Smart Search: finding internal documents in under a second
24.09.2026 – reference case Zuger Kantonsspital
In hospital operations, every minute counts – including when searching for documents. On the Zuger Kantonsspital's intranet, as in many organisations, this ran on a classic keyword search. The problem: anyone who didn't know the exact title or spelling often got incomplete results, or none at all. Ergon developed a search that makes thousands of internal documents findable in under a second.
When keywords aren't enough
The starting point at the Zuger Kantonsspital (ZGKS) was typical of many internal knowledge platforms: the documents exist, but finding them is hard work. The keyword search worked well when employees already knew the correct term or document title. In every other case, it took real effort: trying different search terms, varying the spelling, working through several results lists – and still not being sure the right document had turned up. On top of that came the wait: each query took several seconds, which added up to a real loss of productivity over the course of a day.In a hospital environment especially, every minute counts. Information needs not just to exist, but to be reliably findable. The ZGKS was therefore looking for a solution that would return results even when the wording wasn't exact.
From keyword search to AI-powered search
The new search improves on several fronts, all with one goal: showing the most relevant results as close to the top as possible.Meaning, not just wording
At its core is the combination of classic keyword search and modern vector search. Vector search adds semantic understanding to the keyword search: it finds documents even when the wording differs, spellings vary, or employees are searching for a related concept rather than an exact term.
Better ranking through reranking
To bring the most relevant results further up the list, a reranking model is also used. It orders results so that the documents most relevant to the specific query appear right at the top. This is most noticeable when a whole sentence is typed into the search field: rather than counting individual words, the model takes the meaning of the query as a whole into account.
Narrowing down with filters
Filters – by document type or department, for example – help employees quickly narrow down what they need. A few clicks reduce the results list to what's relevant, without having to rephrase the search.
Greater visibility for important documents
In addition, a PageRank algorithm gives priority to documents that other documents frequently reference. Such references indicate how central a piece of content is within the organisation's internal knowledge. For policies, standards and key process descriptions, this means the important content is more prominent.
Seamlessly built into the intranet
A key success factor was designing the search around everyday working life. It's available exactly where it's needed: in the search bar. Employees enter their query, get a clear list of results, and jump straight from a result back into the relevant ZGKS system. This cuts down on context-switching, avoids duplicate work and makes the search a natural part of the daily routine.The solution is also cleanly embedded on the organisational and technical side: it integrates with the existing SSO, so no additional logins or separate user administration are needed.
Always up to date, thanks to continuous indexing
Internal documents change constantly: policies get updated, processes adjusted, templates replaced. In the past, employees sometimes saved documents locally or printed them out, simply because they were otherwise hard to find again later. With the new search, they can call up what they need directly on the intranet. Because new and updated documents are indexed continuously, changes appear in the search results within a short time. For policies in particular, it matters that the current version can be found quickly.The result: faster, more accurate, a better user experience
Moving from the SQL-based keyword search to the AI-powered Smart Search brought a noticeable leap in quality:
- Five time faster search.
- Higher-quality results: relevant documents appear further up more reliably, even with varying terms or full sentences.
- Fewer searches: employees less often need to rephrase their query or guess what a document might be called.
From feasibility study to production-ready solution
Rather than building at scale straight away, Ergon first established, in a structured way, whether and where the approach would deliver real benefit. A feasibility study showed where the existing search fell short in everyday use, and what vector search, reranking, filters and PageRank could realistically improve. The production solution followed from there: performant, stable, integrated into the existing system landscape, and set up so that content stays continuously up to date.The result is ZGKS Smart Search – a technologically modern search that, above all, does one thing well: it helps employees find relevant information faster, so they can get back to the work that matters.