Assistants & Agents
TellusR lets you set up assistants and agents as part of your organisation’s AI architecture – from simple chat assistants on your own data to advanced solutions with tools, retrieval and cross-system collaboration. Everything is governed in the AI Communications Hub.
With TellusR you can build, customise and operate AI assistants and agents as part of your organisation’s overall AI architecture. In the AI Communications Hub you get a single control point for data, language models, retrieval, access, integrations and workflows.
Build everything from simple chat assistants with access to selected proprietary data to advanced agent solutions that find relevant information, use tools and collaborate across systems. You can assemble specialised assistants with clear tasks and let them interact with a higher-level assistant for a domain or work process.
Get started quickly with ready-made setups, or build tailored solutions with your organisation’s data, instructions, models, integrations and work processes.
Build AI your organisation can trust. With control of data and models, precise retrieval, built-in testing and full traceability you get visibility into how solutions respond – and the tools to test, quality-assure, troubleshoot and improve them continuously.
Uses TellusR AI Search as the retrieval component to find and rank relevant content across selected sources. Boost Graph: dynamic boosting and filtering of content using LLMs. Increases precision in context selection by exploiting knowledge of where chunks sit in the content structure or how they are categorised.
Unique to TellusR
Not everything in a conversation is a question for the language model. Formatting search hits, choosing instructions or checking that the sources exist should happen the same way every time. TellEm is a small, purpose-built scripting language in Dialogue that lets whoever owns the prompt do exactly that – deterministically, with no model call and no deploy.
What can be computed should not be guessed. Scripts run the same way every time.
The logic runs exactly and for free – no tokens, no latency.
TellEm has neither loops nor recursion and runs with no access to the file system, network or shell. A script cannot hang or ask for more.
TellEm rules decide when a tool is offered to the model, format search hits and narrow down which fields it may fill in.
Validation rules check sources and format before the answer goes out. If a rule fails, the model is told and tries again.
Scripts are edited directly in the prompt editor and versioned together with the prompt – no separate deploy.
Less guessing. More control.
DiBK’s internal work tool Arkivslurpen is built on TellusR and processes the several thousand inquiries the directorate receives every year, extracting information from previous answers to make case processing in our response service faster and more accurate. It lets less experienced staff answer quickly and correctly, while we retain full control of data sources.
Significantly faster information retrieval with AI chat than via traditional search or support systems.
Reduces errors and data leaks compared to open AI services – ensures sensitive information never leaves the organisation.
Increased employee productivity: immediate access to up-to-date, quality-assured information – fewer interruptions in the working day.
Modern user experience: the same functionality as familiar AI chat tools, but on your security terms – with no compromise on privacy.
Partners
Invest in insight, security and flexibility – not risk. Send us your most challenging use case!
