Research consistently shows that AI programs collapse not because of the algorithm, but because of the human layer: teams unprepared to experiment, leaders who skip the change management work, and cultures where fear of failure is stronger than the appetite for learning.

In many enterprises across Germany, the US, and South Africa, the pattern is the same: organizations invest heavily in tools and almost nothing in people readiness. The result is expensive software that goes unused and teams that quietly disengage.

The BCG 70-20-10 framework is based on where AI programs actually succeed or fail. The technology component (10%) is the easiest to buy. The hard work is building people readiness: AI literacy at every level, psychological safety to experiment, and leaders who can manage the disruption that comes with role changes.

For a mid-size company, the stakes of getting this wrong are immediate. When an experienced employee sees AI changing her role, she does not experience it as digital progress. She experiences it as an identity threat. That requires deliberate leadership attention, not a software rollout.

The right first use case meets three criteria: high repetition (your team does it weekly in a similar way), low risk if the output is wrong (a human reviews it before it reaches a customer), and visible benefit to the person doing the work, not just to the business bottom line. Offer preparation, standard customer responses, and internal documentation are typical starting points.

The most common mistake is starting with the most strategically important process. That carries too much risk and too much pressure for a first experiment. A smaller, unglamorous use case builds the experimentation muscle the organization needs before tackling anything complex.

An experimentation culture is defined by one specific behavior: the willingness to try something, fail visibly, and share what was learned. In the context of AI, this means distinguishing between reversible decisions (a marketing draft reviewed before sending) and irreversible ones (AI used in hiring decisions). Teams need structural permission to experiment in the first category without fear of judgment.

The simplest tool is a 15-minute weekly ritual: what did we try with AI, what worked, and what did not? Without this structure, the same mistakes repeat and nothing accumulates. The leader’s role is to model this first. In any company, if the leader does not share their own AI failures openly, no one else will.

AI creates two specific anxieties that traditional change management does not address. Employees who use AI worry: “Will people think I cannot do this myself?” Employees who do not use it worry: “Will asking how it works expose me as someone falling behind?” Both reactions are invisible to most leaders and both shut down the experimentation needed for adoption to succeed.

The identity threat is sharpest for experienced employees. A finance manager with 20 years of expertise who sees AI drafting her reports in seconds is not thinking about efficiency. She is asking what her value is now. Teams where this question goes unaddressed develop quiet resistance, disengagement, and eventually turnover among the very people whose judgment is needed to oversee AI output.

Technical AI knowledge is not the critical skill. It can be delegated. The four capabilities that cannot be delegated are: radical prioritization (committing to one use case for six months and resisting the pull of every new tool); visible learning (experimenting yourself and sharing what failed); trust-building under uncertainty (answering “what does this mean for me?” honestly, including when the answer is unknown); and deciding without full evidence (moving forward transparently rather than pretending certainty).

These play out differently by market. German Mittelstand leaders often need to shift from operational to strategic thinking. US mid-market leaders tend to move fast but underinvest in the psychological infrastructure for sustained adoption. In South Africa, the opportunity is to build the right culture from the start rather than inheriting legacy organizational habits.