The Ecology of Work in Digital and AI Environments

An essay on what a year of weekly monitoring has taught me —and on what my own framework was keeping me from seeing about the impact of the ways we execute in a digital era and the impact of AI.

A little over a year ago I designed a weekly radar (a scheduled task run with AI) to observe the future of work through five vectors: intrapreneurship, hybrid and distributed work, agility and self-managed organisations, diversity and inclusive governance, and artificial intelligence and new ways of working.

That was, and so far still is, my ecology of thought after more than a decade training executives in digital transformation, emerging technologies (AI among them), the value of data, the importance of culture and change management until trusted environments are achieved. In my judgement those five vectors sum up how execution works in the time of digital cholera.

Tracking trends across these five vectors was meant to help me understand transformation through observation over time. This longitudinal observation does not work headline by headline, nor video or clip by clip. It works on a slow burn, week by week, signal by signal, letting the patterns emerge instead of imposing them.

A year on, the most uncomfortable conclusion is not what the vectors have taught me.

It is what they have taught me about the framework itself.

The five vectors were born as independent categories. Today I know they are five windows onto the same building. And that the building is not a workplace practice. It is an organisational architecture: the operating system through which an organisation distributes intelligence, execution, autonomy and accountability among people, teams and —now— machines.

That is the thesis of this article. My intention is to defend it, stress-test it and, in the end, if warranted, perform surgery on that intellectual framework.

0. What each vector stopped being

The first thing a longitudinal reading reveals is that every vector has abandoned its founding question.

Intrapreneurship is no longer about culture. We started by asking how to get a large company to think like a startup, and we ended up understanding that the question is badly framed. A corporation can fill its offices with post-its, hackathons and entrepreneurial language without having cut a single approval cycle. The distinction that matters is a different one: a culture of intrapreneurship is not the same thing as an operating model built on entrepreneurship. The main obstacle to intrapreneurship is not a lack of creativity, it is the institutional architecture that slows experimentation down. The right question is no longer "how do we get our people to think like entrepreneurs?", but "what organisational constraints prevent a good team from behaving like an autonomous squad and a source of value creation?".

Hybrid work is no longer about location. It has gone through three phases: flexibility of place, sovereignty over time, and intentionality of presence. The return to the office that dominates the headlines does not invalidate hybrid; it matures it. The most solid empirical evidence —Nicholas Bloom's randomised experiment at Trip.com, published in Nature— shows that well-designed hybrid work not only fails to penalise productivity or team belonging, but cuts attrition by a third. The strongest argument in favour of the office is no longer individual productivity: it is social capital. The question stops being "how many days are you coming in?" and becomes "which activities get enough value out of co-presence to justify it?". We should call this presence with purpose. Today it was worth going into the office.

Agility is no longer about methodology. Perhaps the word "Agile" is losing importance precisely because its philosophy is gaining ground. Too many organisations practise Agile as ritual without having achieved agility: impeccable daily stand-ups and yet decisions that take months. A culture with agile methodologies may not be a flexible, adaptable culture. What matters to measure is the organisation's adaptation latency: how long it takes to detect a change, understand it, decide and modify its behaviour. How long does it take from a possible idea that might change the organisation, to resourcing it, measuring it with data, finding an internal customer to test it, scaling it and turning it into the norm?

Diversity is no longer (only) about representation. The most important conceptual shift of the year is the move from "who is represented?" towards "who takes part in designing the decisions?". When the criteria for "good employee", "high potential" or "acceptable risk" are encoded into AI systems designed by homogeneous populations, diversity stops being a people policy and becomes governance infrastructure. And bias stops being a binary property of the algorithm: it is a property of the complete socio-technical system —person + model + data + interface + incentives + authority + review procedure.

And AI is no longer about task automation. It is about redistribution. AI is the first work-organising technology that simultaneously changes who can execute, how much, from where, with what resources, under what supervision and with what degree of autonomy. That is why it is not just one more vector: it is the solvent melting the other four into a single conversation. Jobs? Old thinking.

1. The patterns that emerge when you connect the vectors

If we overlay the five time series, at least five patterns appear that no single vector shows on its own.

Pattern 1. Governed autonomy: the end of a century-old trade-off

For a century we assumed that autonomy and control were a zero-sum game. More autonomy meant less hierarchy and therefore less control. More control demanded more reporting, more committees, more friction.

AI partially breaks that equation because it radically increases the capacity to observe, coordinate and trace distributed work. A team can have more freedom to decide how to reach a goal while its results, permissions and risks are observable almost in real time. An agent can autonomously execute parts of a process within escalation rules and checkpoints that are 100% human.

The emerging organisational principle of the decade could be formulated like this: more local autonomy + more global observability. We have entered an era of governed autonomy (more autonomy, and more measurement, a counter-intuitive hybrid), and it is the concept that best connects the five vectors. The old organisation said "first we work, then the compliance team verifies". The agentic organisation will say "the rules of governance are part of the system we work with". Governance stops being a department and becomes part of the workflow. Humans work the magic, data and technology do their job simultaneously, trust emerges from governance.

Pattern 2. When execution gets cheaper, judgement gets more expensive

Generative AI and agentic AI reduce the minimum team size needed to put an idea to the test. Market research, prototyping, code, financial simulation, commercial materials: what once required seven departments is today handled by a small cell augmented by AI. This is what we can call the acceleration of the entrepreneurial machinery: fewer people, less capital and less time between hypothesis and evidence.

But beware the easy extrapolation of the "augmented solopreneur". When the cost of execution falls, the scarce resource becomes the quality of judgement: which problem to pick, which hypothesis to test, how to interpret ambiguous signals, what risk to accept. And judgement improves with cooperative practice grounded in a diversity of perspectives —and here the diversity vector walks in through the front door. We are not heading towards "one employee + a hundred agents". We are heading towards small, diverse human teams with disproportionately large productive capacity. It is my old formula operating at cell scale: f(Transformation) = max(People + Data + Technology)^Trust. Technology multiplies and trust is the exponent (it makes the final value grow disproportionately or reduces it to zero).

Pattern 3. Work separates from employment

For decades we treated "work" and "job" as synonyms. AI allows us to break work down at a granularity that makes that equivalence obsolete. A job is a bundle of activities: some get automated, others augmented, others demand more humans, others do not necessarily, and on top of that, new activities are born.

The most relevant unit of analysis is no longer the job: it is the chain Task → Workflow → Roles → Team → Organisation. Talking about "jobs destroyed" or "jobs created" is far too coarse. That precision error is overcome once the elementary unit of organisation stops being the job and becomes the workflow or the dynamic: people, models, agents, data, tools and checkpoints composing value. Microsoft calls it the Frontier Firm; McKinsey, the agentic organisation. I prefer to describe the phenomenon rather than the label: we move from hierarchies to agile teams, and evolve towards human-agent networks. The three layers will coexist. But the third one changes the rules of the other two.

Pattern 4. Literacy is not adoption; adoption is not transformation

The most underestimated pattern of the year. A person can know perfectly well how to use an AI assistant (copilot, GPT, Gem) or an agent and decide not to bring it into their work. Because they do not trust it. Because they fear for their employability. Because their objectives do not reward the time saved. Because their manager does not use it either. Because the old way of working still works well enough.

Fostering AI literacy does not necessarily imply AI adoption, and it guarantees the transformation of the organisation even less. The aggregate data confirms it bluntly: the vast majority of organisations experiment with AI and only a marginal fraction reports material impact on the bottom line. The gap is not technological. It is about work redesign and change management. And I would add a distinction that ideological diversity gave us as a gift: adoption does not necessarily translate into trust. Two people can use the same tool with opposing views on its consequences. Managing that tension is real change management. Using AI to read or write my emails, versus using AI to innovate at every step, in every decision, and to apply a creative filter that transforms the individual capacity to create value.

Pattern 5. The humanist paradox: synthetic abundance revalues the human

The cheaper routine intellectual execution becomes, the more valuable certain human capabilities become. Not in spite of machines producing language, analysis and code in abundance —quite the opposite, it is precisely because of that that human judgement is on the rise.

When content stops being scarce, up goes the price of judgement, intention, reputation, trust, interpretation and accountability. It is what my friend Gam Dias calls the Human Umami: that fifth flavour that cannot be synthesised. That secret sauce that defines augmented intelligence and that connects with hybrid: if AI absorbs the preparation and the administration, physical presence can concentrate on what the machine does not replace well —negotiation, mentoring, trust, shared creation of meaning. AI and the office are not opposing forces. Well designed, they are complementary.

2. Challenging my own architecture: are there vectors missing from this ecology of work?

Here comes the self-criticism, which is where a framework proves whether it works or merely decorates.

My five-vector architecture has three weaknesses that a year of evidence has made visible.

First: the five vectors are no longer orthogonal. They were born as parallel categories and today they are projections of the same phenomenon. Keeping them separate has value when you are researching —it lets you watch for signals— but it has stopped having explanatory value per se. The framework needs a higher level that integrates them: the architecture for distributing intelligence, autonomy, execution and accountability.

Second: a vector is missing, and it is not technological. It is trust and the governance of the socio-technical system. In my formula trust was always the exponent; on the radar, however, it appeared diluted inside each vector. A mistake. The year's evidence shows that governance —traceability of authentic systems, allocation of decision rights, observability, auditability, escalation, mandates and permissions, as well as shared human-machine accountability— is the terrain on which everything else is decided.

I propose promoting it to Vector 6: Socio-technical Trust and Governance (or, if we are faithful to the formula, Vector 0: the exponent that multiplies or cancels out the others). Without it, governed autonomy is an oxymoron and AI adoption will stall in the organisation's festival of pilots.

Third: there is an emerging candidate I am not promoting to vector yet, but watching closely: the sustainability of human energy. The infinite working day documented by telemetry data —fragmented work, meetings colonising the night, pulverised attention— suggests that the first consequence of the abundance of synthetic intelligence has not been to free human time, but to densify it. Sometimes grinding down the capacity for discernment through sheer cognitive overload caused by the continued use of AI. We need more time to process the results AI hands us. If execution accelerates and judgement gets more expensive, protecting people's cognitive and emotional capacity stops being cosmetic wellbeing for talent and becomes a capacity strategy. A strategic pillar for talent teams to protect. I still treat it as a cross-cutting dimension. In twelve months I may have to apologise and promote it to vector. Or is it just one more skill?

And what about skills? We always have the option of raising skills to a new vector to observe —could that be the seventh vector? I am not convinced. And that is despite having written at length about critical thinking. I resist including skills in the equation as one more vector. They are the bloodstream of the "person" system. Two thirds of the competencies organisations will need in five years will be different from today's, but that is not a trend running parallel to the others, it is the consequence of all of them. Turning them into a vector would be confusing the symptom with the force. The energy with the architecture. The degradation of these skills will be a trend and a vector that redefines the ecology of work, and it is still early.

In short, if I had to condense the new ecology of work into an operational definition:

The ecology of work is the system through which an organisation distributes —consciously or unconsciously— intelligence, execution, autonomy and accountability among individuals, teams, agents and structures, under a given regime of trust and observability.

I say "consciously or unconsciously" because that is the real dividing line. Every organisation already has an ecology of work. The competitive difference lies between those who design it and those who inherit it. Between being and wanting to be. Five properties define the ecology of work today:

1. It is distributed: the elementary unit is the workflow, not the job.

2. It is hybrid in three senses: place (on-site/remote), time (synchronous/asynchronous) and the nature of the executor (human/agent).

3. It is governed in the flow: the rules live inside the work system, not in a separate manual.

4. It is diverse by design or biased by default: whoever does not govern the composition of their decision systems delegates it to their historical biases.

5. It rests on measurable trust: adoption without trust is compliance; trust without observability is faith. Trust needs to be measured and adoption must be born of individual conviction. Only then does adoption translate into trust.

3. Strategic implications for the C-suite: A New Ecology of Execution

For the CEO, the implication is one of identity: redesigning the operating system cannot be delegated. The evidence is stubborn: most leaders admit their organisation is not ready for the changes coming, and agentic AI transformation fails when it is treated as a technology programme with an executive sponsor instead of as a redefinition of the operating model led from the top. We stop working in a world where humans designed processes for humans, and move to coexisting in a world where humans and machines (augmented humans) design processes for machines plus humans (the reality of AI agents).

The question the CEO must institutionalise is not "how much AI are we using?", but "how fast do we absorb a good idea and turn it into operational behaviour?". The capacity to absorb change is the new competitive advantage. A mix of governed autonomy (more autonomy, and more measurement, a counter-intuitive hybrid), plus acceleration of the entrepreneurial machinery (autonomous, augmented teams, less intensive in technical and human resources, that speed up the cycle from problem to prototype and later scaling).

For the CHRO, the implication is one of refoundation. If work separates from employment, the entire HR toolkit —job descriptions, performance, workforce planning, compensation, span of control— is built on a unit of analysis that is going extinct. The CHRO of the future manages portfolios of tasks, human-agent workflows, and perhaps skill sets, not inventories of jobs. And they inherit two new mandates: measuring the equity of flexibility in hard outcomes (promotion, access to projects, compensation —not just satisfaction, because proximity bias does not show up in engagement surveys) and developing an emerging managerial competency: the Capacity to Delegate to Human-Machine Systems: delegating to intelligent systems without micromanagement or blind faith, designing their indicator systems or their own translation of governed autonomy into their sphere of responsibility. This new managerial competency will be key to managing the transition from hierarchies to agile teams, and on to human-agent networks.

For the CSO (Strategy), the implication is one of clock speed. The acceleration of the entrepreneurial machinery means the cost of testing a strategic hypothesis has collapsed; strategy stops being an annual document and becomes a living portfolio of governed experiments. Their critical metric —the key KPI for strategy leaders— becomes the organisation's latency in adapting to change. How to measure adaptation from hypothetical scenarios to demanded changes, sponsorship of pilots and scaling, through to the redefinition of regulatory compliance. And their new responsibility will be deciding which capabilities are built in-house and which are bought, because in a world of intelligence on demand, that frontier will define the company of the future. They will unconsciously catalyse the transition from an organisation based on full-time employees to a new swarm of professionals marketing their skills on platforms, temporarily and by project (Augmented Intelligence on demand). And that is part of another story I do not want to develop right now.

For the CIO, the implication is one of role: from systems provider to architect of work, or perhaps architect of the governance model for agent networks. The critical infrastructure is no longer the ERP: it is the human-agent orchestration layer —identity, permissions, observability, escalation, audit. The data shows that the best-performing companies have their technology leaders deeply involved in corporate strategy, not executing it after the fact. The CIO who carries on managing ticket demand (an artefact built with generative AI can already do that) will be replaced by the one who governs agent networks.

For the CDO (Digital and Data), the implication is one of embodied governance. If diversity is decision infrastructure and bias is a property of the socio-technical system, the CDO becomes the guarantor of the informational base of the Human-Machine Decision Architecture: which data feeds which models, with what supervision, evaluated on which populations, with what rights of appeal. Their dashboard stops measuring and guaranteeing only data quality and starts measuring decision quality and how governed autonomy translates into a new informational order. Another reflection is whether this CDO responsibility is a mandate delegated from the CEO's governed autonomy, or whether it is instead a responsibility shared with the CIO as the new architect of the governance model for agent networks.

And a cross-cutting warning for all five: the agenda is not divided up, it is shared. Governed autonomy dies the moment HR manages people, IT manages agents and nobody manages the workflow where both coexist.

The primary challenge of this new ecology of work in a digital AI environment lies in reconfiguring the architecture between human beings and machines. Such a redefinition depends on accelerating the entrepreneurial machinery iteratively, while at the same time cultivating the skill of delegating to hybrid human-machine systems. The ultimate goal is none other than continuously reducing the organisation's adaptive latency in the face of an environment where the rate of change has become exponential, and where organisational flexibility will depend on correctly defining an architecture of agent networks optimised by proprietary data and information, and augmented humans.

The future of work is not being determined by where we work nor by any specific technology. It is being determined by a progressive redistribution of autonomy, intelligence and execution capacity among individuals, teams, organisations and machines coexisting daily, on a minute-by-minute and second-by-second basis.

Intrapreneurship shows that small units can acquire extraordinary entrepreneurial capacity. Hybrid shows that coordination no longer demands permanent coincidence in time and space. Agility shows that adaptive capacity matters more than liturgy. Diversity reminds us that the quality of a decision depends on which perspectives take part in building it. And AI connects everything because it simultaneously redistributes execution, supervision and autonomy.

The central question, therefore, is no longer how to bring tools into work. The challenge now is closest to the art of redefining and optimising the sequence: Task, Workflow, Roles, Teams to create new Organisations. In other words, the main question now lies in how to design organisations capable of consciously distributing intelligence, autonomy, execution and accountability between people and machines —with trust as the exponent.

Because every organisation will have an ecology of work on a digital mat augmented by AI.

The only question is whether it will be designed or inherited.