AI-driven organisations 

AI accelerates work, but organisation-wide change often lags behind 

AI has rapidly taken a more prominent role in the daily work of European organisations. The immediate impact is clear: work gets done faster, processes become smarter and employees receive better support. But when you look more closely, a different picture emerges. Our research into the digital reality of European organisations shows that this progress does not yet consistently translate into a fundamentally different way of working.

According to Bastiaan Sjardin, AI Officer at Conclusion Intelligence, the explanation lies not in the technology alone, but mainly in how organisations apply it: “Using AI within existing processes speeds up the work, but leaves the way of working unchanged.”

Most organisations currently use AI primarily to accelerate existing tasks: processing invoices, summarising emails, extracting information from documents. That delivers time savings, but the core way of working remains the same. And doing yesterday’s work more efficiently is not a strategy for tomorrow. True AI transformation is not just about efficiency, but about reinventing your core services. Those who dare to do so remain distinctive in the market.

The first phase of AI adoption is clearly visible: organisations use AI to become faster and more efficient. This makes sense, as adding AI to existing workflows is relatively straightforward. The next step is more challenging. It is no longer about tools, but about choices.

“Working fundamentally differently requires redesigning processes, organising decision-making differently, assigning ownership clearly and defining where human judgement remains essential,” says Bastiaan. “That directly affects how your organisation is structured. And that is where friction arises.”

Consider processes where AI not only supports but also contributes to decisions or takes over work entirely, such as automatically processing and paying invoices or instantly approving standard claims. Suddenly the question becomes: who intervenes when something goes wrong? This requires fundamentally different processes, responsibilities and governance.

In most organisations, there is an awareness that AI can be more than an efficiency tool. Yet that awareness does not always lead to clear decisions. Technology plays a real role here: many organisations still need to make their cloud and data foundations suitable for GenAI and agents. And where the technology is ready, another barrier often emerges: the willingness to adapt processes, responsibilities and governance.

Bastiaan: “The bottleneck is rarely the availability of technology and expertise. The real question is whether you are willing to adapt your organisation to it. That is also a governance issue.”

So this is not a matter of technology or organisation; the two reinforce each other. Without alignment between data, governance, people and steering, there is no structural impact. In practice, this is evident in organisations rolling out AI solutions and ticking them off, without actively monitoring whether the promised value is actually realised in day-to-day work.

What is consistently underestimated is what it takes to embed AI into everyday operations: architecture, data, integrations with operational systems, MCP and agent-to-agent communication, as well as legal, privacy, security, compliance and human oversight. Ultimately, AI must be embedded into real processes, systems and decisions. That is where lasting value is created. A mature AI lifecycle does not need to be fully in place from day one; what matters is a workable roadmap and a clear framework to grow towards it step by step.

Whether AI truly leads to new ways of working largely depends on leadership—specifically on making directional choices and allocating the resources to realise them.

“Organisations that make progress are explicit about what they want to achieve with AI and align their organisation accordingly,” says Bastiaan. “The difference lies in safeguarding alignment between technology, team structure, governance and business domains. Without that, AI remains a tool: it helps, but does not fundamentally change anything.”

AI literacy in the boardroom is therefore becoming increasingly important—not to master the technology, but to make better decisions about direction, risk and dependencies. Without sufficient understanding, discussions quickly shift to tools or hype, while the real governance questions concern competitiveness and the interplay between technology, processes, capabilities and control. This is where the board makes the difference: AI affects the entire organisation and requires clear direction from leadership and senior management. When that direction is sharp and properly resourced, AI evolves from experimentation into a structural part of daily work.

A common misconception is that organisation-wide adoption will follow automatically once AI is widely available. Enabling employees to work faster with AI tools delivers quick, measurable gains, but core service delivery only truly improves when the foundation is in place: processes, responsibilities and governance. The two reinforce each other: broad use of tools generates use cases that sharpen the foundation, while a strong foundation unlocks the real value of those tools.

Bastiaan: “Adoption does not happen just because everyone can use a tool. It happens when processes, responsibilities and agreements around quality and control are properly organised.”

This also includes the buy-or-build decision. Rolling out a standard tool such as Copilot or building your own solutions on proprietary data and processes are not separate decisions, but part of the same strategy. Standard tooling supports day-to-day work; proprietary development strengthens differentiation. The most value is created when both are driven from a single direction, governance model and architecture.

The real return from AI lies not only in the first application, but in what an organisation can do afterwards: developing new use cases faster and improving collaboration between business and technology. “The first use case delivers visible value,” says Bastiaan, “but the real gain lies in the capability you build afterwards.”

Start with focus. Demonstrate convincingly in a few business-critical processes how AI becomes part of the work. Think of invoice processing with human oversight on exceptions, or AI-assisted programming with clear quality and security standards in the delivery chain. This is how AI evolves from isolated pilots into scalable application. Without that capability, AI remains fragmented; with it, it becomes scalable.

An increasingly important aspect is control. Many AI applications rely on LLMs from large vendors. This offers speed and scale, but also introduces risks in core processes: rate limits, pricing changes, latency, performance differences between model versions, changing licence conditions and the discontinuation of APIs.

According to Bastiaan, the goal is not to avoid dependency, but to make conscious choices:
“You are always dependent on something. The governance question is where you accept that, and where you want to retain control.”

For many organisations, today’s most powerful foundation models are still too large or too costly to host themselves. External LLM infrastructure—often provided by US-based vendors—is therefore a logical choice, but not without implications. Particularly in processes that underpin your services, you must explicitly decide where speed and scale take priority, and where control over models, data, architecture and fallback options weighs more heavily. Building a critical process on a single external model API effectively places part of your operating rules outside your organisation. That is manageable—provided the choice is deliberate, alternatives are available and the architecture remains flexible enough to switch later.

This risk became tangible recently when a US export measure abruptly cut off access to Claude Fable, Anthropic’s top model. The lesson: make your solution model-agnostic so you can switch quickly if an API becomes unavailable, and build internal capabilities to keep open-source LLMs production-ready as a viable alternative. Running models yourself requires in-house expertise in ML engineering and MLOps. What is easiest today may make you most vulnerable tomorrow.

These strategic choices operate at multiple levels. Leaders first determine where AI should create value: productivity, decision-making, service delivery or new propositions. Then comes the question of which AI capabilities to source for speed and scale, and which to develop in-house because they are tied to differentiation, data or governance. Organisations that fail to make these decisions explicitly end up making them implicitly—with all the associated risks of lock-in or loss of strategic control.

A realistic first step is usually not a large-scale AI programme, but selecting one or two business-critical domains where speed, quality and decision-making intersect. In these areas, AI can be applied in a targeted way—provided you simultaneously define the business case, architectural guidelines, human oversight and process design.

It helps to distinguish early between low-risk applications and critical applications with stricter requirements around data, context and compliance. This prevents promising initiatives from stalling later due to governance or architectural issues.

The technology is available and the first results are visible. The more difficult question lies beneath: what will AI mean for your core services in the future? What will your customers expect, and how will the competitive landscape evolve? In many sectors, the source of customer value—and differentiation—is already shifting. Organisations that use AI only to accelerate existing processes optimise the present while the market moves on.

According to Bastiaan, there are two tracks: making current work more efficient and redesigning your services before the market does it for you.
“The tools are there,” he concludes, “what is missing is the willingness to truly adapt processes, decision-making and responsibilities.”

AI is therefore not just a technological issue, but a strategic choice about your position in the market. Organisations that take this step do not stand out because of better tools, but because they dare to redesign underlying processes and treat AI not as a separate application, but as an integral part of architecture, governance and steering. The key question for leaders is no longer what AI can do for them, but what they are willing to change to realise that value.

What will your core offering look like when AI becomes the norm? What role do you want to play for your customers? And are you already building towards that future?