TNO investigated AI adoption at four organizations. At HelloPrint, 80% of customer contact jobs were eliminated, but the work became more demanding. At three other companies, the savings were more limited.

 

AI can lead to significant cost savings in customer contact, but the quality of work for remaining employees may deteriorate. This is evident from an analysis by TNO published in late May 2026. HelloPrint is one of the four organizations studied and also constitutes the most extreme example in the analysis.

TNO's Ellemarijn de Geit, Kathelijne Bax, and Wouter van der Torre took a look behind the scenes at four organizations that have implemented or are developing AI applications: insurer asr, Customs, online print platform HelloPrint, and IT staffing agency LINKIT.

The researchers spoke with a total of 17 employees and managers via semi-structured interviews in the period from October 2025 to February 2026. They used the Technology Impact Method (TIM), an approach in which the consequences of new technology are systematically mapped from the perspective of technologists as well as employers and employees.

The researchers looked not only at productivity effects, but also at what changes in the work itself: the level of difficulty, task variation, autonomy, social contact, and mental load. That aspect remains underexposed in practice, the authors state.

HelloPrint: customer service from approximately 100 to 20 employees

Rotterdam-based HelloPrint has been using an AI application to answer customer inquiries via chat and email since September 2023. The tool, internally named Anna, runs on an OpenAI model and consists of multiple AI agents, per department and per country. Anna now handles 83 percent of all incoming chats and emails entirely independently, including canceling orders, updating addresses, and distributing compensation. This has led to a reduction in the customer service workforce from approximately 100 to 20 employees.

During the implementation, HelloPrint communicated that no one would be laid off, but that those leaving would not be replaced. Given the high turnover in customer service positions, this was sufficient in practice to realize the phase-out.

From the report: “Within HelloPrint’s customer service, an AI application named 'Anna' is used to automatically answer customer questions or messages via chat and email. The application runs on an OpenAI model and consists of multiple AI agents. The AI ​​agents are active per department (such as customer service, artwork, traffic, and finance) and perform various tasks for the respective department. For example, the customer service agent can handle complaints, but also answer product questions. Furthermore, the AI ​​agents are active in different languages; there are four different AI agents per country. Incoming chats and emails are analyzed by Anna and linked to work processes and procedures. The application works essentially the same for chat and email, although there are minor differences. Customers often use chat more interactively.” Anna was developed by Engaige, a start-up that emerged from HelloPrint.

Because Anna handles the routine questions, only the complex cases remain. Questions that require thought, coordination with other departments, or correction of Anna's errors. The difficulty level per ticket has “increased enormously,” says Jesse van Achteren (at the time customer success manager) in the report. The onboarding time for new employees has increased to two to three months. The emotional burden has become heavier, partly because employees regularly deal with customers who are already frustrated by an error or incomplete answer from Anna.

Changes in the work of HelloPrint employees

At the same time, the work has become more varied and challenging, and employees have been given more autonomy, including the freedom to compensate customers independently. An entirely new, full-time position has also emerged that was not foreseen: that of AI Product Manager, responsible for the continuous development and optimization of the AI ​​agents.

Customer satisfaction scores are slightly lower for contact via Anna (an average of 3,45 on a scale of 5) than for human contact (3,75). HelloPrint considers this acceptable: slightly lower quality of service at significantly lower costs and 24/7 availability. The FCR (for the entire customer service) rose from 51 to 62 percent compared to a year earlier.

In addition to Engaige, HelloPrint's experience with AI has also led to another start-up: a facilities contact center. weWow, founded by Robin van Driel Vis. Van Driel Vis is behind the build-up of 200 employees in Valencia that operated for HelloPrint.

Smaller or no profits at the other organizations

The other three cases in the study also show effects, but on a significantly more limited scale. At LINKIT, AI yields small time savings in drafting offer letters and freelance project descriptions. At Customs, a Retrieval Augmented Generation (RAG) application is being built that guides employees through complex customs manuals more quickly, but which has not yet been widely rolled out. At asr, an AI tool was developed intended to predict NPS scores based on call transcripts. That project was halted after the intended accuracy of 80 percent proved unattainable.

TNO states that organizations think almost exclusively in terms of productivity gains during AI implementation. The consequences for the quality of work only become apparent during the rollout, rather than during the design phase. This is therefore a recommendation from TNO: explicitly include which tasks within roles will disappear, which will replace them, and how the freed-up time will be utilized. Those who skip this step run the risk that productivity gains will go hand in hand with increased workload or a reduction in the scope of tasks.

The full report – with detailed explanations regarding the practices per company investigated – is available download via the TNO website.

(Ziptone/TNO)

Also read:

Engaige: radical automation of complex customer contacts

Featured, Human Resources