Banks operate in one of the most trust-sensitive industries imaginable. Customer proximity, regulatory compliance, and data security are not constraints layered on top of the work — they are the work. This makes AI adoption particularly complex: the tools that promise efficiency gains are precisely the tools employees and customers are most cautious about.

Djurslands Bank recognized that simply rolling out Copilot to employees would not be enough. The question was not whether the technology worked, but whether the organization was ready to use it well. Early signals were mixed. Before the intervention, employee scores across all four measured dimensions — AI self-efficacy, psychological safety, growth mindset, and sense of partnership — ranged between 2.8 and 5.4 out of 10. This was not a picture of resistance, but of an organization that had not yet found its footing with AI in daily work.

The bank chose to treat this as an organizational challenge as much as a technological one.

What makes this case interesting from a Hybrid Intelligence perspective is not the Copilot tool itself — it is the deliberate attempt to build the human conditions that make hybrid work possible.

The project was structured around the 4P framework — Projects, Platforms, Policies, and People — with People placed explicitly at the center (Sherson et al., 2025). The framework provided a shared language for assessing where the bank stood across all four dimensions, and made visible the difference between technological readiness and organizational readiness. Twenty-four pilot users from across the bank — advisors, staff functions, and management — were invited into a workshop process rather than a rollout process. The distinction matters: workshops ask employees to develop ideas, identify use cases, and reflect on where AI helps and where it doesn’t. A rollout simply asks them to comply.

Before and after the workshops, Jens Thomas Møller Persen designed and administered the HI Willingness Index — a structured survey measuring AI self-efficacy, psychological safety, growth mindset, and sense of partnership. Cluster analysis of the before-survey revealed three distinct employee profiles that would shape the entire intervention logic. AI ambassadors — already actively using AI and experiencing concrete benefits — made up 33% of respondents. The awaiting, positive in attitude but not yet acting, represented 44% and held the greatest untapped potential. Skeptics, characterized by growing concern rather than growing confidence, accounted for 22%.

The results after two months of follow-up showed meaningful movement. AI growth — covering knowledge, problem-solving, innovation capacity, learning opportunities, and professional development — rose from 5.6 to 6.6, a 22% increase. Growth mindset showed the largest single movement of any dimension, climbing from 5.4 to 6.5. Sense of partnership improved from 2.8 to 3.5, a 35% increase, with nearly all respondents improving on this dimension regardless of which cluster they belonged to. Overall, 39% of employees improved across all four dimensions after the intervention, and the share of employees without a clear stance on AI fell by 63% — a signal that the workshops succeeded in moving people from passive indifference toward active orientation.

Psychological safety, however, showed a slight decline from 4.4 to 4.3. This is not necessarily a failure signal. The business report interprets it as a known paradox in early AI adoption: more knowledge makes expectations more concrete, and for some employees, that makes visible what they still cannot do. Awareness of what can go wrong is a precondition for using AI responsibly — but it requires active leadership support to prevent it from becoming defensive.

The data also revealed a structural tension that no training program can resolve alone. Employees’ expectation of AI’s future significance scored 7.1 out of 10 after the intervention — yet their daily experience of AI actually helping them remained well below 5.0. This expectation-experience gap is not yet critical, but it has a shelf life. High expectations that go unfulfilled over time typically produce skepticism toward future AI initiatives. Closing it is the bank’s most important strategic task in the next phase.

The project also surfaced a structural barrier of a different kind: Djurslands Bank’s standard procedures — the very content that would make Copilot most useful — are stored in a proprietary system that AI tools cannot access. Closing that gap is a platform decision, not a people decision. The 4P framework (Sherson et al., 2025) made this visible in a way that isolated tool training would not have.

The pilot established a baseline for HI readiness in a regional banking context — something that did not previously exist. The HI Willingness Index, the workshop design, and the before-and-after methodology are assets the bank can reuse as it scales adoption beyond the initial pilot group.

The business report points to a self-reinforcing growth dynamic if the conditions are right: targeted competence development leads to increased usage, which produces experienced benefit, which drives motivation, which fuels further development. The data supports this logic — employees who perceive themselves as competent AI users are the same employees who report concrete operational benefits, and this correlation strengthened after the intervention.

Three principles emerge from the case that travel beyond banking. First, employees must co-create solutions rather than merely receive them — involvement is not a courtesy, it is the mechanism by which trust is built. Second, periodic measurement of employee sentiment toward AI is itself an organizational practice worth institutionalizing. Third, a coherent HI vision and narrative at the leadership level is what gives individual experiments collective meaning.

More broadly, the case illustrates that HI maturity in regulated industries cannot be achieved through technology deployment alone. Psychological safety, growth mindset, and a genuine sense of organizational partnership are not soft prerequisites — they are the structural conditions that determine whether human judgment remains active and valuable in AI-assisted work, or whether employees drift toward passive delegation. The 4P framework (Sherson et al., 2025) offers organizations a practical lens for diagnosing exactly where those conditions are missing and what to do about it.

For the sector, Djurslands Bank offers a model: start with the people dimension, measure it seriously, and treat the gap between current and desired HI capacity as a design problem rather than a training problem.