Oossa

Portable skill lets simple RAG match human support steps

A lightweight AI skill achieved 99.2% procedural match to human-approved customer support workflows, though it runs slower than the existing system.

NoteBy Published by Oossa: 1 min read

Istok Bako and team showed that a portable skill—just a natural‑language instruction, a retrieval API and some format checks—can replace a complex document‑retrieval stack for customer support. The skill adds a LLM‑generated navigation index over policy documents and can be built in 1‑2 engineer days. In their test it reproduced the correct resolution steps in 99.2% of cases and provided sufficient evidence for 83.3% of them. The approach runs about ten times slower (≈90 s per case) than the company’s production assistant, which takes 8–9 s.

Why it matters

Support teams could use a much simpler setup to automate workflows, saving engineering effort even if response time is longer.

Was this article useful?
Share

Read next

Oossa · Newsletter

The week in AI, explained

Every Monday: the stories worth knowing, in plain language. Free, no spam.