Human in the Loop – review | Ziptone

Oliver Talens views human control in the deployment of AI as a necessity, not an option. That is why he wrote a book about 'collaborating with AI'. Or as he puts it: “About making choices. About daring to say: we are taking over again here. Not because we are afraid of AI. But because we know what we have to lose.”

Talens (marketing manager at Qaitbay) argues that AI is not a neutral tool, but a magnifying glass that exposes what organizations find truly important. After all, AI follows our instructions, optimizes our KPIs, and reproduces our (un)conscious assumptions. Talens wonders: “What do we want that remains unmistakably human?”

That is more a question from someone who is worried than from a shallow tech enthusiast who sees AI as the new miracle cure.

Talens' book is structured in five parts. He first addresses the organization: what does the arrival of AI mean for functions, service delivery, and management? Next, Talens dissects what AI can mean for the customer relationship. In the third part, he highlights the role and position of the employee. Will the employee become a slave to AI, or will the employee be replaced? In part four, Talens analyzes AI itself: AI is a reflection of human intentions, biases, and design choices. In the final part, Talens makes predictions about the future of work, professions, and roles.

Does European AI exist?

In this self-published book, Talens pulls out all the stops. This is already evident in the introduction, a relatively long build-up. So, let's move quickly on to those five chapters. This choice of structure is sound, yet the five parts do show some overlap. Furthermore, not everything is elaborated upon with equal thoroughness, which frequently leads to the question: 'but how?' For instance, the subject of geopolitics is mentioned a few times, but the exact implications for, say, customer relationships or your distinctive capabilities remain rather underexposed. Can sovereign AI add value to your brand, and if so, how? What does dependence on American tech mean for European companies? Does European AI exist?

Or what about inclusivity—something regarding which companies often have the best intentions, according to Talens. Those intentions are often enshrined in policies, programs, and annual plans, he writes. “But an AI system knows nothing about intentions,” Talens adds; after all, does AI only look at patterns, data, and history? In doing so, the book bypasses the question of what information the AI has access to. Perhaps that is a deliberate choice, because it leads into technical details—and that is not what this book is about. In theory, you can perfectly well have an AI tool investigate the overlap between (inclusivity) policy and practice, or in other words, intended plans and data that say something about execution—provided that inclusivity has been made measurable. Fortunately, Talens argues further on for making 'vague' KPIs measurable.

Meeting agenda

A statement further down the line – 'AI increases structural exclusion' – is, in my opinion, also a bit too simplistic, just like 'you can't defend what you don't measure'. Precisely in conversations between professionals, it is perfectly acceptable to discuss questions such as 'what kind of organization do we want to be'. Organizational culture is, after all, pre-eminently a human product – in other words: a translation of values and norms into behavior. AI, too, should be on that discussion agenda.

The assignments with which Talens concludes each section of his book are well chosen. They include challenging questions for strategists, policymakers, and leaders. Talens aims to prompt these groups to reflect on the impact of AI; the question, of course, is whether this will happen proactively or only after the effects of implementations begin to take shape. However, the questions Talens raises are as challenging as they are useful. In our view, this demonstrates where the strength of this book lies: acting as a thorn in the side, asking critical questions upfront, and engaging in dialogue. Contact center managers might skip this step, or it might remain limited to operational implications: when and how do we tell our employees, and how do we assemble the project team?

From that perspective, I can imagine that in a few years we will face a new crisis in customer contact. Too much automation, too little human contact; customers and employees walking away. The contact center had become a little factory again after all. This book implicitly warns against that, but what if you cannot imagine that future? In other words: what if you don't want to again how you need to think about AI?

What exactly is collaborating with AI?

How and when does the human aspect play a role in an organization riddled with technology? That is the question this book addresses – but in certain respects, the author’s struggle with this issue also seems to come to the fore. The advantage of this is that this struggle, too, prompts reflection. “Human in the Loop in a customer context means that personalization never becomes fully autonomous. There must always be a moment when a human can reflect on the outcome,” says Talens. But aren't we already too late then? Or does collaborating with AI ultimately come down to evaluating (or adjusting) an automated process after the fact? Because that is often how it works now: the employee's help is called upon when the system or customer determines that an automated process is not running smoothly. The human employee then acts more as a 'repairman' than as a matter of collaborating with AI.

That this collaboration between humans and AI is far from being fully crystallized is also evident from another dilemma exposed by this book. Talens points out several times that humanity must be central to the deployment of AI, but at the same time is human in the loop Not really scalable. That calls for a good design (in advance) and (afterwards) for controls and room for exceptions. This is precisely where it gets interesting, but as a reader you are left with the question 'but how are we going to organize this?'

AI is not taking over the job, but clusters of tasks.

Naturally, we read this book with special attention regarding the customer contact representative. Talens also argues that here, it is not the job itself, but rather 'clusters of tasks' within jobs that are being taken over by AI. Like many others, Talens takes the approach that work is changing and that the key question here is how we can strengthen human roles with AI. However, if clusters of tasks are being taken over, surely fewer people are needed for the same work; otherwise, the deployment of AI leads to higher rather than lower costs. This book also overlooks the question that will sooner or later arise within every large contact center team: will my employment contract be renewed or not? For, ideally, the demand for capacity decreases due to the use of AI. Otherwise, 'collaborating with AI' is nice, but it yields no efficiency advantage. And personalized marketing or service delivery won't take off if every AI analysis has to be checked by a human. However, a great deal of attention will have to be paid to the design of these types of applications, something Talens also emphasizes in his book.

Who controls what?

Talens rightly points out that employees tend to monitor the output of AI, which effectively results in double work. In many organizations, it is even policy rather than mistrust. This concerns the practical collaboration between humans and AI, and that is precisely what we would have liked to read more about. What does this actually mean? Is it challenging to work with AI? Do employees train their own AI? Could 'collaborating with AI' also amount to 'training AI' for some professions? Talens also emphasizes that AI does not steal jobs, but takes over work mindlessly – and that this is actually a blessing. Or could mindless work actually be a blessing for some people?

Problem directors and brand ambassadors

The fifth and final part deals with the future of professions. With a greater focus on people, ethics, and culture, Talens outlines the manager of the future. Several other work areas are also discussed, including customer service. “Routine questions and standard procedures can be fully automated, giving employees more time for more complex cases, empathetic contact, and proactive service. AI will become an assistant that provides data, context, and patterns, enabling employees to act faster and more targeted,” says Talens. Talens also expects customer service employees to increasingly transform into problem managers and brand ambassadors.

The question is what exactly that will look like. What does that mean for leadership, recruitment policy, and the cost structure?

Human in the Loop by Oliver Talens is a decidedly challenging exploration of the impact of AI on organizations, customers, and employees. Human in the Loop Fortunately, it does not deal with the technical implementation of the technology or data issues. Enough has already been written about that, and the shelf life of such books is very limited given the speed at which AI is developing. All praise, therefore, for the fact that this book addresses the ethical and organizational questions that arise when we unleash technology on human systems without considering the consequences. Instead of ready-made answers, this book therefore offers heaps of inspiration for anyone who has yet to start formulating an AI strategy.