B2B technology20253–5 months
Client: A B2B software company
One of the project’s screens, in a desktop frame. It is shown again with its caption under “On screen”.
At a glance
Assistant, evaluation set and checks
Product experience and hand-off to sales
AI infrastructure and monitoring
From brief to result
The brief in full, our approach, what we built and what it set out to change.
Sales engineers answered the same product questions repeatedly, and the answers went out of date: an old limit here, a retired feature there. Prospects wanted answers on the spot, but a wrong answer on security or pricing was worse than none.
The brief: an assistant that answers from approved sources, says when it does not know and hands over to a person when needed.
We built the evaluation set before the assistant: questions from sales calls, each with its approved answer and required source. Every model and prompt change has to pass it.
The automatic checks were written as policy first: what the assistant may answer, what it must refuse and when it routes to sales. Prompt-injection tests (OWASP LLM01) run with every release.
Questions collected from call notes, tagged by topic and by risk.
The evaluation set and its rubric: grounded, correct and cited, or a clean refusal.
Search over approved documentation, the assistant interface and the hand-off to sales.
Released only when the evaluation set passes, with injection tests on every change.
An assistant that answers from approved product documentation, passes the evaluation set before each release and hands over to a person when confidence is low.
The assistant is built to answer only from approved documentation, show the page it used and pass pricing and security questions to a person. Every conversation is logged for review.
Accuracy is measured on the evaluation set, and the figures are the client’s to publish.
Picture story
These pictures are illustrative: licensed stock photography and code-built mock-ups that show the setting, not the client’s own work. Credits are at the foot of the page.
On screen
Illustrative, code-built screens of the kind the programme produced, not the client’s own. Open any of them full screen.
Field notes
Outcomes
What each measure is compared with, when it is read and who reports it, with results once the client approves them.
Answer accuracy on the evaluation set
Escalation rate to sales
Demo-to-opportunity rate
Figures pending client approval
We publish results only with the client’s written approval. Until then, the programme is judged on the plan beside this, which we can walk you through under NDA.
Credits
People are named only with their agreement.
Licensed stock photography and code-built mock-ups chosen to show the setting, not the client’s own work.
Keep going
Next case · 04 · Retail & commerce
Get cited in AI answers as well as ranked in search results.
Read the next case
You will speak to a lead who would run the work, and get a straight answer on fit.
01About 2 minutes
You get A reply from a named lead
02About 8 minutesRecommended
You get Options and a first scope after one call
03About 15 minutes
You get Receipt confirmed and a named bid lead
Every engagement starts with a written scope and a quote agreed before work begins. How each package is priced
