AI agents: Please no more proof of concepts but real process automation, this is how it works

As you know, we are clearly convinced that AI is more than just GPT or editing text. Of course it is a good basis, but real benefits can only be generated with AI if the AI takes over the work and is also based on your own data. 

Read the blog or watch the NotebookLM video here.

IT service providers currently sell almost exclusively proof of concepts. This gives the impression that they don't really know where the journey is going either. But what if you took a different path? A path that combines innovation with concrete added value? 

But what if we tried to generate added value on the basis of an innovation process? How do we do that? Quite simply. 

Gather the innovative employees from your specialist departments and analyze the existing processes together. The aim is to identify areas of application for AI - especially where work steps generate a lot of effort but do not deliver any real added value. 

It is precisely these "high volume - low value tasks" that are also the subject of the Eisenhower matrix. 

Based on the current possibilities of artificial intelligence, tasks from the "not important but urgent" quadrant in particular are moving to the forefront of automation  

But which tasks generate large volumes with little value? 

Here are some examples for inspiration: 

Today, customer interactions take place on many channels, whether by email, support tool, form, telephone or messaging app.  

With an AI-based "dispatching" solution, information can not only be forwarded more quickly to the relevant office, but an attempt can also be made to answer the request directly or to respond as quickly as possible if the customer requires further information or documents. As always, however, it is important that such solutions are based on your data or your collected knowledge. Generic solutions based on existing models (LLM) such as GPT 5 without taking your knowledge into account are usually doomed to failure. 

Example 2: 

There are process steps that can be fully automated based on the customer's data and their specifications. Let's take the example of an insurance company that processes minor claims fully automatically based on the customer's details, including a picture of the claim. The AI can analyze the image, assess the offer, take the customer's past into account and also take the insurance company's specifications into account. The important thing here is that, in case of doubt, the human is included in the loop in the form of a check. In this case, doubt means that a statistical threshold value in the assessment is below a certain value. If the AI is only 60% sure that it has made the right decision, it is advisable to pass the case on to a human. 

Example 3: 

As a final example, we would like to mention an example from IT. If you are at home in the world of support, you will certainly be familiar with the problem: a customer's problem has been solved by means of configuration or development. All well and good, but who makes sure that the relevant documentation is updated again? An agent can do this work for you. Based on the processed tickets, the documentation, the configurations or the source code, the agent can adapt the existing documentation and give it to a user for approval. Simple and uncomplicated. 

And how do they find their use cases? Quite simply.  

As a first step, we will prepare them for the topic of AI in a workshop. What is machine learning, how do RAG solutions work, what is the difference between GPT 5 and 3.5 and many other topics we will tackle together. 

In the second step, they will document their possible potential on Post It's and then assess it in the following framework. 

Our aim will be to implement quick wins as quickly as possible and without too much investment, not as a proof of concept but as a real solution that creates genuine added value. 

The most impressive way to demonstrate the benefits of AI agents is through real experiences. That's why we let one of our customers have her say: Simone Siddiqui from SUPRISE KULTOUR AG tells us in an interview how she experienced the workshop and what specific areas of application were identified. 

We would be delighted to be able to uncover their potential too, true to the motto: We know that it works, we will show you the potential of AI agents without further ado. 

Jörg Bieri