4 in 10 scrap more AI than they keep

Four in ten companies have now scrapped most of their AI projects before they ever reached real use.

That's from S&P Global's 2025 survey of 1,006 people at companies across North America and Europe. A year earlier, the figure was 17%.

The top reasons they gave were cost, data privacy and security risks.

I see the small-business version of this all the time. Somebody buys an AI tool because everyone's talking about it, then goes looking for something to use it on.

A month later there are three subscriptions and the same jobs are still landing on the owner's desk.

The tool got chosen before the target, and that's where people go wrong.

So I'd flip the order and choose the problem first.

Write down the one job in your business that eats the most time every week. It might be answering customer enquiries or writing up the weekly report.

Then write down how long that job takes now, and what finished looks like. Make it specific enough to check, like every enquiry answered within the hour.

Only then go looking for the AI tool that fits that job. When you shortlist, ask what each one costs at your size and where your customer data goes, because those are exactly the reasons the bigger companies gave for pulling the plug.

You'll buy less, and what you buy will get used, because it was picked for a job you'd already measured.

That's how you end up keeping more AI projects than you scrap.