Since the 1950s, artificial intelligence (AI) has been a branch of computer science that deals with the automation of intelligent behavior and has enormous innovation potential. An important aspect of AI is machine learning, which is increasingly talked about today because it has a growing influence on economic processes.

In the high-tech, telecommunications, and financial services sectors, these systems, based on data analysis and interpretation, are increasingly being used to replace humans. Linguistic AI is already encountered in faster and better translation programs that can follow multilingual conversations simultaneously, making the former interpreter obsolete. In the field of automation, visual AI is increasingly being used, which plays a prominent role in industrial quality control and manufacturing automation via autonomous pattern recognition and can offer match-deciding competitive advantages. Partially or even fully automated warehouse management is already possible around the clock, replacing human labor on cost-effective terms.

Let us turn to the aspects that particularly affect a small or medium-sized company. On the one hand, there are many opportunities for SMEs to research the topic and develop it in their own interest. On the other hand, there are also risks associated with and fears of touching the topic.

The risks associated with artificial intelligence for SMEs

It is relevant for success to precisely define the use cases to work on in advance – AI literacy is key. There are several examples of companies that have started with use cases because they were simply deemed interesting. Not enough consideration was given to the economic benefit the project would have for the company. Fortunately, those who are just getting started can learn from mistakes made by others before them.

To choose the right use case for a promising start, a certain level of prior AI literacy is essential. By AI Literacy we mean:

1. to understand what AI is in the first place
2. to understand what AI can do (for your company)
3. understanding what it takes to build and maintain a working AI system

Two key success factors

The first factor is that we need to distinguish between smaller and larger projects and the investments they require. Corresponding projects can take between 6 months to several years to implement practically and thus productively, generating a return on investment thereafter.

The second factor is that we need to gain at least a provisional idea of the often-extensive ancillary activities required to implement a project. For example, studies say that at least half of the budget should go to culture change.

Examples

1. who will be the end-users of my project and what kind of training/coaching do they need?
2. what kind of APIs do I need to build to my existing systems to make this work?

It is a great challenge to build a predictive maintenance system. But what happens if there is no connection to the machine on-site or no way to set up a data pipeline from the IoT sensor to the central IT system. Maybe this is just a project not to start with.

Risk 1 – Little interest due to false start

If you are looking at AI and you do not narrow down a use case you want to work on that involves generating some kind of business value, it becomes difficult to get internal support to bring the targeted project to a successful conclusion. If you wonder – yes, this has happened many times in the past.

Example: If the first use case that is created is one that develops a translation engine, there may be vital clashes of interest in the business. This is the case, for example, if 10 translators in the company become unemployed as a result. It will be difficult to carry out the implementation without disruptions.

Reason: You need exactly these translators to build the system. It is unlikely that their support will be available. It is problematic to get skilled people on the team if they’re going to be cut if they’re successful.

This is one of the many reasons why it is important to choose the right use case with which to start.

Risk 2 – Systems developed without consideration of trustworthy AI

In short, AI is considered trustworthy if it is legal, robust, and ethical. That is, it is compliant with the rules, robust in the sense that it behaves the same way in similar situations but is also not harmful. And it is ethical in the sense that ethically defined standards are considered.

For example, Microsoft had developed an AI-powered tweet bot that spontaneously started posting offensive and racist tweets. Here, the developers neglected to address trustworthiness. Microsoft now has a framework in place that takes care of such potential issues.

Neglecting how to make your AI system trustworthy is one of the big mistakes a company can make. If companies develop systems that turn out to be untrustworthy, it will be hard to get back into the system originally envisioned. Cambridge Analytics comes to mind. Is the necessary trust still there?

Risk 3 – Having to go out of business because the competition is using AI

AI is trending, and most likely it is already here without you even noticing. Examples include the cab industry with the arrival of Uber or tourism with Airbnb, where the industry is being disruptively reshaped – with or without the help of artificial intelligence. If a company’s own competitors can automate certain processes and thus cut a variety of costs, they can offer lower prices and put the competing company out of business.

Fear 1 – Job loss

People are afraid of losing their jobs. And to some extent, this is certainly understandable. In the next few decades, certain jobs will be replaced or at least supplemented by AI systems. Certainly, AI systems are not perfect. We know that we need to verify the output of translation systems (in fact – I am doing this now as this text was translated from German). That is, jobs in this industry will not be eliminated, but they will change in terms of their focus. Today, a translator can check a text that has been translated by a system and make sure that the original content is conveyed. The translator can now handle many more texts in the same amount of time. He also can focus on other tasks in addition. Everyone will have to think about how AI could change their job and take the opportunity to retrain or upgrade.

Fear 2 – Machines make wrong or inexplicable decisions

There is a fear that machines will make decisions that are ethically problematic or wrong in a specific context. What happens when self-driving cars cause accidents? Or when you must decide about whether to run over a pedestrian or steer yourself into a tree. These are complex problems that the industry needs to solve and is working on. Novel frameworks are being developed to address them.

The opportunities

We can turn most of the risks and fears surrounding the use of AI into opportunities. Jobs can be made more interesting and more accessible in space and time.

Jobs can become more interesting when the boring, repetitive tasks are reduced to a minimum. This allows new tasks and responsibilities to be opened to employees. There are numerous examples of jobs that have already changed. Self-driving cars are changing the job profiles of cab or truck drivers. There are AI systems that help doctors diagnose cancerous diseases (tumor detection). We all use Google when we search for something. Or we rely on Google Maps when we want to drive somewhere. We use Grammarly or DeepL when we write a text or use certain marketing automation. Artificial intelligence is already there, helping us save time so we have new resources for more important things in life.

Opportunity 1 – Join forces with other companies to develop or invest in solutions

If you are in is a small business, it may indeed be difficult to tackle AI on your own. But that does not mean you should not engage with the topic of AI. Certain solutions can be bought – or used for free. Certain solutions can also be targeted in cooperation between companies to solve common problems in a network – also and especially with or by start-ups.

Example: The German company ControlExpert is currently digitizing claims processing for car damage. They work together with garages and insurers.

Opportunity 2 – Educate the workforce

Educating your own workforce about what is coming helps tremendously in keeping employees on board. Various studies show that C-level buy-in is the key to success. What needs to be shown is what options are available and what the future will look like in the company. At the same time, invest in employees to educate or retrain them in their field. This minimizes professional risks.

The added value of AI for SMEs

In summary, getting started with AI may be difficult for SMEs if they are going about it alone. AI is coming no matter what, and there are opportunities and risks that we need and need to mitigate at the same time. The added value for SMEs is the ability to join forces cooperatively, getting in external help, and quickly implement viable solutions through less complicated infrastructures.