The Corona virus has been shaping business across all of 2020. But there was much more happening in the industry.
The amount of companies launching products in the AI domain has more than doubled in just a few years. And this trend is likely to continue.
In addition, more and more “out-of-the-box” AI solutions are offered to enterprise companies e.g. smart pricing tools, fraud detection software or forecasting solutions. There is a clear trend to build products as SaaS (Software-as-a-service). For startups this essentially means building one product and scaling it across multiple customers with minimal adaption.
However, comparing our most recent client conversations with the years pre-2020 we noticed an interesting twist.
Very often, companies don’t want to buy a generic product, they want to solve detailed and complex problems. For example, we talked to a mechanical engineering company who wanted to extract structured information from highly complex documents and technical drawings. Whereas there are products who promise to help, they discovered that all of them failed on their very own “data”. Products (currently) cannot always meet all requirements that are needed to do the job. Very often they must be developed individually to match the data quality and quantity of each business.
Everyone is talking about AI, but very few companies are using AI solutions in production. Not really a new trend of 2020, but a pattern we often see once new technology hits the market.
Our perception is that most companies haven’t worked with AI and won’t work with AI for the next few years. This is because their business is just “non digital”. However, they start to think about the first steps. In case they would work with AI in the future, what would they need? Even the most non digital companies start to think about data they could collect, although being far away from using AI in production.
When it comes to the type of companies, the landscape is quite diverse. We got to know medium sized companies with a fully implemented AI strategy. They have all their machinery equipped with sensors and run an advanced cloud infrastructure to support data collection and processing. We also got to know billion dollar companies with no data infrastructure, still collecting most of the process information via paper.
Another interesting trend is that we see two different types of mindsets evolving in the industry. The companies that see AI as a chance to be successful and others that perceive AI as “homework” that needs to be done.
One quote from one of our clients who is building large industrial machines:
“In 20 years many companies in Asia will be able to build our machines. And probably at a much cheaper price. However, if we manage to provide the best service for those machines, basically, a machine that never fails. And if we are able to make our machines more intelligent so that they are as easy to navigate as a smartphone, this will truly put us forward in the market.”
We hope that more companies will adopt that mindset in the coming years and start seeing AI as a chance to truly diversify their service and product.
In the past years, many companies hired Innovation Managers for the true purpose of exploration. To show: “we are innovative”. Not a fact, but our perception when talking to businesses in the German “Mittelstand”. Many times these positions were neither equipped with their own budget nor a tech team backing them up with support. In Germany, it is common to say “zahnlose Tiger”.
This is changing dramatically. Many companies allow their innovation and digitalisation managers to do their own investments and the C-Level really starts to listen. A huge step in the right direction.
Decision makers in companies get swamped by cold calls and emails offering tons of services from the AI industry. But is it enough to just buy those services?
We discovered that there is a huge transformation going on from just buying services and products towards building something together. Something that lasts.
Big names in various industries start programs to get in touch with start-ups and service providers but with a totally new motivation. Instead of buying services and products, the new approach is to develop a product together. For example, the enterprise company provides data, the AI company develops a solution based on that data. Together, they sell the solution to clients of the enterprise company.
And let’s be serious, the providers, big companies partnering up with and creating solutions together are the same companies, they bought the service from one or two years ago.
The proof of concept years are over. Whereas some years ago people got excited about a proof of concept, nowadays the focus is much more on showing business value. Companies don’t use AI just to use AI, they want to increase revenue or lower their costs.
And that is challenging, because often it is hard to measure what impact an algorithm really has.
The first question to answer is: “How much business value is behind the AI use case?” What do I spend developing it, and how much does it make me. Sometimes the answer to that question is easy, but unfortunately, more often it is difficult.
For many use cases like optimal pricing, predictive maintenance, recommendation engines etc. it is often possible to do comparisons between processes “untouched” and optimised by the AI solution, so called A/B testing. But how about a forecasting solution? How do you measure success if “testing” means changing a key part of your business strategy?
There is a need to be creative to find provable numbers. Instead of delivering or developing a solution, service providers need to prove the business value first. An enterprise company doesn’t pay dozens of thousands of euros without knowing what they get. Service providers have to create a low entry barrier with a fast proof-of-value.