Why do SMEs struggle with AI?

Why does AI work at home but not in the business?

By now the majority of people have at least played with AI. The decision makers I see in medium-sized companies have usually built something amazing with AI in their private life already. A do-it-yourself project, meal prepping, the gym, some even their household finances. But in their business, they struggle with making AI work.

If you do something with AI in your own environment, where it's just you or maybe your family, the risk is low. If something doesn't work out, that's not problematic. In a business, everything has a level of professionality, and things are supposed to work. You have real customer data. You have concerns you don't have in your private life that much: data privacy, data handling in general, IT systems. And you have the people you always have in mind when you want to change something: your peer, the other executive, your employees who have to be on board.

Why isn't a small prototype enough?

You can do a very small prototype. But often the impact just isn't there. Then it's seen as playing around. Especially if you're the CEO, you have other stuff to do. You have a business to run and new customers to acquire. You have to make sure you're profitable, that your employees are happy and that product quality is up to your standards.

You just don't have the time to play around. So you have to touch something that is actually a value driver.

What happens when AI touches a real value driver?

Everything that's a real value driver in your business is usually at some level of professionality. If you want to build something there, you basically compete with how things are done today. And medium-sized companies are often a hundred years old, or family operations where the great-great-grandfather started the company after the war. A lot of things have always been like that.

So the next step is to get someone professional. But if you involve a full software development company that now specializes in AI, it gets big and expensive really fast. There is political capital involved too. As a leader you're closely involved with your employees. They need to trust you, and in a crisis they need to know you're making the right decisions. So you have to take them along, and the change management makes the scope even bigger. That makes the whole task daunting.

Why is it different for enterprises and startups?

A large enterprise can have a dedicated team that takes care of AI while everyone else keeps doing the day-to-day business. At some point the AI team comes out of their little room and says: this is our solution. A startup can move fast and break stuff. The customer base is often not that big, and people who work with startups know that things sometimes break.

A medium-sized company can't do either. One of its biggest assets is its reputation. Being reliable, a long-standing company that can be trusted, the personal connections. If something goes wrong, people have someone they call and say: hey, something's not working. And then there's a backlash.

What should the first AI project look like?

There's a gap. On one side, a big project with a lot of time, resources, change management and money invested. On the other, something so small that it's borderline without any impact. The middle ground is where you don't find a lot of companies that actually do this. Most AI companies these days are either young, tech-savvy people who want to make money with AI, or professional software developers where everything is automatically in the big category.

What you need is something big enough to make an impact, but not so big that the whole project becomes too daunting and takes up too much resource. Something you can start with, and that grows with you step by step from there, with some experiments on the side. That's the part I work on with established companies.