Self-proclaimed 'AI duo' Remy Gieling and Job van den Berg wrote five books on AI in two years. That is not necessarily positive: a lot of reading material is published on this subject; its shelf life is limited because developments are moving fast. But this practical guide is recommended. Presumably, a sixth book can be published in about two years, about the design and structure of the AI ​​organization. Now first attention for book number five.

 

This book is about AI agents: software that can perform tasks independently. This is different from generating text, sound, still or moving images based on a prompt – typical tasks that generative AI can do for us. This fifth book by Job van den Berg (1989) and Remy Gieling (1988) is first and foremost a cheerful book that drips with enthusiasm. And at the same time, the book is nuanced and realistic about the opportunities and risks surrounding AI agents.

The beginning is also good: the authors managed to get FD journalist and economist Mathijs Bouman to write the foreword. In it, Bouman mainly puts the problem of stagnating productivity on the agenda. Not against the background of an ageing population. But from the perspective of the mountains of digital information that we as screen workers have to digest. That must and can all be done much smarter. Fortunately, AI is a promising solution, Bouman believes. Gieling and Van den Berg then start by saying that they find AI 'overhyped', but the technology 'underused'.

Spoiler alert

AI agents are a real revolution in our information society, because unlike classic software they do not follow predetermined rules, but try to realize a given goal. AI agents can reason logically, have 'learning' ability (at least there is a memory that helps to continuously improve performance) and can interact with other AI agents and tools. Think of making an appointment in the agenda or making a phone call. That is also why this book is - spoiler alert - a must read for customer contact professionals who run a contact center or are responsible for, or involved in customer experience.

AI agents as part of an organization

What is remarkable is that the explanation at the beginning of this book about how AI agents work is reminiscent of the standard work on organizations by the founder of organizational science after just a few pages. Henri Mintzberg (Structure in Fives, from 1983, but nevertheless a must read). AI agents also have roles and tasks, must coordinate with each other and have to deal with control and supervision. And there are different types of AI agents – so task differentiation is also a Mintzbergian principle that returns in the virtual world.

Generic AI agents go a lot further than ChatGPT. They are quite autonomous (for example, they can register a domain name and reflect on their own steps in advance), but many tasks still require checks from the client. In the ideal form, an AI agent asks questions to its owner, but domain knowledge and training are absolute success factors. In addition, the various suppliers are working hard on standards. This explanation makes it clear that a lot is still coming our way.

Interviews

The book contains a number of interviews with both 'ordinary managers' and typical AI entrepreneurs. The latter also makes the book valuable: everyday practices at NN and ABN AMRO Bank, for example, and perspectives alternate. But the common thread is formed by numerous fictional examples that show how this everyday practice will change. Search engine optimization, for example, is first supplemented with (and possibly later replaced by) optimization for AI agents.

Of special note is the interview with Isaiah, the CEO of Bland.ai, which claims to automate 2.000 customer service jobs every week. “Think of phone calls that are completely handled by AI.” It should be noted that Americans have a long tradition of robocalls – wanted and unwanted.

Ziptone also cannot avoid introducing a Ziptone AI agent, which allows readers to retrieve tailor-made information from the field in natural language. On the backlog!

Step by step into the world of AI agents

Where the first chapter takes you on a bird's-eye view, the book offers valuable in-depth insights from the second chapter onwards. It also shows that the thinking around AI is already quite advanced. The most important players, the (huge) investments in it and the underlying economic forces are well explained.

No matter how well AI agents can work, they only have data. They have no gut feeling, no prior experience, no contextual knowledge or insights and cannot take into account what their clients know but do not make explicit. It is precisely this dependence on data that raises new problems, for which solutions are now also being devised, such as other (calculation) models, other infrastructures and other hardware.

Special attention is paid to vector databases, RAGs and knowledge graphs, which are explained very well in this book. Understanding these – plus the explanation of what is on the horizon in this area – is important to understand the potential of AI agents.

“Interbellum”

The fact that models are becoming more powerful due to all the improvements also has a downside. Gieling and Van den Berg warn of a temporary period in which criminals can easily get started with AI agents, while the technology does not yet have a good answer to abuse or a good idea of ​​​​slips - such as buying a carton of eggs for 31 dollars. The design of learning and self-monitoring agents is also work in progress. This will also be discussed Eliette Vaal from The Data Lawyers: there is already a EU AI Act which is based on a structured risk assessment including measures or obligations. Companies must build up knowledge, skills and routine in this.

Useful models and step-by-step plans

Two-thirds of the way through the book, the authors provide practical tools for getting started with AI agents. These are the models, diagrams and step-by-step plans that you don't see in the slick presentations of AI agent providers. Embracing AI has an impact on your people, your processes and your data. Gradually, all these things will have to be properly aligned. The fact that some things will also gradually disappear completely or partially - think of certain jobs - is less discussed.

But at the end of the book, the authors discuss the new organization where AI-gents and people work side by side. And with that, Mintzberg, who we also mentioned at the beginning of this article, returns. Gieling and Van den Berg have taken the trouble to think about new structures and forms of collaboration. This part could have been further elaborated, but perhaps it is too early for that and that is more something for book six. However, a lot of attention has been paid to change management – ​​the change process that you will encounter if AI is going to become a systematic part of your business operations.

Support, not replace

The book does not answer the question 'will AI take over my job'. Functions will change: that has been the prevailing narrative so far. In the customer contact sector, too, almost every manager is quick to emphasise that AI supports employees, not replaces them. We will therefore classify this under the heading 'the knowledge of today'. Anyone who reads this book cannot help but conclude that certain professional groups will be decimated in the coming years. Simply because the bulk of the work - or a significant part of the task package from a function - can gradually be taken over by AI agents. Will customer contact also be affected by this? And if so, in what way? Anyone who reads carefully between the lines will find the answer.

(Ziptone/Erik Bouwer)

 

AI Agents – practical guide on how AI is changing every office job. Job van den Berg and Remy Gieling, 2025. Self-published by AI.nl, 267 pages. www.ai.nl/agentsboek

Featured, Opinions, Review, Technology