Are you on board with a glitchy artificial intelligence?

Artificial Intelligence (AI) is one of the hottest topics right now. Sometimes it is seen as the most important factor for future competitiveness, and other times it is perceived as a threatening, emotionless army of superior robots. Some people even say that chatting with ChatGPT is more satisfying than talking to a human. It’s hard to say whether this is due to the sophistication of AI or if it reflects something about our personal social lives.

One thing is certain: even if our new friend doesn’t need rye bread, the energy consumption forecasts for AI technology will surpass the energy usage of entire countries in the coming years. Something is truly changing.

You will end up riding the AI wave—whether you want to or not

Using AI is a long-term journey that progresses step by step. An organization’s business operations consist of various types of processes, and depending on their content and maturity, the opportunities and priorities for utilizing AI vary. Adopting AI as a tool is not a quantum leap but a long-term, step-by-step journey.

AI does not change the fundamentals or requirements of digitalization management. It is a tool, and everything starts with understanding the need. If you haven’t used image editing tools before, you probably won’t benefit from the fact that Photoshop can now create a unicorn in your forest landscape with the help of AI.

Sometimes strong technology trends are approached through technological possibilities without a real business need. This might impress the management team and employees during presentations, but these tools often don’t get adopted in practice. Nevertheless, these explorations are valuable and can lead to innovative outcomes that no one even dared to dream of. These efforts also measure digitalization capabilities and vision. True breakthroughs always require courage, and sometimes it helps to have a bit of faith in technology and a good grasp of mathematics.

Everything in AI starts with data

Another challenge for trend hunters arises when the maturity of an organization’s digital elements does not meet the baseline requirements of new technology. We want to identify the most likely users of a new service from our customer base, only to realize that customers are not providing accurate baseline data. It is wise to start AI adventures by understanding the content and quality of the data AI uses, as well as the processes and human usage patterns that produce this data.

Depending on the use case, data can be document-based or directive-type point data that can be provided all at once, for example, to a chatbot for customer service needs.

When well-executed, this is one of the most potential use cases for AI, through which most organizations can start their journey.

The potential—and complexity—of using AI changes dramatically when personal customer data is involved. Data protection and privacy requirements impose strict limits on use cases. If these are manageable and customer data is in good shape, the customer experience can reach a whole new level. This is also one of the most feared use cases related to replacing humans, but in reality: there are better uses for humans than answering the same questions over and over again.

Another angle to AI utilization is continuously changing and enriching—or always new—data, which requires a different approach. Summarizing large text masses is a growing and versatile use case. Copilot works excellently for this and can condense large entities into compact packages. However, these summaries have their risks, and they don’t always hit the mark. For some reason, there might be a clear error. If you know the material well, you probably also know where the error originates. If you don’t, it’s hard to identify the flaw in a seemingly convincing presentation.

In this case, it’s about a data error somewhere in the source material. In the bigger picture, this relates to the often-feared AI nightmare scenario, where an AI-generated erroneous summary reinforces further AI use, and an initially single error becomes a prevailing perception. In the future, we will likely see very convincing materials on topics that are complete nonsense, as AI teaches AI.

Next, we might need some kind of AI therapists to fix the quirks of algorithms and bring misbehaving bots back to their senses.

When starting to utilize AI, it is wise to begin by reviewing business processes and identifying repetitive and high-load entities. AI—and automation—often have the most demand in these areas, provided the data content around the process is in order. These often represent a mature playing field for AI.

Another good starting point is to seek understanding from large, changing datasets and identify deviations, trends, and correlations. The boundary between AI and analytics can sometimes be blurry, and it might be wise to start from an analytics perspective. Once the data is under control, AI capabilities can be added to the toolkit. Many systems have started to support these possibilities as ready-made features.

This angle of digitalization must also be mastered by people, and you need to be in the driver’s seat. You can move to the back seat once your data-driven friend becomes familiar. If you want to discuss this further, click on the calendar to book a sparring session with a perfectly ordinary sandwich-eating human. We can both learn new things.

Artificial intelligence sparring conversation