Knowledge Assistant for Contact Center Teams

Customer support quality depends on how quickly and consistently operators can find the right information. At NOS, that was becoming harder as products, processes, and systems multiplied.

Slow, fragmented knowledge search

Operators had to navigate multiple tools and knowledge bases to answer even simple questions. Slow search and unclear navigation translated into long holds, longer calls, and higher effort per interaction.

Scattered and outdated content

More than 15 000 pages of documentation were spread across disconnected systems. Some content was outdated, duplicated, or contradictory, making it difficult to know which source to trust and increasing the risk of inconsistent answers.

Inconsistent service across teams

‍Because each operator relied on their own shortcuts and favourite sources, customers could receive different answers to the same question, making service quality hard to standardise.

UVA was designed as an internal knowledge assistant that sits on top of NOS’s existing documentation and tools.

Instead of manually searching across multiple systems, operators ask UVA their question in natural language. The assistant searches across the internal knowledge base, retrieves the most relevant content, and returns a concise answer with links to the underlying documentation. This turns a multi‑step search process into a single interaction, while keeping the source material visible and auditable.

During implementation, it became clear that content quality was as important as the assistant itself. Some documentation was outdated, duplicated, or inaccurate, which led to incorrect answers in early tests. The solution addressed this by combining retrieval with self‑validation: UVA only surfaces answers from verified sources, and operators can rate responses and flag issues directly in the interface. When an answer is wrong or incomplete, the underlying content is corrected at the source, improving future responses.

Change management was treated as part of the project, not an afterthought. The contact center team was involved throughout the design and rollout, providing feedback on answer formats, tone, and integration into their daily workflow. The assistant was positioned as a tool to make their work easier which helped adoption and ensured that the system evolved in line with real operator needs.

UVA now supports a significant share of NOS’s customer support interactions.

Operators use it to reach the right information faster, reducing handling time by around 10% on the calls where it is used and improving consistency across teams. Responses achieve about 87% accuracy, based on operator evaluations, with incorrect answers flagged and corrected to improve the system over time.

The assistant supports roughly 20% of calls, and the company estimates annual savings of around 120 000€, with upside as usage expands. Operators report spending less time hunting for information and more time actually helping customers, while NOS gains a scalable way to keep support quality high as products and processes evolve.

The roadmap is clear: continue increasing UVA’s coverage across customer support and extend the same pattern to other areas of the company where fast, consistent access to internal knowledge can unlock similar gains in efficiency and service quality.