An air-conditioning installation and service company had eight field technicians, one dispatcher, and a veteran manager who had been there from the first day. Every unusual fault and every older model ended in a call to him.
The starting point: one person as the knowledge center
The figures from the mapping were clear:
- The technical manager received 10 to 15 calls a day from technicians. A one-week sample found that nearly 70% of the questions repeated: fault codes, installation specifications, and manufacturer warranty procedures.
- Onboarding a new technician took about three months before they could work independently, mostly through close support from the technical manager.
- When the technical manager went on holiday during Passover, the company postponed every complex job in advance.
The knowledge existed in manufacturer binders, old emails, WhatsApp groups, and one experienced person’s head. It simply was not accessible.
What we built
- Collect and organize the knowledge. Together with the technical manager, we reviewed the existing material: manufacturer instructions, internal procedures, and solutions to common faults. Much of the knowledge had never been written down, so we recorded a series of short conversations with him and turned the transcripts into procedures. Everything was organized in a Google Drive procedure library, with an owner and update date for every document.
- An internal knowledge assistant in WhatsApp. A technician writes a question naturally, from a fault code to “How do I submit a warranty claim to the manufacturer?” The assistant answers only from approved procedures and links to the source document. If the answer is not in the library, it says so explicitly and sends the question to the technical manager instead of inventing an answer.
- A weekly improvement loop. Every question sent to the technical manager because the library had no answer becomes a new procedure at the end of the week. The library therefore grows at the pace of genuine need.
- Clear limits. The assistant does not quote prices to customers, approve work, or override safety decisions. Those questions always go to a person.
The results after four months
- Daily calls to the technical manager fell from more than 15 to fewer than five. Most remaining calls genuinely require his experience.
- New technicians who joined after the project reached independent work in about six weeks, compared with three months before.
- According to the technicians, average resolution time for unusual field faults fell because an answer arrived within minutes instead of waiting for one person’s availability.
- The technical manager, who had been the most skeptical person at the start, became the library owner. In his words: “Instead of answering the same questions all day, I can focus on giving a good answer to the problems that genuinely require my professional knowledge.”
What can we learn from this story?
Company knowledge held in one person’s head is an asset, but also a risk. Making that knowledge accessible does not replace the expert. It frees that person to focus on work that genuinely requires expertise and helps every new technician become independent much faster.
Questions we get asked about this process
What prevents the assistant from inventing answers? The assistant answers only from the company’s approved procedure library and includes the source with every answer. When the library does not contain an answer, it says so and routes the question to a person. We test that boundary before launch and continue checking it afterward.
How much work is required from us to build the knowledge library? This is the main part of the project, and it cannot be skipped. In this case, it took about a month at a pace of two meetings a week with the technical manager. The surrounding technology is much simpler.
Is this suitable for a business with no written documents at all? Yes, and that is common. Most knowledge sits in people’s heads and WhatsApp. Part of our work is to bring it out through recorded conversations and turn it into written procedures. Even without an intelligent assistant, that process alone has considerable value.
The client’s identifying details were removed at their request. The figures were measured with the client before and after the project.
