"Are you going to train us a custom AI model?"
Almost never. And if someone leads with that, be careful.
There is a real difference between two roles that get blurred together.
An ML engineer trains models
An ML engineer trains models from data. That work is research-heavy, expensive and slow, and most businesses don't need it.
An AI engineer builds systems around models
An AI engineer, which is what I am, takes the models that already exist, like Claude and GPT, and builds real systems around them:
- retrieval over your own data
- agents that take actions
- automations that connect the tools you already use
For almost every business, the value isn't a new model. It is wiring today's models into your workflow so they do useful work.
The better question
The companies winning with AI right now aren't the ones who trained something. They are the ones who shipped something.
So the better question isn't "can you build us a model?"
What in my business is repetitive, rules-based or language-heavy?
That is what can be automated today, with tools that already exist.