I've been at a few startup events in the last month and I've noticed a pattern. Nearly every pitch mentions artificial intelligence. Most of them, when you ask what that means technically, describe a feature that uses an API someone else built, or a simple rule-based system with a machine learning label on it, or sometimes nothing more than a marketing decision.
This is fine as a marketing choice. It becomes a problem when founders start believing their own framing.
What AI-washing looks like
The tell is vagueness under pressure. "We use AI to personalize the experience" is not a technical statement. Ask what model, what data, what metric you're optimizing for, and how you measure whether it's working. If the answers are unclear, the AI is probably not doing what the pitch implies.
I'm not being cynical about this. I've been on that side of the conversation. When you're building under pressure it's easy to reach for the label before the substance catches up.
What real ML in a product actually requires
Data that reflects the problem you're trying to solve. A metric that actually measures what you care about, not a proxy that's easy to improve but doesn't move the thing you need to move. A feedback loop that tells you when the model stops working. And someone who understands enough to debug it when something goes wrong.
None of that is exotic. It's just work that doesn't fit into a pitch slide.
Why it matters
When AI becomes a label rather than a description, two things happen. Users get oversold on capabilities that don't exist, which erodes trust when reality falls short. And founders optimize for the perception of intelligence rather than the substance of it, which produces worse products.
The companies that will be interesting in a few years are the ones treating machine learning as an engineering discipline, not a branding choice.
I'm watching a few of them. They're quieter than the rest.
With gusto, Fatih.