Outcome Thinking is in Your Blood

Our bodies have already solved the problem every organization is now charged to solve in the agentic workplace.

Every cell in your body needs oxygen, and no two cells need the same amount at the same moment. A muscle at work burns through its supply. The same muscle at rest barely touches it. Through a physiological process that rivals magic, your blood deposits oxygen where it is needed, in the amount it is needed, and does not pool uselessly where it is not. There is no dispatcher. No manager of the oxygen. No executive blood function deciding which tissue gets what. It just happens, in the moment, with each beat of your heart.

Science understands this process better than we need to for the purpose of the analogy. In short, chemical changes in the body act as a demand signal triggering the molecule called hemoglobin to release oxygen. The body’s oxygen demand is instantaneously communicated to hemoglobin to provide the necessary supply. If the demand is not communicated, oxygen stays bound and travels on through the blood stream to where it’s needed.

This amazingly efficient physiological logistical system has a property most organizations only dream about. It delivers exactly what is needed and nothing where it is not. Every unit of oxygen delivered is a unit demanded. There is almost no waste, because supply is pulled by demand rather than pushed by a plan. Toyota’s Taiichi Ohno understood this pull system, but could at best approximate it.

AI changes everything.

The push economy

Planners forecast demand, staff to the forecast, and build standing capacity that runs whether or not the work is there. Positions are filled, held and released. Processes run on a schedule. Teams stand ready. Reports are produced every Tuesday, not because anyone is waiting for them, but because that’s the schedule. Resources are committed in advance and expended regardless, and the gap between what was pushed and what was actually demanded is mostly unavoidable waste: idle capacity, work no one needed, effort spent keeping the machine warm.

We have known this for a century. Pull is not a new idea. Lean manufacturing chased it. Just-in-time chased it. But those were heroic efforts confined to a few heavily engineered supply chains, because sensing real demand and moving resource to meet it, everywhere, in real time, was simply too expensive or impossible. The coordination cost was the wall. So organizations did the rational thing. They pre-positioned resource in fixed structures and accepted the waste as the cost of staying in business. The standing team, the permanent position, the batch, the quarterly cycle: every one of them is a hedge against the cost of knowing what is needed and responding fast enough to matter.

That wall is what has now come down.

Why the wall came down

AI collapses the cost of sensing and the cost of coordination. Demand can be detected as it forms. Capability can be summoned and composed on the spot. The thing that forced the push, the expense of knowing and then responding, falls toward zero. For the first time an organization can, at least in principle, run the way the body runs: resource pulled by demand, spent only where it is called for, conserved everywhere else. Not in one engineered supply chain, but across ordinary work, including the knowledge work that lean never truly reached.

This is the business case, and it is larger than the one usually made for AI. The usual case is that AI does the existing work faster and cheaper. That is real, but it is the floor. The larger case is that AI lets an organization stop doing the work that was never demanded in the first place and do the valued work that was not possible. The prize is not a cheaper push. It is the end of the push.

There is a catch, and it is the reason most organizations will not capture this. You cannot pull toward a demand you cannot see, and most organizations have made their demand invisible.

They have done it by describing their work as activities.

An activity is a push by nature. Run the report. Process the invoices. Answer the tickets. Each one presumes the work should happen, and happen on schedule, whether or not anyone downstream is waiting for it. An activity cannot emit a demand signal, because it is not a state someone needs. It is a thing someone does.

The Call for Outcome Centricity

The alternative is to describe work as an outcome: not the doing, but the state you want to be true. Here is where most attempts go wrong, in a way so subtle it requires describing. People write down activities and believe they have written outcomes, because they have dressed the verb as an adjective. The invoices are processed. The controls are checked. The candidates are identified. These sound like states. They are not. “Are processed” still presumes someone is processing. The activity is still there, hiding inside the participle.

Here is a test. Anthropic is one of a very small number of places on earth the best AI engineers most want to work. Write that as an outcome for the talent acquisition function and you might write: the best engineers are identified and recruited. But that presumes an activity, a search, a pipeline, someone doing the finding. The truth is simpler and stranger. The best engineers are not identified. They line up at the door. The state, the best engineers are here, is held by something Anthropic is, not by anything it does. The activity of finding is simply not required.

The best engineers are not identified. They line up at the door.

Call it the test for a real outcome. Does the statement describe a condition that simply is, or does it smuggle in the act that produces it? The best engineers are here passes. The best engineers are identified fails. A true outcome names a state of being. It says nothing about the activity that holds it true, and often the best way to hold it true involves no such activity at all. This statement bears repeating, because it is at the heart of Theory of Constraints. Sometimes, the best activity to perform is no activity at all.

This is not merely a word game. It is the difference between automation and something far more valuable. An outcome with an activity hidden inside it can be satisfied only one way, by doing the activity, faster or cheaper or by machine. That is automation, and automation is the floor. A true outcome, a pure state of being, can be satisfied by anything that makes it true, including arrangements in which the old activity no longer exists. Every invoice is paid on time can be answered only by paying invoices faster. The company owes nothing past its terms can also be answered by settling at the moment of agreement, so there is no invoice to pay and no paying to do. The first statement caps your ambition at doing the same thing quicker. The second opens the door to not doing it at all.

And notice what the true outcome does that the activity never could. It states a condition, and a condition can be true or false, on track or at risk. It can throw off a signal the moment it drifts. It becomes the body’s chemical signal: the demand that calls the resource to itself. Describe your work as states of being, and each one becomes a sensor for its own need. Describe it as activities, and you are back to pushing on a schedule, blind.

In your blood

So the sequence runs one way. Outcomes written as states of being make demand visible. Visible demand can be pulled toward. AI makes the pulling cheap. And an organization that spends only where it is demanded is, like hemoglobin, close to pure value and almost no waste.

The instinct to organize this way is not exotic. It is not even new. It has been keeping you alive your whole life, quietly, in every vessel you have. Outcome thinking is not a technique to be adopted. It is a design your own body settled long ago.

The only question is why your organization still runs on a plan, when your blood never has.