The Conversation I Want to Be Ready For

A little while ago, I had a conversation with an AI voice agent called Valentina. I had heard about what she could do, but hearing about something and actually experiencing it are two different things. I spoke with her myself, and that conversation stayed with me.

Valentina can talk to a customer naturally, understand what that person needs, access information, take certain actions inside a business, and hand things over to a person when the situation requires it. That basic description sounds simple enough, but what interested me was everything sitting underneath it.

How does an AI know what to do next? How does it get information from a company? How does it book something, update a CRM, check inventory or follow up with somebody? Who decides what it is allowed to do? What happens when it reaches a situation it cannot handle? Those questions are what pushed me into learning AI more deliberately.

One of the first things I wanted to understand was how AI actually gets work done. That eventually became an article called How AI Actually Gets Things Done. The biggest thing I took away from that exercise was surprisingly simple: AI doing useful work inside a business is basically a chain of decisions. Something happens, the system understands what is happening, decides what comes next, uses whatever tools and information it has been given access to, takes an action, and if it reaches a boundary, a person steps in.

Once I understood that, Valentina became much easier for me to explain. Instead of thinking about some mysterious piece of artificial intelligence sitting inside a company, I could picture the actual process. A customer calls, Valentina listens, figures out what the person needs, checks the information she has access to, and then does whatever the situation calls for. Maybe she books an appointment, updates a record, answers a question, or realizes that this particular situation needs a person and transfers the conversation.

That gave me my first real ground-level understanding of what I was looking at. For a while, I figured I would just keep learning from there: start at the bottom, learn the fundamentals, keep adding pieces, and eventually I would understand enough of the technology and enough of the business to explain the whole thing properly.

Then I started thinking about the destination. Suppose somebody called me tomorrow and said, “I heard about Valentina. Come have coffee with me and explain what you’re working on.” What would I actually need to be able to explain?

I would need to tell them what Valentina is, how something like this gets plugged into an existing company, where I would start and why. If they asked how long a pilot might take, I would need to understand that. If they asked what it might cost, I would need to understand how those numbers are built. If they asked what kind of return they could get from using it, I would need to be able to talk about things like time saved, calls handled, leads recovered, appointments booked and work that employees no longer have to do manually. And if the conversation went one step further and they said, “This sounds interesting. How could I get involved?” I would want to understand that conversation too.

That gave me a much clearer way to think about what I’m learning. Instead of moving through AI in a straight line and hoping that eventually I arrive at the business conversation, I can start with the business conversation and work backwards. Take every question somebody might reasonably ask me and then learn enough to answer it properly.

If somebody asks, “How would this work inside my company?” then I need to understand integrations, permissions, workflows, systems of record and human handoffs. If they ask, “How would you start?” then I need to understand pilots. If they ask, “What would you actually install first?” then I need to understand where the smallest useful piece of the system sits.

For a dentist office, maybe that piece is answering calls, handling common questions, booking appointments, sending reminders and passing unusual situations to a person. I’ve started thinking about that almost like Lego. The dentist already has the castle. I’m interested in the little mechanism that makes the drawbridge work, one useful piece that can be added to something that already exists. Later, there may be projects where somebody wants you to rebuild an entire section of the castle and make sure it still connects properly to everything behind it. That is a different kind of job, and right now I want to understand the pieces.

Then there is the money side. What does one of those pieces cost to implement? How much work does it replace or improve? How much additional business can it help capture? How do you measure whether the pilot actually worked? And eventually, what does the commercial model around all of this look like? Those are the questions I want to work through next.

The goal I have in my head is fairly ordinary. I want to be able to sit down with somebody over coffee or lunch and talk about Valentina for five minutes, ten minutes, half an hour, whatever the conversation becomes. I want them to be able to interrupt me, ask something unexpected, take the conversation sideways, ask me how it works, what it costs, what happens when it fails, where I would start, or why I think there is a business here, and I want to understand the subject well enough that I can simply talk about it.

So that is where I am now. I started by trying to understand how AI gets things done. Now I’m working backwards from the conversation I want to be ready to have.