AI document search: how the Integrierte Psychiatrie Winterthur (ipw) makes around 1,800 documents searchable with AI agents
ipw Winterthur · Project overview ·
The project at a glance
The Integrierte Psychiatrie Winterthur – Zürcher Unterland (Integrated Psychiatry Winterthur, ipw) supports people from childhood to old age with mental health difficulties and illnesses. After the AI-powered conversation documentation, document search is a further project by Evoya AI with ipw.
For its process management project, ipw has integrated around 1,800 documents into its own knowledge base. Staff ask questions in natural language in the Evoya AI chat and receive targeted, clear answers based on these documents. Several AI agents cover different subject areas and search logics.
- Client
- Integrierte Psychiatrie Winterthur – Zürcher Unterland (ipw)
- Sector
- Healthcare, psychiatry
- Use case
- Document search in the process management project
- Solution
- Evoya AI chat with AI search and AI agents
- Knowledge base
- Around 1,800 documents
- Data storage
- ISO-certified Swiss data centres
Starting point: large document collections, time-consuming research
A process management project brings together a great many documents. With around 1,800 documents, research becomes a challenge: anyone with a specific question has to know which document contains the answer and then go through it by hand.
ipw wanted to make this research much simpler and faster, with a solution that also meets high requirements for data protection and confidentiality.
The solution: AI search with specialised AI agents
1. Own knowledge base: around 1,800 documents are integrated into an organisation-specific knowledge base. Evoya Search makes the content searchable for the AI.
2. Questions in natural language: in Evoya Chat, staff ask questions the way they would ask a colleague and receive targeted, relevant answers from the stored documents.
3. Several AI agents: several AI agents with different prompts have been set up for the workspace. Each one covers a specific subject area or search logic.
4. One central AI chat: besides search, the AI chat brings further functions together in one environment: texts, images, documents, transcriptions and research.
AI search or keyword search: what is the difference?
A classic keyword search finds files in which a term appears. If you do not know the right terms, you find nothing, and if you get too many hits, you have to go through them one by one.
An AI search understands the question. It finds the relevant passages even when different terms are used, summarises the relevant content and delivers a direct answer. Specialised AI agents also ensure that a question is answered in the right subject area and with the appropriate search logic.
Data protection and model choice: Swiss data centres, the right models
Data is stored in ISO-certified Swiss data centres, and the AI chat is aligned with the requirements of the Swiss FADP and the GDPR. The data is never used for AI training.
For sensitive data, Swiss AI models are available that run on sovereign Swiss infrastructure. In addition, leading international models from OpenAI, Anthropic, Google and Perplexity are available. This lets ipw choose the right model for each use case, performance requirement and level of protection needed.
Results: answers instead of hit lists
- Faster research: questions about around 1,800 documents can be answered in seconds.
- Targeted answers: the AI delivers clear answers instead of long hit lists.
- Clear subject areas: specialised AI agents cover different kinds of questions.
- A foundation to build on: the AI chat lays the groundwork for AI agents to take on more extensive tasks and process steps in the future.
Read more: Evoya Search · Evoya Agents · AI in healthcare · AI transcription at ipw
Frequently asked questions about the project
Around 1,800 documents are integrated into ipw's own knowledge base. Staff ask questions in natural language in the Evoya AI chat, and specialised AI agents provide answers based on these documents.
A full-text search finds files that contain a specific term. An AI search understands the question, finds relevant content even when different wording is used and summarises it into a direct answer.
Each agent receives its own instructions (prompts) for a subject area or search logic. This way, questions are answered more precisely than a single, general-purpose agent could manage.
Data is stored in ISO-certified Swiss data centres. The AI chat is aligned with the Swiss FADP and the GDPR, and the data is never used for AI training.
For sensitive data, Swiss AI models on sovereign Swiss infrastructure are available. In addition, international models from OpenAI, Anthropic, Google and Perplexity can be used, depending on the use case and the level of protection needed.
Make your documents searchable with AI
We show you how your staff get answers from your own documents in seconds, securely and in compliance with data protection.