Layout-aware extraction
Umbiko preserves the structure of tables, headings, and paragraphs, and understands which number belongs in which column.
How it works
Turn a stack of PDFs into a comparable table in minutes
Upload all your documents. Umbiko extracts line items, links them to suppliers, highlights risks, and answers questions with direct references to the source.
01
Upload documents
Tender, quotations, specifications, terms. PDF, Word, Excel. Nothing leaves the EU.
02
Extract and structure
Articles, quantities, unit prices, STABU codes. Umbiko reads the small print and normalises it.
03
Compare and question
Compare suppliers side-by-side. Ask questions in plain language — the answer cites the page.
04
Sign off with evidence
Every risk, every exception, every number — traceable back to the source. Export your report.
Under the hood
Built for documents, not chat.
Umbiko combines document intelligence, vector search, and domain-specific prompts to extract and compare supplier quotes reliably.
Semantic search met grounding
Every paragraph is embedded into a vector index. Search and chat answers cite the exact source, so every claim can be traced back to its origin.
Domain-specific agents
Dedicated AI agents handle line item extraction, price normalisation, risk detection, and multilingual support, each with their own prompt and validation.
What we don't do
Boundaries we won't cross.
What AI can do is not always what AI should do. That is why we make three explicit choices we are not willing to compromise on.
No training on customer data
Your documents stay within your tenant and are never used to train or improve models. We use Azure OpenAI with a no-training clause, not the public ChatGPT API.
No US datacenters
All data, documents, RFQs, and supporting files are stored and processed within Azure West Europe. Your content does not pass through external pipelines or CDN caching outside this environment.
No ungrounded answers
Umbiko only answers based on the available source documents. If the information is not in the source, it will not be filled in or invented.
Frequently asked questions
How Umbiko works: frequently asked questions
Frequently asked questions from conversations with customers. Missing something? Schedule a demo.
Do Umbiko’s AI models run entirely within Azure, or are external AI models also called?
Yes. Umbiko runs on Microsoft Azure, and the AI layer is built using Azure OpenAI Service and native Azure AI services, such as Azure AI Search, OCR, entity and PII detection, and translation.
For the processing of customer documents, we do not call the public OpenAI API, Gemini, Claude, DeepSeek, or other external model providers in the background.
Azure OpenAI means that the AI models are offered as an Azure service and are hosted within the Microsoft Azure environment. Microsoft states that prompts, responses, embeddings, and customer data are not made available to OpenAI or other model providers, and are not used to train or improve foundation models without the customer’s permission or instruction. The models themselves are stateless: prompts and outputs are not stored in the model.
For European customers, Umbiko is configured within European Azure regions. Data is encrypted in transit and at rest, and is processed within the configured Azure environment. Microsoft’s EU Data Boundary describes how Azure services can be configured within the EU/EFTA for the storage and processing of customer data.
Where is our data stored?
For European customers, Umbiko runs within Europe. User data is stored exclusively in Europe and processed within the configured Azure environment. This keeps data within the agreed infrastructure and avoids unnecessary processing outside Europe.
Is data encrypted in Umbiko?
Yes. Data is encrypted both in transit and at rest. Umbiko uses Azure’s built-in encryption for data storage, ensuring that documents, analyses and user data are processed securely within the Azure environment.
Is our data separated from other organisations?
Yes. Umbiko separates user data per user and organisation. Data remains within the tenant and is not shared with other organisations. This is important for teams working with confidential documents, supplier information or internal decision-making.
See Umbiko in action
Book a demo and see how Umbiko works in practice for your team. Prefer to explore it yourself first? Request temporary access for a first impression of the environment.