AI Agent
A review of a tender writing platform's AI that cut its running costs significantly and made its answers more accurate and consistent.
Key features
A tender writing platform relied on AI to produce its answers. It worked, but it cost too much to run and the quality of its answers varied from one to the next.
The causes were technical. The effects were business ones: bills that were too high, and answers that were hard to rely on.
First, we met the team to understand their setup, their problems and their business constraints. Then we went through their code to see how the AI actually found and used their documents.
From there, we gave them a list of changes, ranked by impact and effort:
Running costs dropped significantly, and answers became more accurate and consistent. The team can now measure the effect of every change they make, so they can keep improving it after the engagement.
Retrieval-augmented generation (RAG), large language models, vector databases, document processing and evaluation frameworks.
More work
An AI rostering assistant that drafts Onsen Hot Pools' weekly staff rosters, saving at least 5 hours a week for each of its three teams.
An AI running coach built on an ultrarunner's knowledge. It served over 100 runners in its first two weeks and is now a profitable, growing business.
An AI quoting tool that lets businesses with large or complex product ranges give customers an instant, itemised quote on their own website.
About us
A team of product engineers based in Queenstown, NZ. We work with you to understand the problem first, then build the right thing — not just the possible thing.
Book a consultation