Expert knowledge, in the hands of technicians on site
A manufacturer of precision weighing indicators deployed across agriculture, mining and logistics, serving Australia and South East Asia through a growing distributor network.
The problem
Critical technical knowledge was trapped in PDFs and in the heads of senior engineers. As the field service function and the distributor network grew, that slowed resolution, drove repeat site visits and put steady pressure on a small group of experienced people.
- Knowledge locked in PDFs. No searchable, real time access for technicians standing in front of the machine.
- High escalation volume. Junior staff and distributors went to senior engineers for routine questions.
- Costly repeat visits. Without the answer on the first visit, technicians came back a second time at full cost.
- Uneven distributor quality. Service quality varied widely between regions.
What we did
We built a field assistant on the client's own manuals and service history, designed for someone holding a phone in a noisy environment rather than sitting at a desk.
- Ingested the knowledge base. Manuals, fault codes and service bulletins indexed into a retrieval augmented knowledge base.
- A mobile interface. Technicians ask plain language questions on site and get source-cited answers in seconds.
- Smart escalation. Low confidence queries route to the right engineer with the full context already loaded.
- Distributor rollout. Partners get the same depth of knowledge as direct staff.
Phase one is delivered. The distributor rollout is in progress, with predictive service and customer self-service planned after it.
Where it landed
- 73 percent faster fault resolution
- 91 percent first-call resolution rate
- Eight weeks from kickoff to first deployment
- New technicians productive in weeks rather than months
Our view
The knowledge already existed. It was written down, accurate and complete. The only thing wrong with it was that nobody could get to it while standing next to a broken machine. That is a usability problem before it is an AI problem.
Knowledge stuck with your most senior people?
If the same engineer answers the same question every week, that is the use case.