Herman
Herman is the AI credit analyst from Oper Credits, built for regulated European mortgage lenders. He classifies documents, verifies every data point and applies your own credit policy, then hands analysts a decision-ready case; the lending decision stays human.
What is Herman?
Herman is an agentic AI product from Oper Credits BV, an Antwerp software company that sells to regulated European banks and lenders. Oper presents Herman as an AI credit analyst: not a decision engine, but the layer that turns a raw mortgage dossier into a file an analyst can decide on. It sits alongside Oper's other line, the white-label Origination Platform, but Herman is the flagship the company leads with.
The published workflow has four stages. Herman first collects and classifies, taking documents in from a borrower portal, an email inbox or an API call, then splitting, labelling and filing every piece of the dossier. He then extracts and verifies, validating fields and cross-checking the application against its supporting documents, flagging whatever is missing early enough to be sent back to the broker or the branch. He then applies the lender's own written credit policy, using either a full or a simplified set of checks. Finally he produces a decision-ready output: a structured case report carrying a complete verified dataset, the policy results with their reasoning, a recommendation, and the audit trail behind every step.
Oper is emphatic about what Herman does not do. He performs no credit scoring: he applies a written policy identically to every file rather than producing an opaque score on borrower characteristics, and the final decision always stays with the lender's team. The company frames this as a glass box, not a black box, with every recommendation linked to its evidence and its policy reference, and every step logged.
Integration is designed to avoid replacement. Herman runs on top of an existing loan origination system and document management stack with no core migration, through a REST API and SDK with event streams, through email and Microsoft Teams, or through a Computer Use mode that drives systems lacking an API the way a human would. Processing runs in European cloud regions, Frankfurt and Switzerland by default, or inside the customer's own environment. Oper claims case preparation falling from 120 minutes to 3, extraction accuracy above 95%, and up to 90% of underwriting tasks handled without manual intervention, with more than twenty European lenders using its software.
What it does
- Collects and classifies an entire mortgage dossier from a portal, an email or an API call, splitting and labelling every document
- Extracts data fields and validates them, cross-checking the application against its supporting documents
- Flags missing pieces at intake and sends them back to the broker or branch before they block underwriting
- Applies the lender's written credit policy the same way to every file, with full or simplified check sets
- Produces a structured case report and a recommendation, each conclusion linked to its evidence and its policy rule
- Runs specialised document checks: identifier checksum validation, PDF tamper detection and salary verification against public registers
- Re-checks an existing portfolio against a policy change through sensitivity analysis
When to use Herman / When not to
A quick filter to help you decide if Herman is the right fit.
When to use Herman
- Mortgage underwriters and credit analysts at European banks drowning in manual file preparation
- Loan officers and mortgage advisers who need incomplete files caught before they reach underwriting
- Credit risk managers who want a written credit policy applied identically to every single file
- Compliance and operational risk teams at supervised lenders facing DORA, EBA outsourcing and EU AI Act scrutiny
- Integration architects plugging automation onto an existing loan origination system without a core migration
When not to use Herman
- Individual borrowers: Herman is sold to lending institutions, never to consumers
- Anyone wanting self-service: there is no sign-up, no public price and no free plan, only a sales-led demo
- Lenders outside Europe: the product targets regulated European institutions with Frankfurt and Swiss data residency
- Teams looking for automated credit scoring: Herman explicitly refuses to score borrowers
- Small brokers or lenders with no origination system, no written credit policy and no risk function to plug into
How to use Herman
A typical end-to-end flow, from setup to results.
- Book the free 20-minute walkthrough through the contact page; there is no self-service sign-up
- Send Oper a set of historical cases so Herman can be backtested on your own document types
- Review the backtest results before committing to anything further
- Move to a limited set of live cases inside your own environment, settling integration and data-protection questions in practice
- Choose an intake route: portal, email inbox, Microsoft Teams, or the REST API and event streams
- Where a target system exposes no API, fall back on Computer Use so Herman drives the interface directly
- Load your written credit policy so Herman applies the same checks to every file
- Set your retention period, or switch on the temporary mode that deletes everything after each run
- Consult the public developer documentation at developers.opercredits.com for the API and webhook reference
- Roll out fully, with every file processed through Herman inside your existing systems
Pros & Cons
Pros
- Plugs on top of an existing origination system with no core migration, with production promised in months rather than years
- Strong traceability: every recommendation is tied to its evidence and its policy rule, and every step is logged
- Human-in-the-loop by design, with no opaque scoring of the borrower
- European data residency in Frankfurt and Switzerland, with a single storage boundary inside the customer's tenancy
- Explicit commitment that no training datasets are derived from customer data
- Oper Credits holds an ISO 27001 certification and offers audit rights, an exit plan and incident reporting
- Public API documentation, including an OpenAPI 3.1 specification served without authentication
Cons
- No pricing is published anywhere: no page, no grid, no tier, everything routes through a sales demo
- No free plan and no self-service trial; the only free step is a 20-minute walkthrough and a backtest
- Strictly business-to-business, explicitly aimed at institutions and out of reach for an individual professional
- The published Terms of Use cover only the website, so the actual product contract is not public
- The subprocessor list is announced as available on the site, but the download link on the data processors page is empty
- Performance figures are vendor claims repeated across several pages, with no published methodology or third-party verification
- Little public product detail beyond the API reference: no screenshots and no functional documentation
Pricing & Plans
No price is published. Oper Credits operates a quote-based, sales-led model: the complete sitemap carries no pricing page under any name, and the Terms of Use confirm the arrangement by stating that any engagement is subject to a separate commercial agreement. There is no permanent free plan and no self-service trial. The only free element is the pre-sales sequence: a 20-minute walkthrough followed by a backtest of Herman on the prospect's own document types, at no charge. Prospective customers must therefore contact the company through the demo request form to obtain any figure.
Data, GDPR & hosting
A consolidated view of how Herman handles your data.
GDPR overview
GDPR implementation is documented in concrete terms. Oper Credits maps its roles explicitly, acting as controller for site visitors and for administrative data, and as processor for borrowers' mortgage-application data. The privacy notice lists the full set of rights (access, rectification, erasure, restriction, portability, objection, withdrawal of consent) with a one-month response commitment via privacy@opercredits.com, and states that no automated decision-making with legal effect under Article 22 takes place, since lending institutions make every decision. Consent for analytics and advertising cookies is managed through Cookiebot. Transfers outside the EEA are said not to occur in normal operation, falling back on adequacy decisions or Standard Contractual Clauses. The minimum age is 18. No Article 27 representative is designated, which is consistent with an editor established in Belgium.
Who owns the data?
Oper Credits splits its role in two. It is the data controller for website visitors and for administrative data about customers and end users, and the data processor for borrowers' mortgage-application data, which it handles strictly on the lending institution's instructions. Customer documents, extracted data and Herman's outputs stay inside one storage boundary called the Briefcase, within the customer's own tenancy, isolated per client with role-based access and logged production access. The published Terms of Use cover only the website, not the platform: they claim site content for Oper and its licensors, and state that any engagement is governed by a separate commercial agreement that is not public. Belgian law applies, with the courts of Antwerp.
Reuse rights
The published Terms of Use bind website visitors only, so they say nothing about reusing product output. On the site itself, reuse is narrow: you may view content for internal, non-commercial reference, but you may not copy, modify, distribute or reuse it without permission, and automated scrapers or bots are forbidden without prior consent. On the platform side, the customer keeps control: retention periods are customer-defined, a temporary processing mode wipes every document and output after each run, and deleting a Briefcase permanently removes its contents. Oper commits that no training datasets are derived from customer data. Anything beyond that, including reversibility and ownership of Herman's outputs, is negotiated in the separate commercial agreement.
Data retention & training
Hosting summary
Document processing and storage run in European cloud regions, with Frankfurt and Switzerland named as the defaults, and deployment inside the customer's own environment offered as an alternative. Data sits in a single storage boundary called the Briefcase, held within the customer's tenancy, with no separate AI database, no hidden caches and no telemetry copies. Each client runs in its own logical tenancy with role-based access control, limited and logged production access, encryption in transit at TLS 1.2 or above and at rest with AES-256. The privacy notice states that personal data is not transferred outside the EEA in normal operation, relying on adequacy decisions or Standard Contractual Clauses otherwise. Note a discrepancy worth raising: Switzerland, named as a default region, is not an EEA country. The governing law is Belgian, with jurisdiction in Antwerp.
Things to keep in mind
Risks and trade-offs to weigh before adopting Herman.
- Automation complacency: Oper insists on human-in-the-loop review, but an overworked analyst can slide into rubber-stamping recommendations rather than testing them
- The efficiency claims are vendor figures repeated across several pages, with no published methodology or independent audit to support them
- No product contract is public, so ownership of outputs, reversibility and liability are settled in a negotiation the buyer cannot preview
- The subprocessor list is announced in the privacy notice but the download link on the data processors page is empty, leaving a due-diligence gap
- A scope contradiction sits between documents: the privacy notice says all core infrastructure is inside the EEA, while the product pages announce Frankfurt and Switzerland, and Switzerland is outside the EEA
- No training opt-out is documented; Oper states that training on customer data never happens, which is a favourable default rather than a right the customer can exercise
- Beware of attribution: the Mortgage Credit Directive, DORA, NIS2, eIDAS, the EU AI Act and BaFin or BWG expectations are frameworks the software helps banks meet, not credentials held by Oper, which is explicitly not a credit institution or a supervised financial services provider
Setup & Integrations
Technical difficulty
There is no self-service installation. Deployment is vendor-led in three phases, from a backtest on historical cases to limited live cases and then full rollout, with production promised in months rather than years and new use cases in weeks afterwards. The technical effort depends on the route chosen: email or Microsoft Teams requires almost no development, the REST API with SDK and event streams requires work on the origination system side, and Computer Use covers systems with no API at all. No core migration or replacement of the existing origination system is required.
Deployment
Integrations
Behind Herman
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Who is behind Herman?
Does Herman make the lending decision?
Does Herman perform credit scoring?
Where is the data hosted?
Is customer data used to train models?
How much does Herman cost?
Does Herman offer an API?
Do we need to replace our loan origination system?
What compliance evidence does Oper provide?
What is the minimum age to use the service?
Should you pick Herman?
Herman is a deliberately narrow product, and that is its strength. Oper Credits does not claim to automate lending; it automates the grinding stretch between intake and decision, and says plainly that Herman neither scores borrowers nor decides anything. For a regulated European lender, the combination that matters is traceability plus non-disruption: every recommendation tied to its evidence and its policy rule, every step logged, and the whole thing bolted onto an existing origination system rather than replacing it. The supporting apparatus is real, with an ISO 27001 certification held by the vendor itself, European data residency, an explicit no-training commitment, and public API documentation.
The reservations are mostly about opacity. Nothing about the commercial arrangement is public: no price, no plan, no product contract, only a demo request. The published Terms of Use govern the website alone, so the terms that would actually bind a customer, including reversibility and liability, are negotiated out of sight. Two smaller gaps deserve a question during procurement. The subprocessor list is announced as available on the site, yet the download link on the data processors page leads nowhere. And the privacy notice asserts that all core infrastructure sits within the EEA while the product pages announce Frankfurt and Switzerland, which is not an EEA country. Neither is damning, but both are the kind of detail a bank's security review will raise.
The headline performance figures, from 120 minutes down to 3, above 95% extraction accuracy and up to 90% of tasks handled without intervention, are vendor claims without published methodology. The sensible response is the one Oper itself proposes: take the free backtest on your own documents and see what the numbers look like on your files rather than theirs.
- Choosing a selection results in a full page refresh.
- Opens in a new window.