I started where customer inquiries come in.
I worked in frontline contact center operations, then built and ran chatbot and voicebot channels. Today I design AI systems that support agents and help analyze conversations between customers and companies.
Seeing the same contact center from different roles changed what I thought my work needed to solve.
Are we handling the inquiry or solving the problem?
In customer support, it matters that customers get an accurate answer quickly. When I built support channels, the goal was to help customers do what they came to do without waiting for an agent.
Both mattered.
Over time, though, I began to see a difference between handling an inquiry well and fixing what made the customer reach out in the first place.
That changed how I looked at inquiries.
An inquiry was something we needed to handle. It was also evidence of a customer problem the product or service had yet to solve.
A question began to matter more to me than which channel to build or which technology to use.
What customer problem are we trying to solve with this technology?
What if we could find more problems?
Working on AI products showed me another possibility.
It used to be difficult to read large volumes of support records and distinguish the situation a customer was in, what they asked for, and the context behind the problem. Large language models can help us find and organize clues across more conversations than we could previously analyze.
Analyzing the data alone cannot establish a cause. It does broaden what we can observe.
That raised a new question.
If we get better at finding customer problems, will we get better at solving them?
In my experience, that did not always follow.
If the problem came from a product, a policy, or a workflow, the team responsible for it had to act. Often, the team that found the problem and the team with the authority to fix it were different.
The customer faced one problem. Inside the company, it crossed teams with different goals, responsibilities, and priorities.
Better analysis could help, but someone still had to decide who should own the problem, why it needed to be solved now, and how to allocate limited resources.
An improvement that benefited the company as a whole could mean extra costs for one team.
How, then, does solving a customer problem create value for the company in practice?
What have other companies done?
At first, I thought this might be particular to the companies or contact center market I knew.
So I started looking at companies in other countries.
I found companies where the contact center and other teams handled problems within their own areas. Others wanted to use customer problems to improve products and services but struggled to put a process in place. I also found cases where teams managed those problems together and made changes.
I saw these as different ways of working. The examples did not establish a progression from one approach to another.
Why did some practices take hold while others never became part of how the company worked?
The more I looked, the clearer it became that this question predated large language models.
Companies had long been collecting complaints, looking for common problems, and trying to improve their products and services.
What did they try, and what were the results? What has technology made possible since then, and what still gets in the way?
What I’m studying now
I look at this problem from three angles.
Technology. What was hard to discover in customer conversations in the past, and what can we see with today’s tools?
Execution. How can teams interpret and share what they find, turn it into work someone can take on, and check whether the problem was resolved?
Organization. Even with a process in place, why might a company fail to act? How should it weigh customer value and business outcomes alongside each team’s responsibilities and resources?
These questions are still open for me.
I’m starting with papers, reporting from the time, and company cases. I want to document what was tried, what results the evidence supports, and what remains unknown.
I hope to use that evidence to develop hypotheses and an operating model, then try it in a real company and see how well it works.
This site is where I document that work.
I want to understand how finding a customer problem can lead to a real improvement and create new value.