Some problems need AI. Some just need scissors.
It’s tempting to slap “AI-powered” on every new feature or product. Feels smart. Feels modern. But as generative AI becomes more accessible, the real challenge isn’t in building with it. It’s knowing when not to.
Here’s the question behind every smart product decision today: Should this use machine learning? The honest answer: Not always.
What Happened
Generative AI has changed the game. You no longer need mountains of training data to add intelligence to your product. Large language models can fill in the gaps.
But just because you can use ML doesn’t mean you should. Some tasks are still better handled with a rules-based system. A simple decision tree. A list of preferences. A few lines of logic.
Knowing the difference is what separates thoughtful builders from trend chasers.
Why It Matters
In business, efficiency matters more than hype. It’s not about using the flashiest tool. It’s about solving the right problem with the right approach.
Some customer needs are straightforward. Others require flexibility and scale. If a user always inputs the same thing and expects the same result, you don’t need AI. If the inputs are unpredictable and the outputs must adapt, that’s where ML earns its keep.
It’s not just about power. It’s about fit. About using your team’s time, talent, and budget wisely.
How It Impacts You
If you lead a product team or make tech decisions, you’ve heard this question in a meeting. “Should we add AI to this?” Maybe someone thinks it will impress investors. Maybe it’s what the competition just launched. Maybe it just sounds like progress.
But the best leaders slow down before saying yes.
They look at inputs. They look at outputs. They look at cost and precision. And they ask, “What actually makes this better for the customer?”
The best tool isn’t always the newest one. It’s the one that works.
3 Things You Can Do Now
- Map the input-output path What does the user give you? What do they expect in return? Is the logic simple or complex?
- Use the rule vs. ML matrix Repetitive tasks with predictable outputs? Stick with rules. Complex variations with changing outputs? That’s where ML helps.
- Audit one feature this week Pick a part of your product. Would it improve with AI, or just cost more? Simplify where you can.
One more thing:
Good products solve real problems. Great ones know when to stay simple.
What’s one part of your customer journey that might be better without AI?
~Harrison