- Group entities and named officers: Establishments as BLP Digital GmbH in Schorndorf, BLP Digital UK Ltd in London and BLP Digital US Inc. in New York; data protection officer Tim Groeger, Article 27 EU representative Maximilian Rothe (source: https://blp.digital/privacy-policy, as of 2026-07-31)

## Cost of leaving

Moderate: The ERP remains the system of record, so master data and postings sit exactly where they sat before after a switch. What goes is the other half: the exception cases the system has learned, the way the agents are cut, and the document archive inside the platform. Terms of service, notice periods and any route to retrieve that half are described nowhere publicly. The exit is decided in the contract rather than in the technology, and it belongs on the table before signing.

## Regulation and data

| Point | Finding | Evidence | As of |
|---|---|---|---|
| Data processing agreement | unclear — The privacy policy states that the vendor concludes processing agreements with its own service providers. Whether it offers one to its customers in return is stated nowhere, and no document is publicly available. For a tool that processes other companies' business documents this is the core question: procurement cannot see publicly what it would be able to rely on. There are contacts, namely privacy@blp-digital.com and a named data protection officer. | partially evidenced | 2026-07-31 |
| Storage location | unclear — No data centre, no operator and no region is named. For personal data the vendor controls itself, the privacy policy names Switzerland, Germany, the United Kingdom and the United States. Where customer documents are processed and stored it does not say, and there is no security or trust page. That leaves the processing impossible to locate. | not evidenced | 2026-07-31 |
| Subprocessors | no list, categories only — Categories are named instead of companies: hosting and infrastructure, analytics, CRM, email and communications, support tools. The only named service in the legal texts is Google Tag Manager, which belongs to the marketing website. Who sits behind the processing of the documents, and in particular which model providers run the agents, remains open. | partially evidenced | 2026-07-31 |
| Third-country transfer | yes, United States, basis named — The privacy policy names Switzerland, Germany and the United Kingdom as mutually recognised adequate, and the United States as a third country, partly because of the vendor's own New York entity. As a basis it names the Commission's standard contractual clauses, the UK International Data Transfer Addendum and the clauses recognised by the Swiss commissioner. Which recipient relies on which basis is not stated, and the commitment covers the data the vendor controls, not visibly the customers' documents. | partially evidenced | 2026-07-31 |
| Training on customer data | unclear — No statement, in none of the public documents. The product has agents work on other companies' business documents, and the vendor expressly describes a human explaining the context of a special case to the machine and thereby training it further. That is exactly where a commitment would be needed on whether and where what has been learned stays, and whether it reaches beyond the individual customer. It is missing. | not evidenced | 2026-07-31 |
| Retention and deletion | periods only for the vendor's own contract and website data — The privacy policy quantifies what concerns the vendor itself: contract and account data for the duration of the relationship and then the statutory periods, in Switzerland generally up to ten years; enquiry data for a reasonable follow-up period; marketing data until withdrawal. For the documents the platform processes it names neither a period after the contract ends nor any commitment to return them. | partially evidenced | 2026-07-31 |
| Certifications | none named — Neither ISO 27001 nor SOC 2 nor any other attestation is named anywhere. There is no security or trust page: eleven likely addresses were checked and all return 404, and the vendor's sitemap lists no legal or security page beyond privacy, cookies and imprint. The privacy policy describes general measures, namely encryption in transit, access controls and regular review. | not evidenced | 2026-07-31 |
| EU AI Act, Article 50 | unclear — The product contains an AI system within the meaning of the regulation, so the transparency duties under Article 50 apply from August 2026. No public document carries a statement from the vendor on this. In practice the duty mostly falls on the deploying company anyway, because the agents work in-house rather than facing end customers. The gap is still notable, because section 11 of the privacy policy grants a right to human intervention and to information about the logic involved, yet never mentions the regulation. | partially evidenced | 2026-07-31 |
| Audit logging | promised by the vendor, scope and period open — Full traceability and clear ownership of exceptions are a load-bearing part of the sales promise. What the log captures, how long it is kept and whether it can be exported into your own systems is stated nowhere. For a tool that triggers payments and postings this is the point an audit will later start from, and it cannot be planned from public information. | partially evidenced | 2026-07-31 |

## Cost

- Entry: No list price. The website carries no pricing page and the vendor's sitemap shows none either; the way in runs solely through a demo appointment. Instead of prices the vendor cites timings: a production-ready proof of value in under a week, a full implementation in about three months. (as of 2026-07-31)
- Where it gets expensive: In three places that belong on the table before signing. First the pricing key, which is not publicly known: whether billing runs per document, per process or per seat decides whether the invoice grows with the business or with the benefit. Second the integration: more than 30 ERP systems are listed, but yours sits there in a particular version, with its own fields and its own legacy decisions, and reconciling that is project work rather than configuration. Third the exceptions, which is precisely what the vendor sells: every new type of exception means training, checking and sign-off again, and that work lands in your own department, not with the vendor.

## Three routes compared

### Stay with the ERP vendor and take its own automation

SAP, Microsoft and Oracle have their own modules for the same document processes, and the classic document-capture vendors have been in this field for decades. This route buys contractual readiness and proximity to the system of record: one contracting party, one audit trail, certifications procurement already knows and has signed off before. You pay for it with the very thing the AI-native vendor names as its own point of entry. Standard automation handles the normal case; on the deviations the work falls back to people, and in document processes the deviations carry the expensive part of the work.

### Put an AI-native layer on top of the ERP

This is the route the tool offers, and the category has more than one vendor. So there are two things to test, and neither of them shows up on a feature list. First: how deep does the connection to your own ERP actually reach, meaning the write path, the version and the field mapping, not just a logo on an integrations page. Second: what is contractually substantiated when storage location, processing agreement, certifications and audit-log scope stay publicly open. A vendor processing payment-related documents without a security page moves that whole review into the contract negotiation.

### Build the run yourself

Technically the run is well known: take in the document, read out the fields, check against master data, post into the ERP, hand doubtful cases to a person. You build it with a build tool such as Claude Code, an automation chain such as n8n, LiteLLM as the model gateway and Langfuse for traceability and the evaluation loop. Building is not the problem. Running it is: a process that triggers payments needs on-call cover, approval logic, tenant separation and evidence of why a machine decided the way it did. And it needs someone to maintain it when the builder leaves the company. It pays off when a single document type carries most of the volume and your own checking rule is the actual value. It does not pay off when thirty document types from twelve countries run through four ERP versions.

Recommendation by size:

- Solo: Do not buy, the document volume is not there.
- Mid-market: Start with one document type and fix the pricing key before the rollout.
- Enterprise: Buying can be right, but settle storage location, processing agreement, certifications and audit-log scope in writing before any rollout.

## Context

- Implements method: [Constraint Detection (Five-Layer Scan)](https://www.convios.com/en/methods/constraint-detection) — The method looks for the layer where work actually stalls, and this tool pays off precisely when the bottleneck sits in document throughput rather than in demand or in the decision above it.
- Implements method: [Randomized Measurement of AI Productivity Gains](https://www.convios.com/en/methods/ki-produktivitaet-rct) — The vendor claims productivity of five to ten times manual processing, and the method describes how to measure a number like that in your own operation instead of taking it from the sales deck.
- Implements method: [Operating Model Grid (Standardization and Integration)](https://www.convios.com/en/methods/operating-model-raster) — The method separates standardisation from integration, and that is exactly what the decision here turns on: the ERP stays the system of record while execution above it gets standardised.
- Displaces: Keying documents by hand and posting them into the ERP afterwards, Working exception cases out of shared mailboxes and spreadsheets, Hiring more people whenever document volume rises

## Evidence

- Legal entity, registered office, company identification number and contact details — https://blp.digital/imprint (as of 2026-07-31)
- Legal form, register status and most recent register entry in the commercial register of the canton of Zurich — https://zh.chregister.ch/cr-portal/auszug/auszug.xhtml?uid=CHE-295.990.745 (as of 2026-07-31)
- Group entities, data protection officer, Article 27 representatives, third-country transfers and their basis, retention, security measures and information on automated decisions — https://blp.digital/privacy-policy (as of 2026-07-31)
- Founding year, founders, locations and headcount — https://blp.digital/company/about-us (as of 2026-07-31)
- Investment of USD 50 million by Growth Equity at Goldman Sachs Alternatives on 22 April 2026, expressly a secondary transaction, plus 450 customers in over 40 countries — https://am.gs.com/en-us/advisors/news/press-release/2026/goldman-sachs-alternatives-invests-50m-blp-digital (as of 2026-07-31)
- The vendor's own statements on profitability, growth from 12 to more than 150 people, proof of value in under a week, implementation in about three months, traceability and training by subject-matter experts — https://blp.digital/goldman-sachs-alternatives-invests-in-blp (as of 2026-07-31)
- More than 550 customers in over 40 countries, over 1,300 projects, named references Migros, Krone and Testo — https://blp.digital/550-customers-milestone (as of 2026-07-31)
- ERP systems listed by name, connection types, the SAP package built on standard BAPIs and direct database access for Abacus — https://blp.digital/integrations (as of 2026-07-31)
- The product's structure of specialised agents, orchestration and a digital twin, with the ERP remaining the system of record — https://blp.digital/product (as of 2026-07-31)
- No pricing page, no terms of service, no processing agreement and no security page anywhere in the vendor's own sitemap — https://blp.digital/sitemap.xml (as of 2026-07-31)