Comparison

In this Dutch quote, [Manager 2] tells about the emergence of new business models by selling AI solutions to clients. In other words, the development of AI tools also creates new activities.

Furthermore, not only the bank employees are benefitting from the collaboration with the AI tools, but also the clients of Rabobank.

The AI tool is also faster in processing a large volume of data, which reduces the perceived process time. AI is supporting in two forms of collaboration. In the first form, AI is working together on the same tasks with the employee, for example, helping a client in the customer contact center of ABN. The second form of collaboration is working on tasks separately, as happens in the credit risk department of Rabobank. An important observation is the role of AI in relation to the complexity of the job. At the more straightforward jobs, AI is working together on the same case/problem/question. At the more complex jobs, AI is taking over the easier tasks, and the employee is working on the more complex tasks. An underexposed effect of the support of AI is the improvement in job experience, one of the pillars of workplace innovation.

[Employee 1] of ING expressed her immediate enthusiasm about working with AI.

[Employee 4] described the general experience using AI within ABN AMRO as “cool” An overview of the effects of the AI assistance per bank is presented in table 3.

Table 3 Effects of the assistance of AI

Bank Effects of the assistance of AI

ABN AMRO

• Process time reduction

• Improvement in job experience

ING

• Quality improvement

• Improvement in job experience

Rabobank

• Quality improvement

• Process time reduction

Training benefits the collaboration between humans and AI and, therefore, also beneficial to the perceived productivity. Regarding the training of the employees, the banks share the same opinion. ING has a mandatory training program and exam to safeguard the quality of the operations. [Manager 1] of ABN AMRO chooses to train the support agents and monitor their actions closely with the aim of improving the collaboration. [Manager 2] of Rabobank emphasizes the importance of training and guidance to prevent the incapability of

employees from happening. Furthermore, training should also prevent underestimation of the collaboration between humans and AI. However, in the interview, [Manager 2] mentions the lower need for training if an AI tool has a low usage threshold for employees. The level of the threshold totally depends on how smart the tool is developed.

Besides the beneficial effect of training on the collaboration between humans and AI, factors could also inhibit the collaboration, like resistance and a low level of reliability. An overview of the challenges per bank and the mitigation measures are presented in table 4.

Concerning the resistance level against working with AI tools, this level is low at all three banks. Most of the users are happy to work with AI. They recognize the added value immediately and appreciating the support of AI. Despite the low levels of resistance, there is the resistance of a small group of employees. For example, the (older) employee of ING lacking digital skills or not seeing the benefits of AI tools is more inclined to be resistant. Another example is the small group at Rabobank struggling with the fear of replacement. At ABN AMRO, there are no employees showing signs of resistance, but new employees have to get used to the tool. A more general observation is the reduction of resistance when employees experience the benefits themselves.

The reliability of the different AI tools is not on the desired level yet. Even though the tool of ING is relatively reliable, the AI tools of ABN AMRO and Rabobank are dealing with some reliability issues. The AI tools and ML models of Rabobank are a bit biased. Rabobank is applying techniques to reduce the bias. The AI tool deployed in the contact center of ABN AMRO is not on the wished level of correctness. Sometimes the AI tool suggests completely irrelevant answer options to the support agent. The AI tool is quite new and, therefore, still has to learn from the agents' input. The expectation of ABN AMRO is to reach the desired level in the next year.

Table 4 Challenges related to human-AI collaboration and mitigation measures

Bank

Challenges related to the human-AI collaboration

Mitigation actions

ABN AMRO • Reliability issue related to the correctness

• Self-learning through the feedback of users

ING

• Resistance of a small group of (older) employees due to lack of digital skills or not seeing the benefits

• Training

Rabobank

• Reliability issues relate to bias

• Resistance of a small group due to fear of replacement

• Application of bias reducing techniques

AI contributes to the perceived productivity of the team by collaborating with the employees in different settings. The way AI supports humans in this collaboration varies from the quality improvement experienced by ING and Rabobank to the reduction in the response time observed by ABN AMRO. AI could also assist by improving the work experience of employees as experienced at ABN AMRO and ING. As mentioned earlier, the role of AI is either taking over a task or working together on the same tasks. The role of AI depends on the complexity of the tasks. The collaboration between humans and AI can benefit from training.

For example, the two introductions day at ING and the related exam. Instead of training, a low threshold to use the AI tool can also be beneficial, as expressed by the manager of Rabobank.

Next to the benefits, there also challenges companies have to deal with related to the collaboration. The challenges companies need to consider are resistance and a low level of reliability. The low level of resistance occurred at ABN AMRO and resistance at ING and Rabobank. In summary, the collaboration between humans and AI contributes to the perceived team productivity, provided that the challenges are managed. Training can reinforce the

collaboration between AI and humans. In this collaboration, AI can assist humans in different ways.

In the three case studies, the banks expect the role of AI to grow continuously. More applications of AI will drive the future banking sector. Improved techniques will enable the models to be more intelligent and more autonomous. Important demand for this development is the safeguarding of quality, stability, and the human dimension. The developers among the interviewees stress that up-to-date infrastructure is necessary in order to intensify AI development. Also, the lack of available training data should be resolved, according to Rabobank. The end-users among the interviews support the further application of Artificial Intelligence.