Agentic Transformation: what boards and senior executives need to know now

Bas Kemme
Apr 8
10 min read

First came digital transformation. Then came AI transformation. Now a third stage is emerging: agentic transformation.

The objective of this 5-minute read is to help board members and senior executives have the right conversation about what is coming next. That means understanding three broad stages: AI enablement, AI acceleration, and Agentic Transformation. It also means understanding why new forms of process architecture, human oversight, and governance are becoming essential as AI moves from supporting work to taking over execution.

Over the past few weeks of March 2026, in the aftermath of excitement around OpenClaw and the agentic developments at Antrophic's Claude, I have drawn in particular on conversations with Marco van Hurne, Olivier Rikken, and Yassine El Kochta who helped sharpen my thinking on the technical and architectural side of this topic. I also benefited from exchanges with James Johnson, Henri Gentis, Vera Schut; with culture expert Fons Trompenaars; with tech adoption expert Daan Noordeloos; with organization structure expert @Tim de Boer; and with people playing key roles in AI adoption, such as ING's Daicka Gruisen.

So here, in short, are the why, what, and how of agentic transformation.

WHY AGENTIC TRANSFORMATION MATTERS

Agentic transformation matters because it improves performance by reducing direct human intervention in product or service delivery, where appropriate.

This is why Jeff Bezos launched Prometheus, a hundred-billion-dollar investment fund (1). The idea is straightforward: assess agentifiability, execute agentic transformation, and exit with significant value creation. The smartest and winning Private Equity companies will learn this game. It is essentially understanding value creation potential from the playbook behind Amazon’s success:

lower cost through smarter work reallocation and less proportional headcount growth
fewer errors, shorter cycle times, and lower variability
fewer missed opportunities

Closer to home, agentic could help address some of the biggest challenges facing our institutions, such as the Dutch Tax Administration, facing an ageing technology workforce nearing retirement and talent gap, and a legacy system that can no longer realistically be untangled and must instead be managed intelligently.

Independent research by Eigenvector.eu indicates that around 35% of processes appear well suited for agentification today (2). That may sound conservative compared with some of the grander claims made by tech vendors, but it is already enough to unlock major value and more may be within reach, as discussed below.

WHAT AGENTIC TRANSFORMATION IS

To understand agentic transformation, it helps to distinguish between three stages stages, which can overlap.

1. AI enablement

This is where many organisations are today, rolling out AI literacy and Microsoft Copilot training. They train people to prompt better, summarise faster, write more efficiently, and support their existing work with AI.

2. AI acceleration

The second stage is more ambitious. Here, it is about making it possible to do things leaders often dream about, but usually postpone because they seem too time-consuming or too expensive to prioritise.

Think of a new proposition, a new distribution channel, or a new internal tool. These are the kinds of things executives often want, but quietly assume will take months of meetings, requirements gathering, budget discussions, and development effort. Using tools such as Cursor, Claude Code or Claude Cowork, organisations can now build in days, sometimes even hours, what used to take months or years.

A concrete example is Achmea, where, with the help of AI agency ImagineDigital, it took only two weeks to develop a car configurator from A-Z that would normally have taken a year.

A practical way to start this conversation is not with theory, but with experience: let leaders go through a well-designed one-hour exercise in which they use multiple AI tools to create a new proposition at speed. Then ask the question:

What have you been wanting to build for years, but kept postponing because it felt too slow, too costly, or simply unrealistic?

ZHC, zero-human-company experiments such as those being conducted by Olivier Rikken and Yassine El Kochta as we speak point to the next level of this. New companies can be created with a single prompt. They show the rapidly emerging opportunity for fast, serial concept ideation and launch.

AI acceleration therefore uses AI-enabled tooling to compress development time dramatically while humans still remain firmly in charge.

3. Agentic transformation

Then comes the third stage. This is where the redesign becomes deeper. It changes who or what does the work. Agentic transformation is the redesign of business processes so that agents take over meaningful tasks or task clusters, while humans remain involved where needed for judgment and control to ensure safety and compliance.

Sounds risky? Not if you know what you are doing. Think of the 1985 saying by Peter Drucker: “Entrepreneurship is ‘risky’ mainly because so few of the so-called entrepreneurs know what they are doing”. So let’s provide the essential knowledge and make it less risky.

HOW AGENTIC TRANSFORMATION WORKS

Before any organisation starts automating core work, it needs a structured framework that executives can trust to make agentification work: a way to judge which processes are actually suitable and then build and implement an architecture that ensures safe, compliant, and effective automation.

This is where Eigenvector's work comes in. Marco van Hurne developed PASS, the Process Automation Suitability Scan, and PADE, the Process Automation Design Engine, as a structured framework for enterprise agentification (2). PASS helps determine whether a process can be automated and under what type of architecture. PADE then determines and implements the appropriate architecture. It's being applied at ASML today.

This architecture, also called scaffolding, determines whether agents can operate safely, compliantly, and effectively. It includes the way tasks are sequenced, the controls that are built in, the role of human approval and supervision, the data and rules the agent can rely on, and the way actions are recorded for an auditable trail.

Let’s unpack three key components of the architecture: human judgment and control, auditability, and different types of AI.

A. Human judgment and control

Not every process should be fully autonomous, and not every process needs the same kind of human involvement. This is where it helps to distinguish between different forms of oversight (3).

Human-on-the-loop means the agent can operate within predefined boundaries, while a human supervises the process at a higher level and steps in when thresholds are crossed, ambiguity arises, or outcomes move outside expected norms.
Human-in-the-loop means a human actively approves or intervenes before the process can move forward. This is relevant where the consequences of error are high, where judgment is still essential, or where regulation requires direct human involvement.

New roles and job titles, such as Agent Orchestrator, are rapidly emerging and are already being hired for (4)

In addition, the architecture should contain clear guardrails. These define what the agent is allowed to do, which tools it may use, which budgets or actions sit outside its authority, and when the process must escalate. Finally, robust setups also need some form of emergency brake: the ability to interrupt, pause, or override the process when needed.

The key point is that the mature question is not whether humans are in or out. The mature question is: where is human judgment still needed, in what form, and at what moment?

B. Auditability

If agents are taking meaningful action inside core processes, companies need to be able to show what happened, why it happened, what data or logic informed the decision, and where humans did or did not intervene. That is not bureaucracy. That is what makes scale trustworthy. Without a reliable audit trail, a record of decisions and actions, companies cannot properly explain decisions, monitor quality, learn from errors, or demonstrate compliance.

C. Different types of AI for different kinds of work

Where reliability matters, leaders should know the term neuro-symbolic AI. In plain language, it means combining neural AI, the type we know from ChatGPT and others, which is powerful but probabilistic, with a harness of more rule-based or classical linear AI, which is better suited to deterministic outcomes where reliability matters and hallucination must be excluded.

This is also where the upside becomes interesting. Marco van Hurne argues that current effective process coverage of roughly 35% could rise to around 55% with neuro-symbolic AI, and that agents improving not only their outputs but also the architecture around them, could push the theoretical ceiling further, to 70% to 80% in domains with sufficiently strong governance (5).

That is a major reason why architecture matters so much.

HOW LEADERS SHOULD RESPOND

Given this potential, a board-level dialogue is needed on how strategic embedding, adoption, workforce transition, organisational change, architecture design, and programme management should be handled. In other words, a total transformation package.

1. Reassess strategy and the innovation portfolio for new opportunities

If AI acceleration and agent-like ways of working materially reduce time to market, many organisations should revisit their innovation portfolio.

Ideas that once looked too slow, too expensive, or too resource-intensive may suddenly become viable. That creates a need to reassess the innovation portfolio. In practical terms, leaders should ask again: Do we have enough in the portfolio? Do we have the right mix of bets? Are we still allocating resources in line with what is now possible?

At the same time, agentic transformation should not sit as a side initiative somewhere in operations or IT. It needs to be strategically embedded, linked to where the company wants to create advantage, where it wants to improve its operating model, and where it wants to redesign execution for structural benefit.

2. Deeply understand adoption, define the new organisational structure, and prepare people for work reallocation

Technology does not adopt itself. Organisations need to understand what drives adoption, what blocks it, and respond accordingly. Three human drivers matter as Daan Noordeloos explains:

Build on people’s desire to progress, while reducing the fears that often come with change: loss of autonomy, loss of status, and especially the fear of feeling incompetent. That last one is often underestimated. People need help learning how to work with agents, how to supervise them, and how their own role changes as a result.
Manage meaning. People need a clear and credible story that explains what is changing, why it matters, and what it means for them.
Recognise the power of belonging and coalition formation. People do not adopt change in isolation. They take cues from the groups they identify with. Leaders therefore need to build positive coalitions around the change and prevent negative ones from forming, such as “those guys from IT”.

Agentic transformation changes work fundamentally. It requires a different organisational structure, new roles and responsibilities, and a new view of the future workforce. That makes workforce design a critical task for CHROs. Organisational digital twin software, such as that developed by Reconfig.no, can help determine what the future setup should look like, including where work will sit, how roles will change, and how many people will be needed in the new model.

It also means reskilling, upskilling, redeployment, and helping people move into new roles, including roles such as Agent Orchestrator. Redeployment, in particular, means helping employees whose work is heavily affected by automation move into other roles within the company wherever possible. This is where a broader view of talent becomes important. Tools such as the Employability Scan, as developed by THT Consulting, can help companies and individuals look beyond past job titles and formal skill sets, and identify the potential that makes people suitable for roles that conventional screening might otherwise exclude.

3. Work on the culture, especially the tension between speed and safety, and align agent behaviour to your strategy and culture

This is not just a technology issue. It is a cultural one. Many organisations still treat safety and compliance on the one hand, and speed on the other, as a trade-off. That is understandable given the recent past, but it is too simplistic and no longer suited to the challenges companies face, such as businesses responsible for the energy transition, or financial services firms threatened by much faster-moving fintechs such as Revolut.

The better question is: how can we increase speed through more safety, and increase safety through more speed? That is especially relevant in a world of greater uncertainty, where companies need to move faster without losing control.

As agents take over part of the work, alignment of agent behaviour with strategy and culture becomes essential.

This part is easy to overlook, and important not to. If agents work in ways that contradict the strategy, tone, risk posture, or cultural values of the business, then the company has not improved execution. It has simply automated inconsistency. Agentic transformation therefore also means designing agent behaviour in line with how the company actually wants to compete and operate. In other words, culture definition and culture change work will increasingly apply not only to humans, but also to agents.

4. Adapt the architecture and company governance

If agents are operating inside meaningful business processes, architecture cannot be an afterthought. Leaders need to ensure that the required architecture is designed, adjusted where needed, and technically supported over time. That includes the right controls, the right integration choices, the right forms of supervision, and the right mechanisms to improve reliability in higher-risk contexts. As agents take over tasks, governance and liability need to be handled in fundamentally different ways, a topic that deserves a separate article.

5. Put programme management and organisational change in place

This is not plug-and-play. It requires coordination, organisational change, and deliberate leadership attention. New ways of working need to be introduced, teams need clarity on where agents fit, and decision-making often needs to adapt.

Without the dialogue on these five points, companies end up with isolated experiments rather than real transformation. As a board member or C-suite executive, ask yourself whether these points are getting sufficient attention.

FINAL THOUGHT

Most companies are still focused on helping people use AI inside existing work. That is necessary, but it is not where the biggest opportunity lies. The bigger opportunity is to redesign work itself, under the right architecture, with the right controls embedded in that architecture, and with human judgment where it still adds value. That is what I call agentic transformation. And board members and senior executives should get up to speed on it now.

If anything in this piece triggered a question, a concern, or an idea for your own organisation, feel free to reach out. These are exactly the conversations leaders should be having now.

About me and why this article

I work as a strategist and boardroom advisor at the intersection of strategy, AI, and culture. In that role, I collaborate with experts in AI and culture, and see it as my role to translate deep specialist expertise into insights that help boardrooms have the right conversations, quickly and effectively.

IntotheNXT | Entrepreneurship and innovation, by design.

If you want to deep dive on some of the topics in this article, read: