Skip to content

Internal tooling

Reference build

Answers from scattered internal docs, with citations

A retrieval-augmented agent that answers employee questions across scattered internal documentation and cites the source, cutting time spent hunting for answers by 76%.

76%less time searching

At a glance

Duration
13 weeks
Team size
1 person
Engagement
New build
Project type
AI & automation
Industry
B2B SaaS

The situation

The challenge

Internal knowledge is spread across wikis, drives and threads, and it goes stale unevenly. An assistant that answers confidently from an outdated page is worse than no assistant, because it is believed.

What we did

Retrieval over generation, with citations mandatory and freshness treated as a ranking signal.

The calls that mattered

  • Citations are not optional

    Every answer links the documents behind it. Without that, nobody can tell a good answer from a fluent one.

  • Freshness weighted in retrieval

    Recency contributes to ranking so a superseded page does not outrank the document that replaced it.

What changed

reduction in search time
76%reduction in search time
of answers carry a source citation
100%of answers carry a source citation

76% less time spent searching, with every answer traceable to the documents it came from.

Services used

  • Ingestion and chunking pipeline across document sources
  • pgvector retrieval with freshness-weighted ranking
  • Cited answer interface

What we would do differently

Every project has one of these. Publishing it is the point — a case study with no regrets in it is marketing, not evidence.

Retrieval quality set the ceiling on everything. Effort spent on chunking and freshness paid back far more than any prompt work did.

Services used

More work

Ready to start?

Start a project like thisWork