The Organisational Conditions That Determine AI Success
Thu 23 Apr 2026
DISCLAIMER: The views and perspectives shared in this interview are entirely those of the interviewee, informed by her broader professional experience in digital transformation and AI enablement. They do not represent the views, strategies, or positions of her current or former employers, nor do they reference any proprietary or confidential information.
AI adoption is frequently discussed as a technical challenge, centred on tools, models and capability. In practice, the obstacles tend to emerge elsewhere. They appear in how organisations design work, distribute responsibility and build trust across teams.
Across industries, companies continue to invest heavily in AI platforms, training programmes and pilot initiatives. Yet translating that activity into sustained impact remains uneven. The underlying issue is often alignment: between systems, people and the realities of day-to-day operations.
In this interview ahead of Tech Show Frankfurt, Dr Ofure Obazee, Associate Director of RQS Digital Capability & AI Enablement at Merck Healthcare KGaA, reflects on what organisations continue to overlook. Drawing on her experience in data transformation and organisational enablement, she examines why training alone rarely delivers change, how AI surfaces hidden decision-making dynamics, and what distinguishes organisations that embed AI into everyday workflows from those that struggle to move beyond experimentation.
– – – – – –
You’ve worked across both data analytics and organisational enablement. Where do organisations still struggle when it comes to adopting AI successfully?
Most organisations are solving the wrong problem. They invest heavily in tools, pilots, and platforms – then wonder why adoption stalls. It stalls because they treated AI as a technology rollout when it is fundamentally a human transformation. They deployed AI into environments that were never prepared to absorb it – workflows unchanged, decision rights untouched, trust unaddressed. That is why so many organisations can say “we have AI” while very few can honestly say “we are working differently because of it.”
The second struggle is subtler but more damaging. Leaders define success too narrowly – efficiency, cost reduction, and utilisation metrics. When that is the only lens, people experience AI as something being done to them rather than with them. The questions that actually determine whether adoption succeeds are almost always human questions:
What work should be augmented, and what must remain human? Where does judgment sit? What anxieties are people carrying but not voicing?
In environments where professionals have spent decades building expertise-based identities – compliance, regulatory, quality assurance – those unvoiced anxieties are not obstacles or projects to manage around. They are the adoption strategy, which, if missed, may result in an AI roadmap that looks good on a slide but still fails in practice.
In many organisations, AI enablement is approached through training. Where does that approach tend to fall short in practice?
Training matters, but training alone is not enablement. Training answers the question, “Do people know how to use the tool?” Real enablement answers a much harder question: “Has the work itself been redesigned so the tool can be used meaningfully, safely, and sustainably?” That is another point where many organisations fall short. They run workshops, publish prompting guides, maybe even certify people, but they do not resolve the operational realities around workload, governance, escalation, accountability, or workflow friction. A pattern that shows up consistently in adoption research is teams leaving workshops inspired.
The other limitation is that training often assumes the problem is capability, when most times the real issue is trust, confidence or perceived threat. If a frontline employee quietly believes, “I am being asked to train my replacement,” no amount of prompt training will solve that. If a manager fears loss of control, they will slow adoption regardless of how many learning modules they have completed. So yes, teach the tools, but do not confuse literacy with readiness. Readiness is cultural, relational, and structural.
When AI tools are introduced into existing workflows, what tends to change and what tends to remain unchanged?
What changes first is speed. Summarising, drafting, searching, synthesising, triaging – people feel that acceleration immediately. What does not change fast enough is everything around it: decision-making structures, accountability, management behaviours, and performance expectations. So you get an odd and often undesirable situation where the work becomes faster, but the operating model stays frozen. That does not reduce pressure – it increases it, because people are now expected to produce more, respond faster, and absorb more ambiguity without any genuine redesign of their role boundaries.
What also tends to remain unchanged – and this matters enormously – is organisational power. AI may alter tasks quickly, but it does not automatically democratise decision-making. If anything, it can reinforce top-down control. Leaders suddenly have more data, more monitoring capability, and more confidence in standardisation. This dynamic is particularly pronounced in highly regulated environments, where the instinct is often to use AI to tighten control rather than empower judgment. That is why I frame this as a workplace relationship conversation, not a tooling conversation. The question is not only what AI changes, but who gets to shape that change – and whether they feel safe enough to be honest about what is and is not working.
How does the relationship between people and AI systems actually evolve once these tools are in use?
In the early stage, the relationship is transactional. People use AI like an assistant in mostly single-thread chats – draft this, summarise that, speed this up. But as adoption matures, something more revealing happens. AI starts to expose how people actually make decisions, where they rely on tacit knowledge, where they do not trust automation, and where the organisation has been depending on invisible human judgment all along.
This is why I say AI becomes a mirror before it becomes a multiplier. It shows you how work really happens – not how the strategy deck says it happens.
Over time, in ‘human-safe’ organisations – ones that have invested in psychological safety, narrative reframing, and genuine trust-building – AI becomes a collaborator. It expands human judgment, reduces cognitive load, and frees people to focus on nuance, relationships, and exceptions. In less healthy organisations, AI becomes a supervisor by proxy: a source of pressure, standardisation, surveillance, and second-guessing.
That distinction is not theoretical. It plays out in real behaviours. Are people experimenting with AI openly or hiding their usage? Are they asking questions, or performing compliance? Are they integrating AI into their professional identity, or quietly resisting it? The future of AI at work is not decided by model capability alone, but also by whether organisational design is for partnership or subordination.
Many adoption strategies are designed from the top down. Where do organisations miss critical insight from the people using these systems day to day?
They miss it at the exact point where strategy meets lived reality. Frontline workers know where the friction is, where the exceptions hide, what quality actually depends on, and which parts of the job look repetitive from a distance but are full of judgment in practice. A pattern that emerges repeatedly in regulated contexts is that the tasks leadership marks as ‘easily automatable’ are often the ones most saturated with undocumented expertise. When such perspectives are skipped, and the visible layer gets automated, it easily destroys the less-visible expertise holding everything together.
But the insight organisations miss most is emotional truth. Leaders hear formal adoption language: “Yes, we are using the tool.” What they do not hear is: “I do not trust this output.” “This adds rework.” “This makes me feel monitored.” “This technically works, but it makes my job worse.” Those are not soft signals – those are the most important adoption signals you will ever get. The organisations that embed AI successfully are the ones that build mechanisms to surface those truths early – through communities of practice, blameless retrospectives, and persona-adapted feedback channels – rather than waiting for resistance, shadow workflows, and quiet disengagement to announce what people were too afraid to say.
Organisations often have to balance caution with the need to move quickly. Where does that tension become most visible?
It becomes most visible in the gap between safe experimentation and scale. Most organisations now understand they cannot wait forever, but many still have not worked out how to move at pace without becoming careless. So you get a split: some parts of the business want speed, others want control and guardrails – sometimes dressed as assurance – and the workforce is caught in the middle trying to interpret mixed signals with no clear permission to act.
The tension is also visible in leadership language, and this is something I pay close attention to. When leaders say “we need to move fast,” employees often hear “we do not have time to discuss risk, workload, ethics, or unintended consequences.” But when leaders over-index on caution, they create paralysis – and people go around them anyway, adopting AI tools informally without governance, guardrails, or organisational learning. The answer is not speed versus safety. It is structured experimentation: small-scale pilots with clear boundaries, transparent criteria for what success actually looks like, and strong feedback loops that include the people doing the work – not a hand-selected tech-savvy minority. You move at pace without losing trust by making experimentation visible, safe, and genuinely informed by the frontline. That is what human-safe adoption looks like in practice.
Looking across the teams you’ve worked with, what distinguishes those that embed AI into how people work from those that continue to struggle?
The teams that embed AI successfully do three things differently. First, they treat AI as a shift in how work is designed, not a tool to layer onto existing processes. Second, they bring business leaders, technical teams, and frontline practitioners together from the start – not in sequence, not as an afterthought. Third, they define value in human terms: not just “did we reduce cost?” but “are people making better decisions? Is friction decreasing? Are we building capability? Is there room for higher-value work?” They do not ask “where can AI replace effort?” – they ask “where can AI elevate contribution?”
The organisations that continue to struggle usually share a pattern: they layer AI onto broken processes, over-focus on tool access while ignoring adoption conditions, mistake executive enthusiasm for organisational readiness, and treat trust as a soft concern rather than a structural requirement.
Generally, the organisations that win at AI adoption are not necessarily deploying faster – they are redesigning work more honestly. They recognise that sustainable adoption happens when people feel AI is making them more capable, not more disposable. And they invest in proving that – not through messaging, but through the actual experience of the people doing the work every day.
