The Pressure in the Room

There is a moment I keep seeing across the organisations I work with. A leadership team has been handed a mandate, adopt AI, move faster, reduce cost, stay competitive, and they are sitting around a table trying to work out what that actually means in practice. Someone raises the obvious question: which jobs go first? Someone else pushes back. A third person opens a laptop and pulls up a list of tasks that a vendor has told them AI can now do. And the room starts sorting work into two piles: human, or AI.

That sorting exercise is not wrong. But it is incomplete. And the incompleteness costs organisations more than they realise, because the thing missing from those two piles is any clear principle for how to decide. The room is operating on instinct, on vendor claims, on anxiety, and on the memory of every previous technology rollout that promised transformation and delivered friction.

The HUMAND framework is the tool I developed to replace that instinct with a disciplined question. Not a replacement for the conversation, a structure for having it well.

The Binary Trap

The question most organisations are asking about AI right now is the wrong one.

“Will AI replace us?” produces two equally unhelpful answers. If the answer is yes, people disengage, protect territory, and spend energy on survival rather than preparation. If the answer is no, organisations breathe out too early, do too little, and find themselves behind when the preparation window closes.

Neither answer is useful for a leader who has to make real decisions this quarter about real people doing real work.

I explored this more fully in AI Isn’t Taking All the Jobs. It’s Rewriting the Task List. The argument there is simple: the threat is not that AI takes entire roles. It is that AI takes the tasks inside roles, and the roles left behind look different enough to require a fundamentally different conversation about what human contribution means.

That conversation cannot be had through the replacement lens. Replacement is a binary. It answers yes or no. What leaders actually need is a framework that answers: what combination, for which work, produces the best outcome for the people this work is meant to serve?

That is what HUMAND is.

Not prediction. Preparation.

What HUMAND Is, and What Each Element Contributes

HUMAND stands for Human, Understanding, Machine, AI, Network, and Decision. But it is not an acronym to memorise and display. It is a set of questions, one for each element, that, when worked through together, produce a coherent allocation of work across the forms of intelligence available to any organisation.

Here is what each element contributes to that process.

Human is the starting point, not an afterthought. Before any conversation about what machines or AI can do, the framework asks: what is the distinctly human contribution to this work? Not “what can humans do that AI can’t yet?” That is a temporary question, the answer changes every six months. The permanent question is: what does this work require that is rooted in human experience, human relationship, human judgement, and human accountability? What requires empathy, ethical navigation, or the kind of contextual wisdom that comes from years of being in rooms where things went wrong?

Understanding is the lens. Before any allocation decision, the framework insists on a genuine understanding of the work being examined. Not what the work is called. What it actually involves, the tasks, the relationships, the judgements, the moments of complexity that the job title does not reveal. This understanding step is where most AI adoption gets skipped in practice. Organisations automate the description of work rather than the reality of it.

Machine is the category of intelligence that executes repeatable processes with precision, speed, and consistency. Machines do not get tired. They do not lose focus. They do not bring emotional state into a process where emotional state is a liability. The HUMAND framework does not ask whether to use machines, it asks which work is genuinely machine-appropriate: structured, rule-governed, and measurable by output quality. Not everything that feels repetitive is actually machine work. Some repetition carries relationship. Some consistency requires discretion. The Understanding step identifies the difference.

AI is the category of intelligence that explores alternatives, surfaces probabilities, identifies patterns across data sets too large for human analysis, and generates options at a speed and scale no human team can match. AI does not decide. It prepares the conditions for better decisions. This distinction is one of the most important the HUMAND framework makes, and it is the one most commonly collapsed in practice. When organisations hand decisions to AI, they are not using AI well, they are using it in the space that belongs to the Decision element, and the consequences of that collapse appear downstream, often quietly.

Network is the connective tissue, the systems, relationships, and structures through which human and machine intelligence communicate, coordinate, and operate together. A well-designed network makes the allocation of work between humans and machines feel natural and navigable. A poorly designed one creates friction, confusion, and the kind of silent stall where transformation loses momentum without anyone being able to name why. Network is not a technical question alone. It is an organisational design question. Who communicates what to whom? Where do AI outputs enter human workflows, and at what point? Where are the handoffs, and who owns the accountability at each one?

Decision is where the framework lands, and where leadership is exercised. More on this shortly, because Decision deserves its own section. It is not the end of the HUMAND process. It is the point the whole framework is in service of.

How to Apply HUMAND in Practice

The framework does not work as a diagram on a wall. It works as a conversation, structured, facilitated, and grounded in the specific work your organisation actually does.

Start with a work function, not a role. Choose something specific: customer complaint resolution, procurement approval, performance review, content moderation, clinical triage. The more specific the function, the more useful the HUMAND conversation.

Work through the Understanding step first. Map what the function actually involves. Not the job description, the reality. What decisions does it require? Where does judgement enter? What relationships does it depend on? Where does it fail when the person doing it has a bad day, lacks context, or is operating under pressure? Where does it fail when a process is applied without discretion? This mapping exercise alone produces insights that most technology adoption conversations never surface.

Then bring in the Human question. Of everything this function involves, what is genuinely human? Not what humans currently do, that is just history. What requires human experience, ethical weight, relational trust, or accountability that cannot be delegated without consequence? This is often a smaller list than people expect, and a more important one.

Then ask the Machine question. Of what remains, what is genuinely machine-appropriate? Structured, rule-governed, measurable, and free of the contextual judgement that the Human step identified? This is often a larger list than organisations have acted on, not because they lack the tools, but because they have not done the Understanding step well enough to know where the boundary sits.

Then ask the AI question. Where in this function would surfacing patterns, generating options, or processing data at scale improve the quality of the human decisions that follow? AI does not take the work away from the function. It changes the conditions under which the human elements of the function are exercised.

Then ask the Network question. Given everything the previous steps have identified, how does the work need to flow? Where do outputs from AI enter human judgement, and how? Where do machine processes need human review, and at what threshold?

And then you arrive at Decision.

In The Workforce Revolution, I wrote about the shift from jobs as containers to work as something that flows across different kinds of intelligence. The HUMAND framework is the tool for designing that flow deliberately rather than inheriting it by default.

Why Decision Is the Anchor

The D in HUMAND is not a final step in a sequence. It is the point of the whole exercise.

Every other element of the framework is in service of one question: who decides, based on what, and with what accountability?

This is where leadership is exercised. Not delegated, not transferred, not handed to a system. Exercised. By people who carry the authority and the responsibility for the outcomes of the work their organisations do.

In the research behind Who Decides 2025, the pattern was consistent and illuminating: organisations are comfortable delegating routine decisions to AI. They are deeply uncomfortable, and often flat-out unwilling, to delegate decisions that involve money, people, or reputation. The question of where that line sits, and whether it is drawn deliberately or by instinct, is one of the most consequential strategic questions any leadership team can have.

HUMAND gives the Decision element its proper weight because it arrives at Decision having worked through everything else. The allocation of human, machine, and AI roles. The design of the network through which they interact. By the time the framework reaches Decision, the question is not abstract: it is grounded in a real understanding of what the work involves, who is responsible for it, and where judgement is irreplaceable.

I wrote about this specific tension in AI Is Not Replacing Judgement. It’s Exposing Where It’s Missing. The places where AI is failing us are almost always the places where human judgement was already weak. AI does not create the gap. It surfaces it. HUMAND is designed to make that gap visible before AI arrives in the system, not after.

What HUMAND Is Not

It is worth being direct about this, because the framework gets misread in practice.

HUMAND is not a checklist for automation. It does not produce a list of tasks to eliminate or a roadmap for reducing headcount. Organisations that come to it looking for cost justification will find the wrong tool. The framework is designed to produce better allocation decisions, not cheaper ones. Sometimes the conclusion from a HUMAND conversation is that more human attention is needed in a function, not less, because the Understanding step has revealed complexity that was being managed through volume rather than judgement.

HUMAND is not a way to avoid the hard conversation about what humans are for in the age of AI. That conversation is the framework. Every element of HUMAND is designed to bring that conversation into sharper focus, not to sidestep it.

HUMAND is not a technology framework. It does not require a particular system, platform, or vendor. It is a strategic thinking tool that can be applied in a boardroom with a whiteboard, in a workshop with a leadership team, or in a facilitated session with a workforce redesign group.

And HUMAND is not a one-time exercise. The allocation of work between humans and machines is not a fixed answer. It is a living question, because the capabilities of AI are changing, the nature of work is changing, and the expectations of the people doing the work are changing. Organisations that treat HUMAND as a project, something to complete and file, miss the point. It is a discipline.

The Ripple Effects of Getting This Wrong

When organisations skip the HUMAND conversation, the consequences do not appear all at once. They arrive gradually, in ways that are easy to misread.

A function is automated without the Understanding step. The process works technically but produces outputs that require more human correction than the original work did, because the edge cases, the exceptions, the context-dependent decisions, were never mapped. The time saved by automation is absorbed by the new task of managing its outputs.

A team’s roles are redesigned around AI capabilities without the Human question being asked. The work changes. The people in the roles feel their contribution narrowing. The judgement they brought, the relationships they managed, the context they held, these are not in the new role description, and the new role does not feel like theirs. Engagement quietly drains.

A governance structure delegates decisions to AI in domains where the organisation said it would not. Not through intention, but through gradual creep, each individual decision to let AI make the call seemed reasonable, but the accumulated pattern has moved the organisation to a place it never chose to be.

These are the ripple effects of skipping the disciplined allocation that HUMAND provides. They are not hypothetical. They are the patterns I keep seeing across the sectors and leadership teams I work with, in the organisations that adopted quickly and designed slowly.

The Preparation Window

The organisations that do the thinking early, before the pressure is undeniable, before the vendor is at the door, before the board has made urgency the frame for every decision, make better choices. Not because they are smarter. Because they have time.

Time to understand the work, not just react to it. Time to involve the people whose work is being redesigned, not just announce to them what has changed. Time to draw the Decision line deliberately, not in the heat of a competitive moment when instinct replaces judgement.

The preparation window for AI-informed work design is open now. The choices about how deep AI goes, where it governs, and what it replaces in the experience of human contribution have not yet been made at the level of intention and principle that HUMAND is designed to support.

Starting the HUMAND Conversation

If the HUMAND framework has a single entry point, it is this question: of everything your organisation does, where does it most matter that a human is responsible?

Not “where do humans currently work?” That is the present state, not the principle. And not “where can’t AI work?” That answer is shrinking by the month.

Where does it matter, to the people your organisation serves and to the culture your leadership team is responsible for, that a human carries the responsibility, the relationship, and the accountability for the outcome?

That question produces a list. And that list is the foundation on which everything else in the HUMAND framework is built. Machine and AI allocation flows from it, not the other way around. Network design follows from it. Decision authority is anchored in it.

This is what I mean when I say that HUMAND starts with Human. It is not sentimentality about the role of people in organisations. It is a strategic principle: design the human contribution deliberately, then build the system around it.

The signals are already here. The preparation window is open. The question of who decides, for which work, on what basis, with what accountability, is the strategic question of this decade.

Not prediction. Preparation.

Choose Forward.