Of data, use cases and the acceptance of AI

Data, data, data! The secret to successful use of AI in your SME

The abilities of AI tools like ChatGPT or Dall-E never cease to amaze. But how can artificial intelligence be used profitably in your company? bbv expert Michael Maurer outlines the most important success factors – suitable data, a dash of courage and the willingness to experiment with creative ideas.

10.05.2023Text: tnt-graphics0 Comments
KI Projekt

There is no perfect recipe for using artificial intelligence (AI) in your company. “But there are definite factors that make an AI project successful”, says Michael Maurer, Senior Consultant and Solution Architect at bbv.

Maurer is convinced that artificial intelligence will sooner or later find its way into every company. Its potential is enormous and the possible applications are unlimited. We already discussed some use cases in our last blog post.

“It is essential that companies think about whether and how they want to use artificial intelligence”, says Maurer. Regardless of whether a manufacturing industrial company or a service provider. “The main thing is not to ignore the topic.”

Which requirements help to introduce AI successfully in your company? And where do the challenges lie? Michael Maurer provides clarity.

 

Requirements

Yeah, data, baby!

“Data, data, data!” This is how one workshop participant summed up the most important requirement for using AI. Michael Maurer, joint leader of bbv’s AI workshop, agrees: “Nothing works without data. Ideally the data is already available in a structured, i.e. usable form. This can be your own data or data sourced externally.”

But there is still a lot that can be done with semi-structured data once it has been converted into a usable form. It is also possible to unleash AI applications on Excel datasets or video files, for instance.

Yet the data does not necessarily have to be available from the outset. “Depending on the use case, a company starts by collecting the necessary data – for example with custom software developed by bbv – or it sources the data externally”, says Maurer.

If the data pool is too small or the data is available in a form that is not usable, it is possible to buy in ready-made AI services. Providers like Microsoft, Amazon and Google offer APIs that can be easily integrated into your own applications, for example for speech recognition, tagging of images or for text recognition. OpenAI with its Dall-E and (Chat)GPT models can also be integrated via APIs.

The be-all and end-all: the use case

Without a use case, however, it doesn’t matter how much data you have, says Michael Maurer. He therefore sees two basic approaches for making AI usable. “Either artificial intelligence helps to improve the quality and efficiency of existing processes. Or else it justifies a completely new business case.”

The first option involves finding the hidden benefits of existing datasets. When developing a new business case, on the other hand, it helps to have a data-driven mindset and the support of a consulting company like bbv.

 

Success factors

Everyone starts small

A clear objective, a good ROI, a strict roadmap: What is usually best practice often prevents companies from experimenting with AI, says Michael Maurer. “It makes more sense as a first step to have an open mind and see what the data can produce using different models.

This could be customer data that is clustered to gain useful insights, or databases that can be analysed for patterns in order to develop a use case. “It takes courage in the beginning to try out such applications. If you start small and the results look promising, you can extend the scope of the project.”

A prototype can then show whether potential exists.

 

Leveraging existing AI tools

ChatGPT has made us realise that the barrier to productive use of AI in everyday life is very low. “You don’t need either your own data or a data concept for this, rather simply an idea for new applications,” says Maurer. A lot of computing power or a large cloud infrastructure is also not necessary, because external AI tools can be integrated into your own applications via APIs.

Use cases for AI tools can be found in practically every industry. They can be used to translate, write and improve texts, create and analyse images or write entire code snippets.

 

Promoting acceptance of AI in the company

Transparency and information on what AI can and cannot do is key to its acceptance in the company. “AI makes certain tasks obsolete, thereby allowing us to concentrate on activities that add value”, says Michael Maurer.

Management must therefore be able to clearly demonstrate the benefits and potential uses of artificial intelligence.

These can include:

  • Routine tasks can be automated and completed more quickly with AI. This increases efficiency and leaves more time for those tasks that still need to be done by humans.
  • Dangerous activities can be completed by intelligent machines, such as tests in contaminated environments.
  • AI helps to combat the shortage of skilled workers: Speech-to-text applications allow medical professionals to dictate reports and convert them into text, for example. This leaves more time for patients.
  • Incoming and outgoing goods can be analysed with AI and forecasts produced as to which goods are in demand and when. This optimises capacity utilisation and inventory levels.
  • Artificial intelligence provides insight into customer segments and customer behaviour. This allows the customer journey to be improved.
  • AI is already used and accepted in many places, such as in product recommendations or individualised newsletter content.

 

Challenges

Corrupt data

The results produced by AI are always only as good as the data used to generate them. If the data source is biased, incomplete or incorrect, this leads to problems. Large amounts of high-quality data are therefore important. But it is also crucial that the results of AI are examined by specialists with experience and know-how. People and machine together produce the best results.

 

Uncertain legal situation

The number of applications based on AI has increased considerably in recent months. The law as it stands has been unable to keep pace. Content creators therefore complain that AI applications like ChatGPT or Dall-E use legally protected data, text or images to train their algorithms.

But changes are becoming apparent: “It’s now possible to block data for individual tools”, says Michael Maurer. “For example, image databases block uploaded images for machine learning activities on request.”

 

Conclusion: Courage and good ideas are needed

For Michael Maurer, artificial intelligence is now an integral part of our everyday lives. He therefore encourages companies to gain initial experience with AI. “Above all, this takes a dash of courage and a good idea”, he says. When it comes to identifying untapped potential and then utilising it, bbv is also on hand with its expertise to provide advice.

The expert

Michael Maurer

Michael Maurer is a Senior Consultant and Solution Architect at bbv in the financial service provider & transport fields. He focuses on digital transformation, innovation management and the design of (cloud) software solutions. Thanks to his many years of experience in the SME area, he understands their challenges and can highlight targeted solutions and implementation options.

Safe and ethical use of AI

The AI Act: Regulatory framework for artificial intelligence

AI/KI
Industry specific AI solutions

Beyond the hype: How bbv’s AI Hub is shaping the future of specialised AI

AI/KI
Artificial Intelligence

AI training with limited data: How you can optimise your AI models even with limited data

AI/KI

Attention!

Sorry, so far we got only content in German for this section.