What if your AI company is a restaurant?

An interview with Mieke De Ketelaere, better known as Madame AI. Professor at Vlerick Business School and has more than 30 years of experience in the big tech industry.
What if your AI company is a restaurant?

What if your AI company is a restaurant?

An interview with Mieke De Ketelaere, better known as Madame AI.
Professor at Vlerick Business School and has more than 30 years of experience in the big tech industry.

First things first

Mieke: “A lot of companies make the mistake of thinking data is enough to start being an AI-driven company. It’s not. Even when you hire a data -scientist to create models on your bunch of data, it won’t mean anything unless you align the decision you want to make on the outcome of the models to your business plan. You have to start from your business problems, not from your available data.”

AI is one of the biggest hypes right now. But as it is with any hype, you should probably first think before you dive in head first. Just because you have data and might hire a data scientist, doesn’t mean you’re “doing AI” in a proper way.

Mieke De Ketelaere

Check your ingredients

“When you’ve analyzed from your business challenges which data you would need, you’ll still have to make sure it’s useful data. Much like in a restaurant, it’s not because you have a bunch of ingredients, you’ll make a delicious dish and create a good running restaurant over time. And it also depends on how fresh and complete the ingredients are in order to make your dish. The same goes for data: analyzing click behavior of a target group in the retail sector 6 months later will probably not give you accurate behavioral insights on your consumer.”

A star chef isn’t enough

Congrats, you have quality ingredients – uhm, data – and you hired a star chef – uhm, the best data scientist there is. But does that fix your problem? Probably not…

“A star chef will be able to clean and cut carrots, but he won’t get much energy from that. Same goes for your data scientists. If you follow the 80-20 rule where 80% of your time goes into preparation and 20% in the actual implementation, you know your data scientists will soon quit after tedious hours of preparation. You need to hire different profiles to do that work, namely data engineers, just like you hire kitchen personnel for your restaurant.

Don’t promote yourself to head-chef

A lot of companies really want to get into the AI game, but do it without any strategy. Managers push their employees to learn on the spot, risking not only quality but also team fatigue. Give people either the time to learn, supported by an expert, or hire a professional team. Anything in between will set you up for failure.

Forget about the certainty you have with rule-based systems

You might get mad when AI makes “mistakes”. But don’t forget to realize that AI applications are not rule-based software.

“AI doesn’t obey the “if – then – else” rule. An AI application will always provide an answer, and that will always be with some probability but never 100%. As a company using AI, you need to realize this margin of error will always exist. So don’t use an AI system if you want to have a perfectly consistent system. You should rather use it as a helpful system that provides you with insight you might not have been able to detect yourself due to limitations of human intelligence in certain conditions.”

Hire a team of experts

“An entire AI team takes time to set up and organize, but don’t go off buying untransparent pre-trained models online. Because the thing about AI not having rules, is that they are trained with specific data in a specific environment. As long as you don’t know these details – and you probably won’t – it’s a risk using it for your company as the context will be different. So my advice? Ask for a transparent explanation on the training data and context and hire an external team to guide you to retrain the system for your context if you don’t have an AI team yourself.”

Start small and grow

“Just like a start-up restaurant, you don’t just start with an elaborate menu of 30 dishes. You start with maybe 10, perfect those and then add 5 new dishes. Also work agile in this process: start with the integration of one or two AI systems in your processes, evaluate, optimize and build further on.”

“Also get your entire company involved: only when everyone realizes they are an important part of the chain, will it work. Streamline your processes and people from the bottom to the top and vice versa.”

Ask for a review

After you have everything in place, check if the client also likes your menu.

“You could serve any dish of any quality: a hungry client with no other options will still eat it – even if you just microwaved a pizza. But the client will never come back. Also the context of your restaurant might change over time. So make sure to check the results of your cooking efforts. Are customers coming back or not?

Maybe your data analyses and actions didn’t do what you expected, didn’t resonate with your potential clients, or were simply not meant for this target audience and you need to reevaluate.”

Curious about how to get started with artificial intelligence or how to accelerate it in a valuable, meaningful way?
Menu Close