Guessing — or actually observing?

How can I make Buildmate reach more people? Since I started this initiative, the challenge has been that I have to guess how to make Buildmate useful. The low level of usage has always been an obstacle to be overcome before moving further.

Buildmate’s core functionality was built to help consultants create reports based on natural language. Navigating Buildmate and understanding how it interacts with Hypergene is still a wall to climb.

Report building wasn’t necessarily a bad idea, but I could tell it would take some time for colleagues to actually use it. Also, building reports might have been a bit overly complicated for many users.

Buildmate takes away much of the necessity of learning the report editor by replacing it with natural language. However, it does require the user to learn a new way of working, which is a one-time investment in learning Buildmate in general.

This will only happen if I can prove a visible and direct positive effect for the users.

I needed to rethink what Buildmate can do and who it should help.

We recently hired quite a few customer support specialists. That opened up a new use case. I hoped that new employees would be more open to new ways of working.

I could replicate common patterns in my colleagues’ daily work. Buildmate could then mirror the know-how of a senior customer support specialist. I set out to do just that. New support specialists can now get knowledge and best practices from Buildmate instead of relying on asking others.

A project in Buildmate is paired with a ticket from the customer support system. Everything the user and Buildmate talk about in a project is fully documented.

If the ticket is solved with the help of Buildmate, the documentation serves as a case study for future similar issues. If the ticket isn’t solved there is still a lot to gain. Apart from documentation, an unsolved ticket helps to improve Buildmate the next time the issue appears.

Previously, I had actively tried to get people to use Buildmate by inviting them to workshops. This time I didn’t invite colleagues to a workshop, but asked someone in customer support to see who were interested in participating.

Maybe because of this, the workshop was more active than previous ones. I hoped that was a move in the direction I wanted: to see some demand for Buildmate from colleagues.

Up to this point, I used my own instincts to improve Buildmate. Increased usage was the first step and I saw a jump after the workshop. This cleared the way for a new path. For the first time I could observe how Buildmate was actually used.

During the following days I kept track of Buildmate and the questions asked in order to iteratively improve the correctness of the answers.

The usage itself is a prerequisite for the main goal, which is to let consultants solve problems more easily and help new colleagues be productive faster.

I track a lot of data in Buildmate. Some is easy to measure accurately, and some is harder. The missing metric so far is whether a solution was ultimately provided or not.

The success or failure reflected in such metrics is, at this point, unknown. But it will decide if this is a track to follow or not.