Why 95 percent fail – and it’s not because of the technology 

The technology works. Large language models draft contracts, generate code, analyze documents, and automate customer service requests. Yet the organizational impact remains limited in most companies. The Deloitte study “State of AI in the Enterprise” confirms this: more than 70 percent of companies have implemented generative AI — but only 34 percent are truly reshaping their business by changing business models, roles, and ways of working. And just 6 percent have fully implemented agentic AI — the next stage of intelligent automation. 

Why is that? The joint study by DLR, Saarland University, and Frankfurt School provides a clear answer. It identifies five success factors for AI transformation: strategy, processes & implementation, technology, governance, and culture. The greatest contribution to success comes from the topics around processes and organizational implementation. In companies that are particularly successful at scaling AI — the so-called “AI transformation leaders” — all five factors are strongly developed. The insight: anyone who addresses only one factor in isolation puts the success of the entire transformation at risk. 

Anyone familiar with the McKinsey 7S framework will recognize the pattern immediately. Strategy, structure, systems, skills, staff, style, and shared values — all seven elements must be aligned for change to succeed. AI adoption is no exception. It is a transformation project that affects all dimensions of the organization. Anyone who treats AI as a pure IT project will not fail because of the technology — but because of the lack of organizational maturity. 

What does that mean in concrete terms? Let’s look at the five most common stumbling blocks. 

Without process understanding, AI only shifts the bottleneck 

Imagine a clerk in order processing using AI to create quotes twice as fast. That sounds like a clear productivity gain. But if the downstream approval step — without AI — continues at the same speed, the output piles up at the next process step. The organization’s overall throughput remains unchanged. The improvement evaporates — or worse: it creates pressure exactly where the team is already operating at the limit. 

What is happening here is precisely described by Eliyahu Goldratt’s theory of constraints: if you improve throughput at a point that is not the bottleneck, you improve nothing — you even increase waste. The output of a system is always determined by its weakest link. Using AI at the wrong point is like putting a larger funnel on a bottle with a narrow neck. 

There is also an individual effect described by Parkinson’s Law: work expands to fill the time available for its completion. If AI creates room for the individual but no organizational realignment takes place, the time gained is often filled with other tasks — without any measurable added value for the organization as a whole. 

The consequence is uncomfortable but logical: before AI is deployed, reliable process documentation is needed. The process must be understood end-to-end — as it actually happens in operational reality, not as it was documented in the last PowerPoint presentation. Only then can you assess where the actual bottleneck lies, which process steps add value, and where automation — whether through AI or classical methods — really makes a difference. And only then can you sensibly evaluate whether tasks that are currently in the wrong place can be moved to where they belong — to put the organization in a better overall position. 

BLIKS IO enables exactly this first step: the platform captures processes, roles, and systems directly from employees — bottom-up, in plain language, and without months of workshop marathons. This creates a reliable snapshot of the organization’s current state, on which informed decisions about AI deployment, task allocation, and capacity planning can be built. 

If people fear for their role, they won’t surface potential 

In addition to process understanding, there is a second, often underestimated hurdle: the human dimension. The Pronova-BKK study “Working 2025” shows that more than half of AI users now perceive their work as more error-prone or less secure. According to the Xing labor market report, one in six employees in Germany now believes that their own job is at risk because of AI. Three quarters of those who are worried expect this could happen within the next five years. 

Now imagine the extreme scenario: an experienced employee realizes that their entire area of responsibility could be automated by AI. Will they proactively disclose this potential? Rationally speaking, that would endanger their own position. The natural reaction is not open sabotage — it is more subtle and deeply human: opportunities are not reported, improvement suggestions do not come forward, and the willingness to change quietly, but effectively, declines. 

And the threat goes deeper than fear of losing a job. Someone who has built up 15 years of expertise in a specialized field and sees AI producing comparable results in seconds is not just experiencing economic uncertainty — they are experiencing an identity threat. It is about professional self-understanding. That is why rational arguments like “Your job is safe” often fall short. What employees need — and what good change management must deliver — is a concrete, honest picture of what their role will look like in the organization of the future — not a vague promise, but a comprehensible plan. 

BLIKS IO takes this dimension into account from the outset. The platform captures not only processes, but also roles, responsibilities, and participation in their real-world form. When an organization introduces AI and tasks shift, such a digital twin of the organization makes it possible to plan role changes early and transparently — instead of improvising them reactively. The perspective shifts: away from “AI is taking something away from me” toward “we are shaping together how our collaboration will evolve.” And only with this perspective do people open up to the potential that actually lies dormant in the organization.  

AI amplifies what is already there — including the bad 

Even when processes are understood and roles are secured, a third hurdle remains, particularly tricky in established organizations: the reality of the systems and data that are actually being used. 

AI agents that automatically pull data from CRM systems, generate reports, or classify customer inquiries — these are the promises of agentic AI. But what happens if employees aren’t even working in the CRM? Shadow systems are ubiquitous in established organizations: Excel lists instead of a project management tool, email distribution lists instead of a ticketing system, local drives instead of cloud storage. Anyone who has ever introduced an ERP system knows the phenomenon. The reasons are rarely malicious — they are practical: the official system is too slow, too cumbersome, or does not reflect the actual workflow. 

Anyone who wants to use AI and automation sustainably needs a common foundation: employees who work according to defined standards, and systems that accurately map actual processes. The best AI agent is ineffective if it never receives the data it needs — because operational reality takes place in shadow systems. 

To recognize these patterns, you need broad knowledge of the organization. Not as a guess from the IT department, but as structured feedback from the people who work with the systems every day. Why is calculation done in Excel rather than in the CRM? If the answer is that the CRM has performance problems or does not cover certain use cases, then the first task is not AI integration — it is eliminating the causes that allow shadow systems to emerge in the first place. 

BLIKS IO makes exactly these patterns visible. Its decentralized data collection captures how employees really work — which systems they actually use, where media discontinuities occur, and where operational reality diverges from the documented target state. This is the foundation without which any AI deployment risks amplifying existing problems instead of solving them. Because AI is an amplifier: good processes and clean data become better. Bad processes and fragmented data become worse — only in accelerated form. 

Where no one owns the full process, every AI initiative runs aground 

The fourth hurdle is so commonplace that it is rarely addressed openly: in many established mid-sized companies, no one is responsible for a process end-to-end. Sales is responsible for customer contact, order processing for fulfillment, controlling for billing — but who is responsible for the entire flow from first contact to invoice? Often: no one. 

Without end-to-end process management — without someone who has oversight of an end-to-end process — there is also no one who can make an informed decision about where AI use would make sense. There is no one to drive the change, address resistance, and measure success. AI projects without clear process ownership end up as isolated point solutions — technically functional, organizationally ineffective. 

Then there is the silo problem: the greatest AI potential often lies in cross-departmental processes — precisely where responsibilities end and knowledge stops in most organizations. Anyone who has ever fought through five rounds of alignment because a different stakeholder was missing each time knows the result: energy is wasted on coordination instead of outcomes. 

BLIKS IO creates the foundation here on two levels. First, it captures processes, roles, and system dependencies across the organization and brings them together in a digital twin — as a shared basis for decision-making for everyone involved. Second, it makes visible in which networks people actually collaborate across departmental boundaries. This finally makes it transparent who needs to be at the table for which topic — not based on organizational charts, but on lived collaboration. This creates working groups that can work productively toward an outcome from the very beginning, instead of spinning in endless coordination loops.  

The demographic race: secure knowledge before it’s gone 

And then there is a fifth hurdle that cannot be solved with better project management, but is simply a matter of time. In many mid-sized companies, the most experienced knowledge holders will retire in the next five to ten years. These employees carry the tacit knowledge that keeps processes running — not in documents, not in systems, but in their heads. They know why the special case for customer X is handled differently, which Excel logic bridges the interface between two systems, and whom to call when the standard process does not work. 

Without this knowledge, AI lacks the foundation on which it can work in a value-adding way. An AI agent trained on documented target processes fails against operational reality because undocumented special cases, workarounds, and experiential knowledge are missing. Capturing process knowledge from employees — in a structured, systematic, and timely manner — is therefore not a nice documentation exercise, but a strategic necessity. Those who wait too long lose not only experience, but also the basis for any sensible AI deployment. 

BLIKS IO addresses this pressure for action directly: the platform makes it possible to convert knowledge from the minds of employees into structured, machine-readable process descriptions — before it is irretrievably lost. 

The right sequence: first understand, then change, then automate 

AI adoption is not an IT project that can be delegated to the technology department. It is an organizational development project that must address strategy, processes, roles, culture, and technology simultaneously. The sequence is crucial: 

First understand the organization – capture processes, roles, and systems as they actually function in operational reality. Not as on the latest org chart, not as in the quality manual, but as employees work every day. 

Then optimize the process flow – identify bottlenecks, address shadow systems, clarify responsibilities. The question is not “Where can we use AI?”, but “Which parts of our value chain are not working optimally today — and why?” 

Then automate and use AI where it makes sense – on an organizational foundation that can support it. Not as an isolated tool for individuals, but as an integrated lever for the entire organization. 

BLIKS IO was built for exactly this sequence. As a platform for digital potential inventory, it captures processes, roles, and systems holistically — thereby creating the organizational foundation without which AI adoption remains an expensive experiment.