The internet juicer you could beat with your hands

Juicero sold a countertop press, connected to the internet, that squeezed juice from sealed packs of chopped fruit and vegetables delivered to your home. The machine pressed with a reported 8,000 pounds of force and scanned a code on each pack to reject old ones. The company raised about $120 million from investors including Google's venture arm.

The press launched in spring 2016 at $699, with packs costing about $5 to $7 each. In January 2017, the company cut the price to $399.

On 19 April 2017, Bloomberg reported that its reporters had squeezed Juicero packs with their bare hands and got nearly as much juice as the machine produced. The story spread widely.

On 1 September 2017, Juicero stopped selling its press and packs and offered refunds. It said it could not build the national fresh-food supply chain it needed on its own, and would look for a buyer.

Juicero built impressive engineering around a problem that a simpler method already solved. The internet connection and powerful press made the product expensive, but the core job, getting juice out of a pack, did not need them.

Many AI projects fall into the same trap. Teams add a chatbot, a model or an agent because it sounds modern, when a checklist, a search box, a spreadsheet formula or a better prompt would do the job. AI adds cost, upkeep and new ways to fail, so it has to earn its place.

The idea behind this caseLesson: Fine-tuning vs prompting →

1. 1Write down the problem in one sentence before choosing any technology.
2. 2Try the simplest possible fix first, and only add AI if it clearly beats it.
3. 3Compare the full cost of the AI version, including running and maintenance, with the simple version.
4. 4Test with real users whether the AI part changes what they get, not just how it looks.
5. 5Drop features that only exist to make the product sound advanced.