Magnifica Humanitas, Pope Leo XIV’s encyclical published just a few weeks ago, is a long engagement with what digitalization, robotics, and artificial intelligence are doing to human institutions. One observation in it caught me when I read it and has stayed with me since.

Coincidently, Julie Whitten’s The People Stack came out at almost the same time. The book is a study of why organizational change fails despite formal success and offers a perspective on transformation grounded in the human architecture underneath the visible work. The two pieces, although radically different in nature, talk about something that resonates deeply with my own work in organizational change and transformation leadership.

Both readings name one thing I have been thinking about for years without finding language sharp enough to hold it. The encyclical names the pattern at the level of nations, where authority has migrated from States to major economic and technological actors whose resources outstrip most governments and whose design choices set the rules everyone else has to obey. The book names it at the level of the firm, where another invisible architecture beneath the visible work decides whether anything actually moves. Different architectures. Different scales. The same structural intuition: the visible layer of organizational action is not, increasingly, the layer that does the actual deciding.

What follows is the reflection that came of them. The argument that develops returns to the form I usually work in.

What the org chart still claims

The conventional picture of authority inside a large enterprise is the cascade picture. The executive committee sits at the top, with function heads and business unit leaders working beneath them, regions and country offices below those, and line managers running the day-to-day work of teams whose individual contributors do the producing. Authority moves downward through decision rights, delegation matrices, approval thresholds, and policy. The picture is clean enough to fit on a single slide, which is partly why it persists.

Most transformation programs treat the cascade picture as the operating reality of the firm. They design new operating models against it. They cascade change communications along it. They map sponsorship to it. The methodology presumes that the cascade picture is roughly true and that the work is to adjust its shape, sharpen its accountabilities, or reallocate decision rights within it.

The picture was once true, or true enough. It described a world in which decisions lived in policy documents and in the judgment of the people exercising delegated authority. The deciders were people. Decisions happened in deliberation, slowly, with traceable rationale. The cascade story held because the architecture of decision-making was, in practice, a network of human authorities working through documents and meetings and signatures.

Those conditions are no longer dominant. Inside most large firms today, a second architecture sits underneath the cascade picture. It does not appear on the org chart. It does not show up in the operating model. It does not feature in the change-management plan. But it has absorbed, quietly and over years, a meaningful share of the authority the cascade picture still claims.

The cascade picture has become, increasingly, a description of how the firm narrates its decisions rather than how the firm makes them. Programs that work above the second architecture without naming it are operating where authority no longer lives. The first move toward seriousness is to acknowledge the second architecture exists.

What the platforms decided in advance

The second architecture is the platform layer. ERP, CRM, HRIS, workflow systems, communication tools, the analytics layer, and increasingly the AI assistant that mediates between any of the above and the human supposed to use them. Most large firms run dozens of these systems.

Each platform encodes decisions. Which fields are required. What gets measured and how it gets aggregated. Who can see which records. What exceptions are visible to which roles. Where approvals route and at what thresholds. What an out-of-policy condition looks like, and what the system does about it. What behaviors generate signal and what behaviors disappear into a log no one reads. None of these are features. All of them are decisions about how the firm operates, made on the firm’s behalf and embedded in the architecture the firm runs on every day.

The decisions were not made by the people the org chart names as deciders. Vendors shipped the product with defaults. Implementation teams selected from configuration menus during the rollout, often working under deadline pressure and tradeoffs against budget. Project managers deferred edge cases, leaving the system’s first-draft behavior to handle situations no one had thought about carefully. Outside consultants made choices on the firm’s behalf with its tacit endorsement. People who left the firm years ago made decisions that still shape how customers are categorized, how performance is scored, and what conversations happen inside which meetings. The encoded decisions outlast every executive in the room, and they continue to govern the firm long after the implementation team has moved on.

This is not a critique of platforms. They are doing exactly what the firm purchased them to do, which is to encode decisions in software so that the firm does not have to make them by hand every time. That is a real efficiency, and the firm benefits from it. The argument is that the encoded decisions are decisions, that the firm has been carrying them for years, and that almost no one inside the firm currently treats them as decisions worth governing.

Most operating-model conversations skip past this layer entirely. They assume the platforms execute decisions the executives have made, rather than recognizing that the platforms have, in many cases, already made the decisions the executives are about to ratify.

The fiction layer of transformation

The consequence of the platform layer’s authority is that the apparent decisions of the org chart are, in many cases, downstream of decisions the platform has already made. The executive who decides a policy is selecting from options the platform supports. The middle manager who approves a request is operating inside a workflow whose escalation logic the implementation team set during configuration years ago, and the manager’s approval is a button click on a screen the system rendered using rules the manager did not write. The front-line worker who exercises judgment is doing so inside a window the system carved out for human attention, with upstream automation having already filtered the larger context and decided what counted as the relevant facts.

In a routine quarterly business review, the analytics dashboard has pre-selected which variances appear at the top of the discussion, in what visual form, with which suggested root-cause categories already tagged against them. The workflow tool has pre-populated the corrective actions available for selection from a list whose menu the implementation team built three system releases ago. The executive arrives prepared to decide. The decision was assembled before the executive entered the room. The meeting is the moment at which the firm ratifies and narrates the decision the platform layer has already proposed. This is the texture of contemporary executive decision-making across most large firms.

The implication for transformation work is structural. Most transformations operate at the layer above this one. They redesign processes that sit on top of platforms whose decision-encoding goes untouched. They redraw operating models the platform layer will not, in fact, support, then commission expensive workarounds when the misfit becomes visible during pilot. They cascade authority into roles whose authority the system has already pre-allocated, and the role holders discover that the new accountability the program assigned them comes without the decision rights the role description implies. The transformation hits its targets on paper while the operational reality drifts in a different direction.

The result is the pattern most senior leaders have lived through more than once. The program hits its case, the leadership team declares success at the milestone reviews, the firm announces and rolls out the new operating model. Inside two years the gap between the declared model and the operational reality is wider than it was before the program started. The decisions the program intended to relocate have stayed where the platform layer placed them years ago. The transformation succeeded at moving the visible scaffolding without moving any of the structure underneath.

The structural diagnosis is that most transformation work is occurring above the layer at which authority actually lives. Programs that do not name and engage the platform layer as a decider are running interventions in a space where decisions no longer happen. The energy goes into adjusting the cascade picture while the substance of the firm continues to operate from the architecture that the cascade picture no longer accurately describes.

Decision architecture as governance

The discipline that does the actual work is decision-architecture analysis. It treats the platform layer as a co-author of every decision the firm makes. The platform layer participates in those decisions, sometimes as the dominant participant. Governance that overlooks this participation is incomplete by design.

The discipline has four moves. The first is to map where decisions actually happen: which decisions the system makes alone, which decisions the system makes and the layer of humans with screen access ratifies, which decisions humans still make in deliberation. Most firms have never produced this map. When they do, the distribution surprises them.

The second is to identify the encoded decisions explicitly, as decisions. This requires walking through the platform configurations, the dashboard logic, the workflow rules, and the prompts and guardrails of the automated assistants now embedded in those tools, and asking, at each one, what choice did someone make here, and on what basis. The exercise is more uncomfortable than it sounds because the answer is often that someone made the choice under deadline pressure during the original implementation, on a team that has since dispersed, with reasoning no one can now reconstruct.

The third move is to ask, for each encoded decision, whether it is the decision the firm would have made if it had been asked. The question separates the encoded decisions that still fit the firm’s intent from the ones that no longer do. Both categories will turn up. Some of the legacy encoding will be correct. Some of it will not. The firm has been treating both as equally settled.

The fourth move is to choose which encoded decisions to move closer to the work and which to leave encoded where they are. Moving them closer means redesigning the screens, the workflows, the routing logic, and the platform configurations so that the decision rights the firm intends are the decision rights the platform supports. Leaving them encoded means accepting that the firm has chosen, deliberately, to let the platform decide that question on its behalf, and accepting the consequences of that delegation.

This is governance work, conducted at the layer that now does the deciding. It touches platform configurations, vendor relationships, and technology strategy without being any of them.

The most useful exercise an executive can run this year is to walk through their last major decision and ask honestly how much of it the surrounding architecture had already settled before the discussion in the room began. The answer will not flatter anyone. It will name the platforms that pre-shaped the options, the dashboards that pre-selected the variances, the workflows that pre-populated the corrective actions, the consulting choices made years ago that still constrain the menu the executive sees today. It will also reveal which fraction of the decision the executive actually made, in the room, with judgment that could have gone differently.

That fraction is the floor of recovered authority. Any transformation that intends to actually move authority where it claims to be moving it has to grow that fraction first. Programs that work above the platform layer cannot do this. Governance that treats it as a co-author can.

References

Pope Leo XIV. (2026, May 15). Magnifica humanitas: On safeguarding the human person in the time of artificial intelligence [Encyclical letter]. Dicastery for Communication – Libreria Editrice Vaticana. https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html

Whitten, J. A. (2026). The people stack: Why change fails (and what actually fixes it). Julie Whitten Consulting.