AI agents

Knowledge that answers questions.

Planning firms and administrations sit on decades of knowledge: in folders, PDFs, old proposals and case files. Our AI agents make this knowledge accessible. In natural language, with source references, and without digging through archives.

Two agents, one principle

Both work on your own document repository, not on the internet.

Archive Agent

In the lab
Make planning archives searchable

An AI agent that makes planning and administrative archives searchable. Finds decisions, reports, plans and case files in natural language, with source references, instead of digging through folders and PDFs.

  • Semantic search across document archives
  • Finds decisions, reports, plans and case files
  • Queries in natural language
  • Answers with source reference and location
  • For municipalities, cantons and planning firms

Proposal Agent

In the lab
Draft proposals faster

An AI agent for proposals and offers. Searches previous proposals, fee rates and text modules in the company archive and supports the drafting of new offers, consistently and in minutes instead of hours.

  • Semantic search across previous proposals
  • Finds fee rates and reference projects
  • Text modules from proven proposals
  • Draft new proposals at the push of a button
  • For planning firms and consultancies

The Archive Agent in the dashboard

In practice

A question in natural language, an answer with a source reference, drawn from thousands of documents. This is what the Archive Agent (internally «Archiv+») looks like in the SPEKTRUM dashboard.

dashboard.spekt.ch · Archiv+
Archive Agent in the SPEKTRUM dashboard: AI-assisted search with source references across thousands of documents

Archiv+ in the dashboard: searchable company knowledge with a traceable source location.

Retrieval instead of guessing

How it works

The agents are based on RAG (Retrieval Augmented Generation): the AI first finds the relevant passages in your own documents and only then formulates the answer. No expensive training of a custom model, no invented content. Every statement can be traced to a source.

1

Organise the data

First clarify data storage and permissions. Only then does the search become reliable and secure.

2

Connect selectively

Not everything at once: we deliberately decide which knowledge the agent may search.

3

Stay independent

Solutions that secure data portability. Without lock-in and without dependence on US providers.

Is it worth it for your firm?

Further reading

Before connecting all company knowledge, it pays to take a sober look at data hygiene and dependencies. We have summarised the main reservations and the right order of steps in a blog post (in German).

Blog · SPEKTRUM
Access to your own database and archive: connecting AI to company knowledge?
Andreas Rupf · spekt.ch/blog

Interested in an agent for your firm?

The first step is a conversation about data storage and permissions. Or write directly to info@spekt.ch.

Request AI consulting