DocAI

AI Spaces is an AI consultancy in St. Gallen, Switzerland. Their clients are mid-sized and large enterprises, including a regional railway operator and a scaffolding contractor, and they kept running into the same wall: large document sets that had to be read by hand.

A construction firm assessing a tender, a real-estate team running an ESG review, an energy business checking an EPC contract for a solar project. Hundreds of pages, read manually, with results that varied by who did the reading and no record of how a decision was reached.

They did not need another slide deck about AI. They needed a product their clients could run against real documents and trust.

We built DocAI, a web application with one simple loop.

• The user defines an analysis task once, along with the questions that matter for it.
• They upload the documents, structured and unstructured.
• The system extracts what is relevant to that task and answers each question.
• Every answer carries a source reference back into the original document, so a reviewer can check it in seconds instead of trusting it blind.
• Results export to PDF, to email, or straight into Microsoft Teams, which is where the decisions actually get made.

The hard part was not getting a model to produce an answer. It was making answers over a company's private documents specific, citable and repeatable, so the same question on the same tender gives the same result every time.

DocAI is the client's flagship product and is sold commercially today. AI Spaces publishes the outcomes on their own product page: 10x faster analysis of complex tender documents, 85% fewer human errors, and 99% higher user adoption. Those are the client's figures, not ours.

It is now sold for tender analysis for construction firms, ESG analysis for real-estate projects, building-regulations analysis, and EPC contract analysis for photovoltaic projects.

This is the shape of work we do best: an AI feature over a client's own data, inside a real product, built to survive real use.