For years, large companies have treated transformation as a program: a centralized office, a single roadmap, multiple parallel streams, and regular reporting to the top. This model creates structure, but it also introduces limits of its own. Decisions that should be made quickly and close to the ground wait for approval. Teams are pulled into several initiatives at once. Transformation offices may understand the methodology, but not always the day-to-day reality of a specific business unit.
An alternative approach treats change management not as a temporary project, but as an organizational capability. In this model, senior leadership defines long-term strategy and priorities, while concrete plans are shaped closer to the teams that will execute them. Middle managers are not simply implementing a plan handed down from above. They become owners of change in their own areas.
As Dmitry Papusha, Strategy & Operations Director and co-founder of consulting company STADIK, explains, “When changes — including radical ones such as the spread of AI — happen more frequently and require faster responses, this approach is no longer optional. It becomes a necessity.”
The expert behind the argument
Dmitry Papusha is a senior strategy and operations leader with 15 years of experience improving execution in complex environments. His work spans large-scale transformations and the development of capabilities that allow companies to sustain change after consultants leave. He has led enterprise programs under the McKinsey umbrella across industries including banking, gaming, internet businesses, and FMCG.
One of the defining projects of Papusha’s career was the transformation of Sberbank, the largest bank in Russia, Central and Eastern Europe. Dmitry co-led the operating model redesign for the Daily Banking department, a mission-critical mobile platform with around 85 million monthly active users. His work included rebuilding the execution engine end to end and strengthening internal change capacity by scaling the agile coach function, as well as upgrading onboarding.
The broader transformation program affected more than 40,000 employees and became one of the first initiatives of its kind in the market. There was no ready-made playbook. The methodology had to be designed from scratch while being tested in real organizational conditions. That created a specific kind of pressure. “The cost of experimentation becomes very high,” Papusha says. “Every mistake or inconsistency feeds resistance.” The program had to move forward without the luxury of a long experimental cycle.
For Dmitry Papusha, the most important factor was not the methodology itself, but the involvement of the bank’s leadership, who remained active participants throughout the process. The program successfully moved from pilots to full-scale adoption. New approaches to organizing production teams became part of the bank’s operating rhythm, while engineering culture took root in key departments. This created a foundation for further digital initiatives and helped support Sberbank’s later development as a technology ecosystem.
Dmitry Papusha encountered a different transformation challenge at Playrix, now one of the world’s top mobile gaming companies. At the time, it was already generating around $1.7 billion in annual revenue and preparing for further operational scale. “This was not about fixing something broken,” Papusha explains. “It was about finding bottlenecks in processes that were already working well, but could become constraints under new scale.”
That required a different approach from large transformation programs. Instead of creating a central office and a visible company-wide initiative, small cross-functional teams of three to five people were formed around specific problems. These teams had relevant expertise, broad authority, and autonomy. They worked in short iterations, tested hypotheses, and involved key leaders only when a solution was ready for final alignment. The work happened locally and in parallel, without a loud organizational announcement.
In practice, this led to changes in traffic acquisition planning, creative production prioritization, and onboarding programs. For example, Papusha orchestrated the implementation of collaboration and prioritization models that improved marketing forecast accuracy by 10% and reduced campaign production cycles from four weeks to 1.5 weeks, respectively. For Dmitry, this showed that when changes are targeted and the organization is mature enough, small autonomous teams can move faster and avoid unnecessary coordination overhead.
Another important project was the transformation of one of Southeast Asia’s largest FMCG conglomerates: a company with more than 70 years of history, over 13,000 employees, and brands known across the region. “The task was complex because product-oriented approaches had to be applied in an environment very different from digital development,” Papusha says. “Many processes had been refined for decades. Feedback cycles were measured not in days, but in months. An error was not a software bug that could be patched later, but a defect in a physical batch of products.”
The pilot included six teams from very different functions, including new product development, marketing, and logistics. Their work covered both online interfaces and physical goods — from improving customer-facing processes to experimenting with the amount of cookie crumb in a chocolate bar to create the right sensation of crunch for consumers.
The initiative later scaled to thousands of employees and led to a 40% improvement in cross-department alignment, shorter delivery cycles in key product lines, and greater predictability in quarterly commitments. One critical condition remained: capabilities had to accumulate inside the company rather than stay with consultants. For Papusha, the project reinforced a broader lesson — transformation approaches cannot be copied mechanically from one industry to another. They have to be adapted every time to the culture, history, and real constraints of a specific organization.
From transformation experience to repeatable frameworks
Over time, Dmitry Papusha turned this experience into several repeatable frameworks that he sees as especially relevant for companies facing the next wave of transformation. The first is an approach to building internal centers of expertise as a way to internalize external knowledge. It includes defining the right talent profile and hiring system, creating internal training programs and academies, building mentoring structures, and designing long-term onboarding and development plans for new people.
The underlying principle is that any capability that becomes strategically important can be grown inside the organization if the company treats it as a system rather than a hiring problem. “In the AI era, this becomes even more important: the necessary skills are scarce, and they quickly become outdated,” Papusha says. “The ability to accumulate, reproduce, and renew expertise internally can become a long-term competitive advantage.”
The second framework is designed for piloting new and especially risky initiatives. Papusha used it to test and refine operating models before scaling them. But its value goes beyond validation. A well-designed pilot solves three problems at once: it tests the methodology in real conditions, builds middle management autonomy in leading change, and strengthens the organization’s internal muscle for implementing, measuring, and controlling transformation.
“This last point is particularly important today,” Papusha emphasizes. “As changes become more frequent, companies cannot depend on external teams every time they need to adapt. They need to learn how to run change themselves — with enough structure to manage risk, but enough autonomy to avoid turning every initiative into another centralized program.”
These frameworks reflect a broader principle in Dmitry Papusha’s work: change capability has to be built into the organization rather than attached to it from the outside. For this to happen, three conditions should be met.
- Change management has to become part of the managerial profile. A manager should be able to diagnose resistance, explain the meaning of change, and keep a team productive in uncertainty.
- Managers need to cascade strategy downward. This means translating not only financial goals, but also operational priorities and metrics to every level of the organization.
- Companies should be prepared to support managers in making more decisions at their own level. Without this support, the first two capabilities do not fully work. A manager may understand the strategy and know how to guide a team through uncertainty, but if every decision waits for approval, change still moves at the speed of the hierarchy.
AI raising the stakes
All three criteria become even more important in the AI era, as the speed and frequency of change continue to grow. In many companies, today’s AI use is still advanced automation: agents take over part of an employee’s tasks and accelerate processes. This already creates value. But the next level — an organization where agents make decisions autonomously, interact with each other, and respond to their environment in real time — creates a different set of constraints and requires a new operating model.
Older models were designed around another logic: a human makes a decision, a hierarchy approves it, and a system executes it. “An AI agent works differently,” Papusha says. “It does not wait for approval — it acts.” That is where the conflict appears. Hierarchical chains absorb speed like friction. Blurred responsibility becomes a critical risk. Fragmented data makes autonomous decisions unreliable.
In this context, companies need to redesign operating models through the logic of multi-agent interaction. The structures best suited to AI agents differ not mainly by technology, but by how they are organized internally. The first characteristic is flat decision chains with clear rights. An agent needs to know what it can decide independently, what it must escalate, and under what conditions it must stop. Organizations that define these boundaries in advance gain an advantage because speed is not achieved by removing control, but by making control explicit.
The second characteristic is data as operational infrastructure. An agent can only make decisions as well as the information available to it. Companies ready for agentic AI treat data as a product, with owners, quality standards, and a single access point. Data cannot remain a byproduct of disconnected systems if autonomous decisions depend on it.
The third element is a redefined human role. Employees move from task execution to orchestration of outcomes. “People set goals and define constraints while remaining final arbiters — but they should not become the bottleneck,” Dmitry Papusha says. This also forces organizations to answer a practical question before it becomes a crisis: who is accountable when an agent makes the wrong call? Companies that fail to define this upfront will face the issue at the worst possible moment.
Ultimately, AI raises the standard for management. As companies move from isolated automation toward more autonomous agentic systems, transformation can no longer depend on temporary programs, central offices, or methodology alone. The organizations best prepared for the AI era will be those that can distribute ownership, clarify decision rights, build reliable data foundations, and allow people to supervise outcomes rather than slow them down.
