How to Achieve a Positive Future with AI: Four Dimensions in Balance

Within a few months, two very different voices have shifted the debate on artificial intelligence. In May 2026, Pope Leo XIV presented Magnifica Humanitas, the first encyclical explicitly devoted to protecting the human person in the age of artificial intelligence. At the end of August, Bill Gates followed with a long essay on a turbulent AI era and the choices that now have to be made.

Neither text is a rejection of technology. Both say something different, and yet the same thing: what determines the benefit is not the pace but the direction.

I share this assessment, and I hold it from a very practical perspective. We build software for corporations, banks, energy utilities, and public institutions with agentic platforms. Every day I see what works. And I also see where organizations harm themselves because they confuse automation with value creation.

What agentic development actually changes

Agentic development is not the next stage of code completion. It changes the entire chain: strategy, analysis, specification, build, test, operations. People define the goal, review, and take responsibility. Agents take over the execution.

For large organizations, the greatest leverage lies where nobody could reach before. Legacy applications hold decades of domain knowledge that was never documented and whose authors have long since retired. Precisely this knowledge can now be reconstructed from custom code, transferred into a robust specification, and then rebuilt. Modernizations that used to be considered unaffordable become plannable.

The orders of magnitude are considerable. We see acceleration of up to fivefold, a quality improvement of around 30 percent, and massive relief in the pure coding share. Across projects, a corridor of 20 to 45 percent efficiency gain is evidenced, depending on the starting position, data quality, and maturity of the organization. Anyone quoting higher figures without context is selling expectations instead of results.

For the public sector, the same applies with a different sign. Public administration is rule-based, procedural, and subject to documentation duties. That is not a weakness but the ideal precondition. In Der selbstfahrende Staat we described how digital twins of legislation make legal norms machine-readable, so that procedures become faster, more traceable, and more comprehensible for citizens and businesses. Not less rule of law, but more of it, because decisions become explainable.

Governance is not a brake

The most frequent objection is that control costs speed. The opposite is true. Unchecked AI generation demonstrably produces security vulnerabilities on a considerable scale. What is created without review gates is paid for twice in operations.

That is why we work with human validation at defined points, with security and quality gates, with automated acceptance tests, and with expert review. That is not a brake. It is the condition under which a board may take the results into production at all.

For public administrations, I formulated the principle in the DNA of the self-driving state: even when software automates procedures, responsibility remains with people, and every decision must be documented in an audit-proof manner. Anyone who cannot guarantee that should not automate.

Autonomy needs three counterweights

In my books I set three dimensions against autonomy as a counterbalance: sustainability, resilience, and humanity.

Autonomy is the engine. It describes how far a system acts without human intervention, and it generates the leap in productivity.

Sustainability asks what operations cost. Data centers consume energy and water. The Pope, too, points to this ecological price in Magnifica Humanitas. An efficiency calculation that leaves out resource consumption is incomplete.

Resilience asks what happens when the system fails, is manipulated, or is wrong. A corporation that has lost its ability to fall back to manual operation is not efficient but vulnerable. For critical infrastructure, that is an existential question.

Humanity asks whether the people in and around the organization live better as a result. It is the most demanding of the four dimensions because it is hard to capture in key figures.

The decisive point: no company can balance these four dimensions on its own. A single provider that voluntarily slows down loses. It takes rules that apply to everyone. That is precisely the task of policy makers. Their role is not to slow autonomy down but to keep the four dimensions in harmony.

What the encyclical contributes

Leo XIV formulates an insight that should sound familiar to everyone responsible for technology: technology is not an adversary of humanity per se, but it is never neutral either. It bears the traits of those who design, finance, regulate, and use it.

I consider three points directly applicable to corporate practice.

First, the dignity of work. The encyclical states that new forms of work are not automatically better ones, and criticizes systems that force people to adapt to the rhythm of the machine rather than the other way round. That is a design criterion, not a pious wish.

Second, subsidiarity. Leo XIV observes that in the digital space the top level has long ceased to be the state; it is the large platforms that define conditions of access, visibility, and economic opportunity. Anyone who wants sovereignty for their organization must actively shape this dependency.

Third, the fundamental question. According to the encyclical, the real alternative is not between enthusiasm and fear but between two paths of development: progress that serves people and nations, and progress that subjects them to a logic of power. That is a leadership decision, and it is taken in corporations just as in governments.

What Gates contributes

Gates argues economically and arrives at a related point. He proposes deliberately reserving certain activities for humans and calls this Human Reserved. His image is the nature reserve: one could build, but decides against it because the loss would be too great. He has two forms in mind, permanently human activities and temporary protection for employees who realistically can no longer retrain. In the most far-reaching variant, he speaks of an order of magnitude of up to around 40 percent of jobs.

Added to this is a tax proposal. Gates argues that today's system pushes companies toward the machine because staff triggers payroll costs while equipment is depreciated. A tax on computing power and robots is meant to correct this imbalance.

I consider the diagnosis accurate and the proposal worth discussing, not finished. Arguments against a token tax include problems of delimitation, international evasion, and the risk that European users of all people are burdened while value creation takes place elsewhere. But the debate must be held, and now, not only once displacement has taken place. That framework conditions must be set so that this transformation proceeds safely and for the benefit of people is, for me, not a concession but a prerequisite.

Human Reserved does not begin in parliament but in the company

The part that companies should not leave to the legislator is the business part.

We already see the disadvantages today in organizations that have automated too far. Processes run faster, relationships erode. Customers no longer reach anyone who is allowed to decide. Applicants drop out because they never meet a human being in the process. Suppliers lose their willingness to help in a crisis because there is nobody they would be helping. These costs appear in no business case, but they are incurred.

People have a deliberate place in value creation, namely wherever people interact with people. In B2B business this is particularly clear:

• Sales. A multi-year framework agreement is not signed because of a product page but because someone trusts a person who will stand up for it when in doubt.
• Recruiting. Anyone who wants to win people has to meet them. A fully automated process mainly selects those who can operate automated systems.
• Leadership. Responsibility, feedback, and bad news cannot be delegated, least of all to a system.
• Procurement and negotiation. The value of a supplier relationship shows in the exceptional situation, not in normal operations.
• Conflict, complaint, crisis. Anyone who automates empathy here saves at the most expensive point of the customer relationship.

My rule for this is simple: automate the preparation, not the relationship. A sales team whose analysis, proposal drafting, and documentation are agentically supported gains time for exactly what only people can do. That is not a renunciation of efficiency. That is its meaningful use.

What needs to be done now

For boards and heads of administration:

1. Determine your maturity level honestly. Digital, automated, or self-driving. Without a position assessment, every roadmap is fiction.
2. Start with what exists. Reverse engineering the legacy landscape delivers value faster than any new frontend.
3. Define review gates before you buy speed. Human validation, security review, acceptance test.
4. Put your own Human Reserved fields in writing. Which points of contact remain human, regardless of what becomes technically possible.
5. Keep the fallback level. Resilience is the part of the calculation that only becomes visible when it is too late.

For policy makers:

1. Keep the four dimensions in balance. Autonomy without sustainability, resilience, and humanity produces short-term gains and long-term costs.
2. Regulate traceability rather than technology. Anyone who automates procedures must be able to explain and document decisions.
3. Hold the debate on Human Reserved and on the tax imbalance now, openly, and together with business and the social partners.
4. Prevent concentration. When models, data, and infrastructure lie with a few actors, that is an economic policy risk and not merely an ethical one.

In closing

I am an optimist, and from experience. Agentic development makes modernizations possible that failed on budget and capacity for ten years. It frees skilled professionals from work that nobody misses. It can bring public administration back into a relationship with citizens that deserves the name service.

But autonomy is not an end in itself. It is one of four dimensions. Companies that understand this will be faster and at the same time preserve what constitutes their value. Societies that understand it will shape this transformation instead of suffering it.

The question is not whether we develop agentically. The question is what for.

Sources and further reading

• Pope Leo XIV, Magnifica Humanitas. On the Protection of the Human Person in the Age of Artificial Intelligence, encyclical, signed May 15, 2026, published May 25, 2026. vatican.va
• Bill Gates, A Turbulent AI Era and Critical Choices to Make, GatesNotes, August 2026. gatesnotes.com
• Florian Schnitzhofer, The Self-Driving Company. A Conceptual Model for Organizations of the Future, Springer, 2023. About the book
• Florian Schnitzhofer, Patrick Pils, Philipp Seper-Ambros, Der selbstfahrende Staat. Ein Denkmodell für das Zusammenleben im Staat der Zukunft, Springer Gabler, 2024; English edition The Self-Driving State, Springer, 2026. About the book