For the last few years, technology companies have been asking the same question:
Where can we add AI?
It's an understandable question. AI can classify images, recognize objects, generate descriptions, identify patterns and make enormous collections of data searchable in ways that weren't previously practical.
But when the data is deeply personal, I think product teams should start asking a second question:
Where should we deliberately not use AI?
We build Yogile, a private photo storage and sharing service. One change I've noticed is that people increasingly ask us about AI.
They used to be much more interested in storage capacity. How many photos can I keep? How much space do I get?
Yogile offers unlimited photo storage, but today we're increasingly hearing questions like: Are my photos private? Does AI scan them? Can my photos be used to train AI?
That's an interesting change in what consumers consider part of a storage product.
A photo library isn't just a collection of files
Consider what's inside the average family's photo library.
Children growing up. The inside of their home. Family members and friends. Holidays. Weddings. Schools. Locations. Perhaps decades of someone's personal history.
AI makes it technically possible to derive considerably more information from those files than the owner explicitly provided.
That changes the relationship between the customer and the storage provider.
The customer may have uploaded a photograph because they wanted it stored safely. They didn't necessarily upload it because they wanted the provider to understand who's in it, where it was taken or what everyone is doing.
At Yogile, we've therefore made a deliberate choice not to use AI to scan customers' personal photo libraries.
We don't need to understand someone's photographs to provide what they came to us for: safely storing their original files and letting them share those photos privately.
I think this distinction is going to matter far beyond photo storage.
Technical capability isn't automatically product value
AI has created an unusual product-development environment.
Normally, we build technology because users have a problem and the technology helps solve it. With AI, there's sometimes a temptation to reverse that process: we have an extraordinary capability, so where can we apply it?
That's how products accumulate features that are technologically impressive but don't necessarily make customers' lives better.
For sensitive data, the cost can be greater than unnecessary complexity. Every additional analysis can create new derived information about the user.
A useful product question is therefore:
If we removed this AI feature, what problem would the customer be unable to solve?
If there's no good answer, perhaps the analysis wasn't necessary.
Data minimization can improve the experience
We've applied a similar philosophy to sharing.
If someone creates a private album and invites 50 wedding guests to contribute their pictures, we could require all 50 people to create Yogile accounts.
From a growth perspective, that looks attractive: 50 registrations, 50 email addresses and 50 potential future customers.
But most guests didn't arrive because they wanted another cloud-storage account. They wanted to see the album or add their photos.
So guests can participate without creating an account.
That means collecting less information, but it also means fewer steps for the user.
This is something I think the technology industry gets wrong about privacy surprisingly often. We frame privacy as friction: more consent screens, more settings, more controls.
Sometimes the most privacy-conscious design is simply not asking for the information in the first place.
Trust will increasingly include what technology doesn't do
Storage capacity, reliability and preserving original files still matter enormously. People want confidence that the photos they upload today will still be there years from now.
But AI is adding another dimension to trust.
Consumers increasingly want to understand not only whether a company can keep their data safe, but what the company can learn from that data once it has it.
That's why I think “no AI” will become a legitimate feature in certain categories.
Not because AI is bad. We use AI elsewhere where it solves real problems.
But good technology isn't about maximizing the amount of technology inside a product.
It's about using technology where it creates value, and having the discipline to leave it out where it doesn't.
For products handling our most personal data, restraint may become one of the most important technology features we can offer.