5 notes on how to succeed with AI

Don’t waste your time on Proof of Concept — it only ends up in a drawer somewhere anyway.

I’ve been around in the consulting industry since 2014. Before that, I was a very critical buyer of consulting services, which I believe has made me a better advisor and sparring partner for my current customers. In these glorious and adventurous days of artificial intelligence, this critical sense has proved particularly important. The market is so full of magic dust and unrealistic promises about what artificial intelligence may bring in terms of prosperity, success and eternal customer satisfaction. Working in this landscape every day, with some of the most brilliant data scientists I have ever met, I’m also convinced that the world as we know it is changing, and artificial intelligence plays an essential role. When I listen to lectures, read articles and books, or just reflect on the teams’ work, I make notes in my very analogue black Moleskine, which is always tagging along in my mobile handbag office. Here are five of my most important notes:

1. It’s not about technology, it’s about the business value you want to achieve

I’ve been in many a customer meeting where the customer wants to assess different technologies for AI. Seriously, it doesn’t matter. Artificial intelligence is about applying mathematics and statistics to data, usually solved through the programming language Python. It’s like spending hours on debating the brand of screwdrivers with your carpenter. Don’t. They will choose whatever works best for their purpose given the context of the job at hand. What matters to them, and even more to you, is what you want your house to look like when the work is done. If you fail in expressing what you want to achieve, you are very likely to be extremely unsatisfied, and you’ll have to redo the work.

2. Get out of Proof of Concept hell!

The terms Proof of Concept (PoC) and pilot are often used interchangeably. They do mean different things though, and the outputs are potentially VERY different. A PoC is about testing a concept or a technology, to verify whether it makes sense in a given context. Typically, a PoC is executed on very limited prerequisites, with limited data and low precision requirements, and without interaction with the surrounding operational environment. Consequently, the results of the proof of concept typically ends up in a drawer far away from the CFO’s nose. A pilot on the other hand, is a limited production scope, which means you start small, but still aim for a full production scope, and you have to think through the operational implications from the very start. It also means that you’ll be encouraged to prioritize processes that are truly relevant for your core business.

3. Your most important asset is STILL your employees

Let’s have an example: A utilities company wants to develop a machine learning model to predict downtime of windmills, in order to schedule maintenance with a proactive, and preventive approach. The technical manager is an experienced guy with a long track-record in maintaining heavy equipment. No matter how good the predictive maintenance model you develop is, you still need to convince the technical maintenance manager that his experience and gut feeling is about to be replaced (or better, complemented) by an algorithm. If he is not in on this from the beginning, don’t waste your money (or your time), because in his eyes, the algorithm will fail, no matter how useful it could’ve been. Yes, artificial intelligence is about algorithms, but your business is ultimately about people (your employees), who again work with people (customers).

4. Making the decision to invest in AI is great! But remember that THAT’s when the hard work begins…

I’ve seen this scare some leaders into decision paralysis. Don’t worry, just start small, and always keep your eyes on your business targets. Artificial intelligence is not magic. It’s hard work, experience and science. All those data you have collected for years are lying around waiting to be utilized. Structured, systematic work, by resources that are specialists in understanding the science of data, in combination with your best domain specialists. Together, they will unveil new insight about your business processes. Potentially worth millions. Perhaps hundreds of millions. I’ve seen it happen.

5. The speed of change is increasing exponentially

Do you still believe that you can spend six months on strategy revisions? And another six months concluding on the business case? Sorry, those days are over, and they’re not likely to come back any time soon. You still need your long-term foundation, and, on top of that, you need to develop an agile top management approach, continuously learning from customers, sparring partners, and expert resources in the market. The days when a company could build and nurture all required competence and capacity internally are officially over.

Følg 3min på Facebook, så blir du oppdatert på de ferskeste sakene!