1. The False Debate
Henri Allegra, Homo Promptus and Niels Quinten, Best Goed studio
Ask most business leaders how to succeed with AI transformation and you will get one of two answers.
The first: fix the technology. Get the right tools in place, integrate them properly, make sure they are adapted to how people actually think and how the organisation already works. If AI is hard to use, unintuitive, or culturally tone-deaf, adoption will stall regardless of its capabilities. The solution is better design; human-centred, empathetic, grounded in real organisational contexts.
The second: fix the people. Train them, upskill them, shift their mindset. If employees resist, misuse, or blindly delegate to AI, the problem is not the tool, it is the user. The solution is cognitive and cultural transformation.
Both arguments are compelling. Both are backed by evidence. And both are incomplete. On one hand, a perfectly designed AI deployed into an organisation of cognitively unprepared users will underperform. On the other hand, a workforce trained in AI collaboration but handed poorly designed, context-blind tools will hit a ceiling fast. The uncomfortable truth is that neither movement succeeds without the other.
Research confirms what practitioners already sense; human-AI combinations do not automatically outperform the best of human or AI working alone, and on average the gains are neither guaranteed nor evenly distributed (Vaccaro, Almaatouq & Malone, 2024). They depend entirely on how well the collaboration is designed and how intentionally humans engage with it.
In other words: the question is not which transformation matters most. It is why most organisations are attempting only one and wondering why results disappoint. The real shift is neither purely technological nor purely organisational; it runs through both simultaneously. What follows is an argument for why treating these two positions as alternatives may be the most expensive mistake leaders can make right now.
2. Two Necessary Movements
Movement 1: Cognitive Evolution
Henri Allegra, Homo Promptus
The human side of this transformation is the one most organisations underestimate, and most AI vendors purposefully ignore. What most enterprise AI programmes call training is closer to product onboarding: people learn the interface, not the collaboration. Cognitive evolution is something else entirely, it is what happens when professionals begin to change how they think, not just what they delegate to.
Real cognitive evolution starts with understanding what happens in the human mind when AI enters the equation. Decades of cognitive science tell us that humans naturally offload memory, effort, and judgment onto external systems when those systems are available (Risko & Gilbert, 2016). This is simply how we manage cognitive load. The problem arises when offloading becomes the default: when we stop asking whether we should delegate a judgment and simply do it because we can. This is a different question from the automation debate, not what AI does to tasks, but what humans do to their own thinking when AI is available.
Efficiency gains from offloading tasks can coexist with measurable deskilling over time (Grinschgl et al., 2021). The more fluently we use AI, the more invisibly we may be eroding the very capabilities we rely on. This is the delegation trap, and most AI adoption programmes walk organisations directly into it.
Augmentation vs. Delegation: The Core Distinction
At Homo Promptus, we draw a sharp distinction between delegation and augmentation. Delegation says: let AI do it. Augmentation says: let AI make me better at doing it. The difference is not semantic, but cognitive and cultural.
Augmentation requires intentionality: knowing why you are using AI and what for, what you want to amplify, and what must remain human. But intentionality alone is not enough. It must be paired with agentivity, the felt sense of being the author of your own thinking and decisions, even when AI is deeply involved in the process.
Together, intentionality and agentivity are the twin conditions of true augmentation: you know why you are engaging AI and what for, and you remain the author of what emerges. That combination requires metacognition; the capacity to observe and regulate your own thinking while working with AI (Sidra et al., 2025), and both can be deliberately developed.
Prompt Fluency as Professional Literacy
This is why prompt fluency matters, not as a technical trick, but as a new form of professional literacy. A prompt is not an instruction to a machine. It is the externalisation of your thinking. How clearly you prompt reveals how clearly you think. Iterating with AI on a complex problem is, at its best, a form of structured self-reflection.
Cognitive evolution does not stop at the individual. It is also cultural. Organisations that evolve cognitively create environments where AI use is transparent, where outputs are questioned rather than accepted, where accountability remains human even when assistance is artificial. They move from the assisted level, fragmented and inconsistent AI use, through augmented, where AI is integrated into workflows with intention, toward AI-native, where hybrid intelligence is embedded in culture, strategy, and collective decision-making.
Clark and Chalmers argued in 1998 that cognition does not stop at the boundary of the brain, it extends into the tools and environments we inhabit (Clark & Chalmers, 1998). The question is not whether AI will become part of your cognitive infrastructure. It already is. The question is whether you are building that infrastructure intentionally or passively letting it build itself around you.
Movement 2: Design Evolution
Niels Quinten, Best Goed studio
Most organisations discover their AI is not working not in the boardroom, but in the field. A system that performed well in the pilot begins to fracture when it meets real users: the clinician under pressure, the analyst mid-context switch, the operator on the floor. The technology has not broken. The context has simply revealed what the design never accounted for.
The Context Gap: AI’s Most Common, and Least Discussed, Failure Mode
This is the context gap: the distance between how AI performs against idealised users in controlled conditions, and how it actually behaves when deployed into the grain of real work. This gap forms when it is assumed that a well-built tool is inherently well-suited to its users, that technical quality and human fit are the same thing. They are not. A junior analyst and a senior clinician may face the same interface, but they bring entirely different expectations, stress levels, and situational knowledge to it. If the AI knows nothing of that difference, it cannot serve either of them well.
From Tool-First to AI-Native: Three Design Orientations
Addressing the context gap means moving through three design orientations. Tool-first design optimises for capability: what the AI can do, largely indifferent to who uses it or when. Context-fit design shapes AI around actual usage, responsive to variability in users, roles, and conditions. AI-native design goes further: AI is no longer a tool inserted into a workflow, but a constitutive part of the system itself, evolving alongside the people and practices it supports. Most organisations are still somewhere between the first and second. The move toward AI-native is a shift in design philosophy, one that demands the same intentionality on the design side that cognitive evolution demands on the human side, and no amount of technical upgrading gets you there.
The philosopher Merleau-Ponty once argued that cognition is never detached from the world it operates in, it is embodied, contextual, and relational. Dourish (2001) observed this long before the current AI wave: people always interpret technology through existing social and cultural practices, never from a blank slate. That is why AI-native design is critical: AI that ignores people’s reality will be worked around or simply abandoned, no matter how capable it is. The question for leaders is not: have we deployed capable AI? It is: have we designed AI that can actually inhabit the organisations we have, with all their messiness, diversity, and human complexity?
3. The Hybrid Intelligence Loop
When Human Judgment and Machine Generation Meet
Henri Allegra, Homo Promptus
For 300,000 years, Homo sapiens has been defined by one competitive advantage: knowledge. The capacity to accumulate, organise, and transmit what we know. That advantage is now shared. AI knows more than any individual human, retrieves it faster, and almost never forgets. The question is no longer who knows most. It is who thinks best, and with what.
Hybrid intelligence only emerges when both movements happen simultaneously: when human judgment and machine generation enter into a fluid, iterative collaboration that produces outcomes neither could achieve alone. As early as 2019, Dellermann and colleagues argued for the deliberate development of socio-technological ensembles, systems that combine the complementary strengths of human and artificial intelligence to collectively achieve superior results (Dellermann et al., 2019/2021). Yet, human-AI teams consistently underperform their potential, not because the technology fails, but because the collaboration is poorly designed. Weak team cognition, misaligned trust, and absent collaboration architecture are the real culprits (Schmutz et al., 2024). Access to AI is not the bottleneck. Collaboration maturity is.
From Individual Augmentation to AI-Native Organisations
The organisations that will lead in a hybrid intelligence world are those that embed this capacity into their decision-making architecture, not as a policy layer added on top, but as the structural logic of how people and AI think together.
Westby and Riedl demonstrated that AI agents capable of modelling the mental states of human teammates can generate targeted interventions that improve the collective intelligence of the whole team beyond what any human member could achieve alone (Westby & Riedl, 2023). The implication is architectural, not managerial: when collaboration between humans and AI is deliberately designed, the collective becomes genuinely smarter.
Three Conditions of an AI-Native Organisation
An AI-native organisation, something fundamentally different from a digitalised one, is where AI is woven into the fabric of how people think and work together. It rests on three interdependent conditions:
1. Fluid, intentional, and responsible AI use: not as a technical layer added on top, but as a natural reflex applied with awareness of its limits and impacts.
2. Augmented human cognitive capabilities: where every person learns to engage AI as a thinking partner, not a substitute.
3. Collective hybrid intelligence: where humans and AI share a cognitive space in which information flows, structures itself, and improves continuously.
In an AI-native organisation, AI is not at the centre. The human always is.
From IQ and EQ to HQ: The Hybrid Quotient
Building this kind of organisation requires a new kind of intelligence. Not purely analytical. Not purely emotional. Something that combines both and transcends them.
Researchers are beginning to measure what they call the Artificial Intelligence Quotient (AIQ), an individual’s stable capacity to collaborate with AI across tasks (Qin et al., 2025). But collaborative intelligence is not only analytical. At Homo Promptus, we argue it also requires augmented emotional intelligence, the human capacity for consciousness, empathy, and intentional guidance of an intelligent system. Together, augmented IQ and augmented EQ form what we call the Hybrid Quotient (HQ): not a test score, but a new cognitive relationship between human and machine.
The Homo promptus is its expression, the cognitive evolution of Homo sapiens. Not the one who knows everything. Not the one who outperforms the machine. But the one who knows how to think in partnership with it, combining intuition and calculation, emotion and reason, creativity and rigour, while remaining irreducibly human.
Designing AI for Contextual Empathy
Niels Quinten, Best Goed studio
Knowledge alone is not intelligence. Real intelligence is embodied: it is in the clinician’s hands steadying under pressure, the operator’s instincts honed by repetition, the way a team’s rhythm turns individual actions into something greater. AI that ignores this is not only limited, but it will also simply feel out of place.
Contextual empathy is the design-side answer to the call for fluid, intentional, and responsible AI use. It is what happens when AI is not just built for humans but built to attend to them: to a person’s history, their level of expertise, the pressure they are under, and the environment they inhabit. A system that adjusts itself, its language, and its level of detail in relation to these realities does more than improve usability, it creates the conditions for genuine collaboration, not just assistance.
Symbiotic Co-Creation
Symbiotic co-creation is what happens when AI is designed to enable the Hybrid Quotient in practice. The AI surfaces possibilities the human might miss, while the human refines, challenges, and contextualises what the AI offers. Together, they produce outcomes that neither could achieve alone.
That is why at Best Goed studio we believe in design that does not just integrate human input but that empathises with the human context: understanding the rhythms of expertise, the weight of intent, and the unspoken needs of the moment. The result is a collaboration that feels instinctive, where the human leads not because the system steps back, but because the system is fully present, amplifying their judgment without overshadowing it.
4. What Gets in the Way?
Cognitive Laziness: The Delegation Default
Henri Allegra, Homo Promptus
The path of least resistance with AI is delegation. Ask, accept, move on. No reflection, no verification, no judgment. This is a cognitive reflex, not a character flaw, and it has a history.
In 2008, Nicholas Carr asked in The Atlantic whether Google was making us stupid (Carr, 2008). His argument was not about the search engine itself, but about what it was training us to do cognitively: skim rather than read, retrieve rather than reflect, bounce between links rather than think deeply. Nearly two decades later, that cognitive habit is hardwired. That reflex is now the default posture most people bring to AI. And it is precisely the wrong one.
Typing a keyword into Google was always a retrieval act. Collaborating with AI is a thinking act. A prompt is not a search query, it is the externalisation of your intent, your context, your judgment. A well-formatted prompt is still cognitive level zero if the thinking behind it is shallow.
An analysis of over 13,000 publicly shared ChatGPT conversations found an average of just 1.7 messages per session (WebFX, 2025). A September 2025 NBER working paper analysed approximately 1.1 million ChatGPT conversations and found that roughly half of all interactions are classified as “Asking”: seeking information or clarification, the same cognitive posture as a search query in a more sophisticated interface (Chatterji et al., 2025).
This is why at Homo Promptus we go beyond prompting. We have developed an approach based on learning to converse with AI, not just instruct it. The goal is not better prompts. It is deeper collaboration: one where the human remains the author of the thinking, not just the sender of the instruction.
Business Model Distortions
Behind most enterprise AI adoption sits a commercial logic that is rarely made explicit. The most influential actors in this space are not neutral enablers; they are major AI vendors whose primary interest is protecting and extending their existing business models: cloud services, productivity suites, enterprise software subscriptions. For these actors, AI is not a transformation agenda, but a retention strategy. The goal is to integrate AI into familiar tools and environments, enough to justify continued investment, not enough to fundamentally disrupt how their clients think and work. Cognitive disruption is their risk. Cognitive dependency is their product.
This logic produces a specific and largely invisible outcome: the proliferation of prompt libraries, pre-defined use case inventories, and curated AI functionalities, presented as enablement but functioning, in practice, as cognitive ceilings. When an organisation trains its people to select from a menu of pre-approved prompts, it is not building AI collaboration capability. It is replacing one form of passive consumption with another. Using AI with a predefined prompt is a retrieval act dressed as intelligence. Thinking with AI, iterating, questioning, co-creating, is a fundamentally different cognitive engagement. The first scales easily and keeps the vendor in control of the interaction architecture. The second requires human development that no platform can package and sell.
The Techno-Centric Trap
Niels Quinten, Best Goed studio
Most AI development begins with the technology: what the model can do, what features it offers, what capabilities it unlocks. That is a natural place to start. It is also, too often, where in terms of development the thinking stops. The techno-centric trap is not about bad intentions; it stems from a default orientation that places the system at the centre and asks people to organise themselves around it. Speed, power, and novelty become the primary metrics of success. Human needs, workflows, and judgment become secondary, things to be accommodated later, once the technology is built. The problem is that later rarely comes.
Designing for an Imagined Person
The people who build AI systems and the people who use them inhabit almost entirely different worlds, different incentives, different vocabularies, different definitions of what success looks like. Builders optimise for what they can measure: performance, accuracy, speed. Users navigate what they can feel: whether the system fits their workflow, whether it earns their trust, whether it helps them do what they are actually trying to do. When those who design AI have little sustained contact with those who will use it, assumptions fill the space. The user becomes a persona, an archetype, an average. And AI built for an imagined person will always struggle to serve a real one.
Smoothing over the Edges
By design, AI systems distil patterns, favour probabilities, and default to the mean. But the value of human-AI collaboration is not found in averages, it is in the friction, the contradictions, the stubbornly human ways a team thinks, debates, and stumbles toward something new. When AI smooths over these edges, it does not just standardise outputs, it risks eroding the very qualities that give an organisation its competitive edge. The way forward is not to tweak the model, but to design for the exceptions: to build systems that do not just tolerate human uniqueness but depend on it. Systems that treat tension as a feature, not a bug.
5. What It Means for Leaders
Henri Allegra, Homo Promptus and Niels Quinten, Best Goed studio
1. Stop measuring adoption. Start measuring collaboration maturity.
The question is not how many people use AI. It is how well your organisation thinks with it. Track not just usage, but how often AI outputs are questioned, refined, or overridden by informed human judgment. The shift: from Return on Investment to Return on Intention.
2. You cannot delegate a transformation that requires you to change first.
Cognitive and design evolution must be modelled from the top, not commissioned downward. Put yourself in the room as a participant in your organisation’s AI learning, not as a sponsor watching from a distance.
3. Stop deploying AI and start embedding it, AI-native design is your competitive edge.
Ask the question: have we designed this from the actual context of use, not just the ideal one? Embedding AI means building from real workflows, pressures, and quirks, not just testing for functionality, but for fit.
About the Authors
Henri Allegra
Henri Allegra is the founder and president of Etikord, a consultancy promoting responsible business models and practices, and the founder and Programme Lead of Homo Promptus, Etikord’s AI transformation programme for European SMEs. His work sits at the intersection of cognitive evolution, strategic transformation, and responsible AI. Henri works with leadership teams across Europe in English, French and Italian. Connect with Henri on LinkedIn.
Niels Quinten, PhD
Niels is the founder of Best Goed studio, a strategic design studio that helps companies evolve when AI starts doing what they do. It can be frightening to realise that the products and services you’ve built over years can be replaced overnight by AI. But disruption always creates new ground. Best Goed studio works with companies that want to find it, and believe the way forward is to bet on the human. Niels works at the intersection of embodied intelligence, human-centred design, and the practical realities of AI development. Connect with Niels on LinkedIn.
A co-authored article by Niels Quinten (Best Goed studio) and Henri Allegra (Homo Promptus)
We wrote this article because we believe the most important conversations about AI are not happening yet, not at the depth, and not in the right rooms.
If this piece made you think differently about what your organisation is actually building, or raised questions you do not yet have answers to, we would welcome the conversation. Reach out to either of us directly.
Niels & Henri
Key themes: hybrid intelligence, AI transformation, human-AI collaboration, cognitive evolution, collaboration maturity, augmentation vs delegation, AI-native organisation, human-centred AI design
