This is what companies need for the development of artificial intelligence.

The use of artificial intelligence (AI) enables data-driven companies to automate elaborate processes in next to no time. Programming your own AI solutions, however, requires more than just a good idea and the necessary funds. The successful implementation of one’s “own AI” project requires three important attributes from companies above all.

TimeTime
  • If there are no suitable AI systems available to purchase on the market, companies can program their own.
  • To do so, however, companies have to meet a few criteria, as Malte Zuch, Lead Data Scientist at EOS knows.
  • In addition to the data quality and a competent team of developers, it depends above all on the mindset.

“Artificial intelligence holds great potential for working in a resource-efficient manner – away from expensive manual processes and towards automatic data processing. EOS realized this early on,” says Malte Zuch, Lead Data Scientist at EOS in Hamburg, Germany. The sector is too specialized for off-the-shelf software. The result: At the in-house Center of Analytics (CoA), his team is successfully programming its own AI solutions for the debt collection and investment business. Three attributes are crucial for this: well-trained staff, a positive error culture and last but not least data, data, data.

AI development in companies: Malte Zuch, Lead Data Scientist at the EOS Center of Analytics.

The better the data, the better the AI.

Data is the food that nourishes AI. Quality and quantity go hand in hand: the more data and the higher its information density, the more effectively it can work and the faster it evolves. “The learning process of AI is comparable to that of a child: a child can still read just as much – but if they only read comics, the learning curve flattens very quickly,” says Zuch. “The quality of the data becomes visible in the outcome of the AI. Let's take a simple yes/no prediction with an equal number of ‘yeses’ and ‘nos’. An accuracy rate of over 50 percent is already better than a random generator – anything below that indicates poor data quality. Depending on the desired result, you can try to improve the rate by enhancing the data.”

Data scientists wanted: well-trained a must – diversity a plus.

In addition to data, of course bright minds that think about the systems they are working on are also needed. But how is the perfect team of developers put together? According to Malte Zuch, it should be as diverse as possible: “In the field of data science, it is often about thinking outside the box. The more cultures and life paths that come together, the more angles we have from which to look at problems. Our team is made up of physicists, mathematicians, biogeneticists and, of course, computer scientists. Once we even had a bird biologist on board to find out more about swarm behavior.” Other important qualities include a natural curiosity, a high willingness to learn in a fast-paced sector and, of course, a certain affinity for numbers.

However, the search for suitable staff is often not easy for the company: “Artificial intelligence is still a relatively young industry with relatively few specialists – for a long time it was difficult to get into due to a lack of training options. In my time, there were still no serious training opportunities.” This certainly looks different today: In the past few years, many universities have expanded their curriculum to include courses relating to AI, which gives hope for less pressure on the job market.
AI development in companies: A data scientist looks at a data construct.

The mindset has to be right – especially at management level.

In addition to data and the staff required to process it, there is one thing companies need above all – a vision: “From my point of view, the most important factor is the right mindset with which to approach the subject – not only among the staff but especially at management level.” In a way, AI development is an investment in the unknown. Companies need a positive error culture to be able to deal with possible setbacks, believes Malte Zuch: “Not every AI solution brings the desired success straight away. Failure is practically a part of day-to-day business. Above all, companies have to be patient and find the courage to say: ‘We will do it better next time!’” After all, there are plenty of examples that show patience pays off. And the EOS Center of Analytics is one of them – at EOS, there is evidence the use of AI has already contributed to the company’s success.
Photo credits: Getty Images / Maskot, Getty Images / Bloom Productions, EOS