.png)
A summary based on the 2025 research of PwC Switzerland–IMD, and McKinsey & Company
AI development over the past decade has been gradual, but after 2023, its pace brought a qualitative leap. From generative models — which produced text and images — we have arrived at the era of autonomous AI agents: systems that independently perceive, decide, act, and learn.
The deployment of AI agents is substantially reshaping organizational structures, governance models, and decision-making processes. Artificial intelligence is delivering the most significant organizational paradigm shift since the industrial and digital revolutions — this time affecting knowledge work.
The value of AI investment is now determined by whether an organization can deploy it in a strategically directed, measurable way, or whether it remains permanently stuck in the experimental phase.
This summary draws on two major sources: the joint study by PwC Switzerland and IMD (One Agent to Rule Them All, 2025) and McKinsey's research (The Agentic Organization, 2025).
Before deploying AI agents, it is worth examining what they are being built upon. PwC's Enterprise Data Strategy (EDS) framework assesses an organization's data strategy across five dimensions:
This framework retains its relevance in the age of agent-based AI — though the emphasis shifts (PwC–IMD, 2025).
With agents that make decisions autonomously, poor data quality can trigger flawed decision chains that propagate through the organization without human detection. Deliberately curated proprietary data assets — from customer behavioral data to product usage patterns — become one of the most critical competitive advantages in the agentic era (McKinsey, 2025).
AI agent deployment can be realized through two complementary approaches:
A centralized, dedicated team of specialists develops specific, end-to-end use cases. A single agent captures significant value, with a longer deployment timeline.
A good example is the Vendor Master Data Management Agent (PwC–IMD, 2025). Vendor master data management previously involved monotonous, manual work: copying data from documents, monitoring for duplicates, with data ending up, at worst, in a spreadsheet on a shared drive. The agent took over this entire process: it extracts data from uploaded documents, flags any discrepancies it detects, and can even locate the right vendor based on a plain-text description.
Key characteristics of the enterprise-led model:
Employees create their own personalized agents, with rapid deployment and smaller but cumulatively significant value.
A strong example is the Financial Commentary Agent (PwC–IMD, 2025). A financial analyst noticed that consolidating quarterly reports consumed a disproportionate amount of time — even though the work essentially amounted to data collection and assembly. The agent guides the analyst through selecting the relevant period, extracts data from subsidiary reports, and produces a finished consolidated report. The analyst can then focus on what truly requires understanding: the insights behind the numbers.
Key characteristics of the user-led model:
The two approaches complement each other well. Those who build from the bottom up quickly discover that a well-functioning solution is worth deploying more broadly. Large-scale, centrally developed solutions, in turn, can later branch into more specific use cases. The goal is to establish an internal "agent marketplace" where agents can be published, adopted, and further developed (PwC–IMD, 2025).
AI agents structurally decouple the relationship between growth and cost: as processes transform, marginal cost can decline toward the cost of compute capacity. Proprietary internal data assets become the most critical competitive advantage — organizations that continuously expand and integrate their own data into agent-based systems will consistently outperform those that rely exclusively on publicly available models. Companies that own the customer relationship can create value well beyond their traditional industry boundaries (McKinsey, 2025).
The fundamental organizational unit changes: functional hierarchies give way to small, outcome-focused, mixed human–machine teams. A human team of two to five people can already oversee a system of 50 to 100 specialized AI agents covering an entire business process end-to-end. Organizational charts are replaced by task-based networks that extend beyond organizational boundaries to include external partners (McKinsey, 2025).
Traditional planning and control cycles are too slow for the operational tempo of AI agents. The role of leaders shifts from overseeing execution to defining policies, managing exceptions, and providing strategic direction. Cybersecurity represents the highest risk factor across all corporate functions: agents' autonomous access to sensitive data and systems opens new attack surfaces that can be addressed through deliberate architectural safeguards (PwC Switzerland–IMD, 2025).
AI agents take over a significant portion of the routine tasks performed by knowledge workers, while end-to-end problem-solving, systems thinking, and exception handling that requires human judgment become increasingly valued. Human judgment remains irreplaceable where situational awareness is uncertain, data is incomplete, or ethical considerations cannot be automated. A new executive role is also taking shape: the Chief Agent Officer, responsible for the strategic direction of the entire enterprise agent portfolio, coordination among agents, and the enforcement of ethical guardrails (McKinsey, 2025).
The management of IT systems and data is becoming democratized: agent-based development environments enable employees without technical backgrounds to independently create software solutions and data pipelines. Inter-agent communication protocols are replacing traditional system integration solutions, enabling faster and more cost-effective connectivity. Avoiding technology and vendor lock-in becomes a strategic priority, requiring an architectural approach that separates organizational knowledge and process logic from the current technology layer (McKinsey, 2025).
1. Build organizational learning capacity through quick-win deployments.
Organizations that build their own agents today gain direct experience: they learn which models work, what data must be prepared, and where human intervention should be retained. When a ready-made market solution emerges, these organizations will be able to evaluate and configure it rapidly. Early adopters build organizational learning capacity that multiplies in the next development cycle (McKinsey, 2025).
2. Data strategy precedes agent strategy.
Without high-quality, well-governed data assets, the potential of AI agents will remain unrealized. Strengthening the five dimensions of the enterprise data strategy framework is a prerequisite — without it, agent-based deployment is built on an unstable foundation (PwC Switzerland–IMD, 2025).
3. Agent governance must feature on the senior leadership agenda.
The CEO and leadership team must establish a comprehensive position on the role of AI agents: which ones to develop in-house, which to source externally, which areas to prioritize, and what ethical guardrails to put in place (PwC Switzerland–IMD, 2025).
4. Make culture the ethical compass.
Organizational culture is a decisive factor: embedding values and long-term purpose into agent-based systems prevents short-term efficiency-seeking from undermining organizational cohesion and internal trust (McKinsey, 2025).
Agentic artificial intelligence is a tool for amplifying human work. The most successful organizations integrate AI agents into their operations in a structured, data-driven, and responsibly governed manner, while freeing their people to focus on higher-value tasks.
In the agentic era, the source of competitive advantage lies in which organization builds faster, learns continuously, and scales responsibly.
As an AI-native process automation platform, Fluenta One helps organizations transform the potential of agentic artificial intelligence into measurable, institutionalized digital transformation — from the experimental phase through to scaled, governed operations.
Contact the Fluenta One team and begin your agent-based operations today.
Sources:
.png)
.png)
.png)
.png)