AI-Transformation Playbook

Most companies have now purchased, rolled out, and maintained AI tools for months. The results are sobering: after a year, on average only 10–15% of employees actually use the tools productively, and only 3% are classified as “proficient” or “expert.” 85% simply use AI as a better search engine.

This is shown by the Section AI Proficiency Report, based on data from over 150 AI transformation projects worldwide. The numbers are clear, and the real problem runs deeper.

The fundamental problem: AI is treated like a software rollout

In a presentation during the “Human X” session, Greg Shove, CEO of Section AI—an AI consulting firm that has already guided over 150 organizations worldwide through AI transformations—summed it up:

"The root of all challenges in AI transformations inside large organizations is that they think about it as a software deployment. It is not. This is not an extension of Office or an upgrade to an ERP or a CRM. This is co-intelligence. It works weirdly and wonderfully, and it threatens people's livelihoods."

AI is not an SAP rollout. Those who treat it as such may be implementing tools, but they won’t achieve a change in behavior. The reason pilot projects fail is almost always the same: most AI initiatives are imposed from above without a genuine understanding of the workforce and their day-to-day work. Executives define use cases at the board level, while employees—the actual experts in their processes—are not involved. The result is solutions that are out of touch with reality and, consequently, are not adopted.

What Successful Organizations Do Differently

The key metrics for measurable added value are clear:

• 80% of the workforce should have advanced proficiency in AI,
• 80% should actively use the tools on a weekly basis for substantive tasks, 
• 5% should serve as internal advocates,
• and the ratio of agents to employees should be one-to-one in the first year and three-to-one in the second year.

The Right Order: Augment → Automate → Reinvent

Many programs start with moonshot projects. That’s the wrong approach. The order matters:

Phase 1 – Optimize (Augment the Workforce): Employees learn to use AI daily for their own tasks. Low risk, enormous potential.

Phase 2 – Acceleration (Workflow Automation): Recurring tasks and processes are systematically automated.

Phase 3 – Reinvent (Rethinking the Entire Business): Only now are entire business processes reimagined—for example, in customer service or contract management.

"Let's start with the workforce, and let's augment them as quickly as we can."

Anyone who starts Phase 2 or 3 without having completed Phase 1 is building on sand.
 

Not a policy paper, but an AI manifesto

A common mistake: Companies start with lengthy governance documents and lists of restrictions.

An effective AI manifesto answers three questions: Why are we introducing AI? How do we intend to use it? What standards apply to day-to-day operations? Section AI, a company that is itself among the most advanced AI users, has defined five operational principles:

• Use AI daily and deliberately
• Hunt for obsolete work
• Create, don’t just cut
• Be bold but not careless
• Share breakthroughs and failures

Key internal message:

"AI is not cheating."

The manifesto should be no more than one page long, endorsed by management, and updated quarterly.

Managers Are the Bottleneck—Not the Tools

Transformation rarely fails because of technology. It fails because of middle management.

Greg Shove describes what managers need to do:

“We need super managers to build super companies. They need to be empowered and able to really manage this change [...] We need to get [Leaders] on their own learning curve, actually using these tools and capabilities and understanding [them] better.”

The counterexample mentioned during the session: A manager who, for four years, walked through the hallways every day, sat down with employees, and led “Lunch & Learn” sessions.

No manager should talk about AI goals unless they have at least ten hours of hands-on experience themselves.

The best agents don’t come from the central IT team

An often underestimated lever in AI transformation: The most effective automations don’t originate in a central department or a specialized project team, but directly from the employees who know their own processes best. Those who delegate the development of AI agents exclusively to experts are squandering precisely this knowledge.

The approach to dealing with failures should be pragmatic:

„Build them fast, build them cheaply. Some will stick and be used. Some will just fade away. It doesn't matter. They didn't cost a lot to build."

The 50-50 rule applies: 

• 50 percent of experiments deliver significant value
• 50 percent deliver none. This is normal and a healthy sign that an organization is truly experimenting.

Where will the “blast radius” strike first?

Every company must honestly assess how close it is to the epicenter of AI disruption. The sectors most severely affected are software development and knowledge-intensive services such as legal consulting, management consulting, and auditing. For now, the sectors least affected are those that rely on a physical presence: construction, energy, and healthcare.

For board members and executives, this means conducting an honest strategic assessment: Where are we vulnerable? And where do the greatest opportunities lie? The key question here is:

„There are two strategies in AI: What are you going to cut? And what are you going to create? Super companies do both at the same time."

What to Do Now

In 30 days

Within the first 30 days, a clear AI manifesto should be adopted—a document endorsed by senior management and communicated to everyone. At the same time, access to powerful AI tools should be made available to all knowledge workers, without distinctions based on hierarchy or department. And there needs to be a person in charge of the AI transformation who has a business background, not just a technical one.

In 90 days

During the first 90 days, the goal is to assess the current state of AI usage, recruit 5% of the workforce as internal advocates, and hold an initial hackathon in which all employees develop at least one AI agent of their own.

Within 12 Months

The goal after twelve months: More than 60 percent of the workforce uses AI weekly for substantial tasks, at least 15 percent have reached an advanced level of proficiency, and a ratio of three agents per employee is within reach.

Either you build a Super Company—or you’ll soon be competing against one.

Lust auf mehr?

Wir unterstützen deutsche Unternehmen auf ihrer Tranformationsreise. Wenn du dir Insights & Support aus dem KI-Innovationsökosystem Nr.1 - dem Silicon Valley - holen möchtest, melde dich bei uns.