Home / Research / Drift versus Design
Drift versus Design
Who is really making your decisions?
Your meetings are 30 minutes long. Nobody chose that. A calendar did.
Your sat nav took you a different way home last night. You did not question it. You arrived.
The report you signed this morning was drafted by a model, tidied by a template, and shaped by the way your predecessor always did it. You read it. You approved it. Did you write it?
Most of what happens in a working day was decided by something that is not in the room. That is not a scandal. It is how organisations work. But there is a difference between letting something decide for you because you chose to, and letting it decide for you because you never noticed. The first is design. The second is drift.
Drift feels like freedom. Design feels like effort. That is why drift wins by default, and why design has to be chosen.
What drift is, and what design is
Drift is when a decision keeps being made, and keeps being signed off, but nobody present is actually making it any more. The routine makes it. The default makes it. The model makes it. The person with their name on it has become a signature.
Design is when someone chose who, or what, should make that decision, knows why, and could change it. Design is not doing everything yourself. A good designer hands plenty of decisions to a checklist, a colleague or a machine. The difference is that they chose to, and they can say so.
So the question is never “did a human make this decision?” Plenty of decisions are better made by a system. The question is “did anyone choose to let it?”
That is the whole idea. Everything below is a way of seeing it in your own week.
Where are you on the map?
Two questions place anyone. Do you notice what is deciding for you? And could you change it if you wanted to?
- Sleepwalkers do not notice and could not change it. The week happens to them.
- Programmed people are effective and busy, and run entirely on scripts they have never examined. They are often the most senior people in the room.
- Stuck people see the drift clearly and cannot do anything about it. This is where most frustration lives.
- Designers notice, and act. Not on everything. On the decisions that matter.
Nobody lives in one square. Most of us are Designers at home and Programmed at work, or the reverse. The point is to know which square a given decision is in.
The seven ghosts
When nobody present is making a decision, something else is. We call these the ghosts: the absent authors of your working day. There are seven that turn up again and again.
| Ghost | What it sounds like | Where it hides |
|---|---|---|
| The Predecessor | “That’s how it’s always been done” | Routines, handovers, the way the last person set it up |
| The Peer | “Everyone does it this way” | Norms, what the team expects, what looks normal |
| Your Past Self | “I always do it like this” | Habits, the groove you wore years ago for reasons you have forgotten |
| The Policy | “We’re not allowed to” | Rules written for a situation that no longer exists |
| The Product | “That’s the setting” | Defaults, the 30-minute meeting, the template, the dropdown |
| The Proxy | “It’s what we measure” | Metrics and targets standing in for the thing you actually care about |
| The Predictor | “The system recommends” | Models, scores, AI drafts, anything that forecasts or generates |
The ghosts are not villains. Each one was useful once, and most still are. A ghost becomes a problem only when it is making a decision that somebody should be making, and nobody has noticed the swap.
The newest ghost is the Predictor. It is the one to watch. The others leave fingerprints. A rule is written down. A habit is yours. A model’s draft arrives in your own voice, polished, plausible and signed by you. It is the first ghost that is hard to see even when you are looking.
The Ghost Test
Pick one decision you made this week. Any decision. The meeting you booked, the email you sent, the number you approved. Ask four questions.
- Who, or what, actually made this? Name the ghost if there is one.
- Did anyone deliberately choose that?
- Could you change it? Not in theory. Tomorrow, if you wanted to.
- Would you choose it today? Knowing what you now know, would you hand this decision to the same author again?
If you can name the author and the answer to the other three is yes, it is designed, whoever or whatever is making it. A no on question two or four and it is drifting. A no on question three and you are Stuck, which is worth knowing too.
Most people who try this find the same thing. Not that their decisions are bad. That a surprising number of them have no author at all.
Try it on three decisions. Then try it on your team’s.
One layer deeper: the ghostwritten decision
For anyone who runs a team.
Every decision in an organisation has three owners. Someone has the authority to make it. Someone is accountable if it goes wrong. And someone, or something, actually authors it: does the thinking, forms the view, writes the words.
Organisations are very good at the first two. There are schedules for who may sign what and charts for who answers for what. Almost nobody tracks the third. Until recently it did not matter, because the person with the authority usually did the authoring. AI has pulled those apart. The decision is signed, it is answered for, and nobody present authored it. That is a ghostwritten decision.
You will know one when you see it. The slide deck nobody in the room could defend. The hiring shortlist that “the system” produced. The policy that was approved because it read well. Ask the author’s question of anyone who signs something:
Can you explain why this is right? What would make you change your mind? What did you change, or consciously accept, from what you were given?
If they can answer all three, they authored it, however it was drafted. If they cannot, a ghost did. Neither answer is a crime. But the second one needs to be known, because it is where the Verifier’s Discount lives: the moment a reviewer stops checking because the draft is fluent, and signs.
Two warnings for managers. Drift runs both ways. Handing a decision to a model nobody understands is drift. So is overriding a good model out of reflex, to look in charge. Design is choosing the right author and being able to say why. And it rises with quality: the better the draft, the less anyone reads it. The best tools create the most ghosts.
For boards: who chose the author?
For chairs, directors and executive committees.
At board level the question moves up one step. It is not “did a person author this decision?” For many decisions the right answer is no: a validated fraud model should classify transactions, and a reorder point should set stock. The board’s question is whether anyone authored the allocation: chose that author, on evidence, with a date to look again.
That is the layer most governance is missing. A delegation of authority schedule says who may sign. A risk register says who answers. Neither says who, or what, is forming the judgement, or whether anyone decided that it should. The failure is rarely that a machine made a decision. It is that nobody chose to let it.
Four questions settle whether an allocation is designed:
- Was the choice of author made by someone with the standing to make it?
- Was it made on evidence, rather than by default or by drift?
- Is it recorded, with a date to review it?
- Would the person accountable choose this author again today?
Some decisions should never be fully handed over, whatever the model’s accuracy, because the person affected is entitled to have their circumstances weighed by someone who answers for it. Terminations. Clinical calls. Safeguarding. Credit refusals with a face attached. We call those constitutive: the human judgement is part of what the decision is. The rest are instrumental, and for those the only question is which author does the job best, and whether the organisation could still do it unaided if it had to.
The practical step is small. Take the existing delegation of authority schedule and add two columns: who authors each decision class, and whether that allocation was chosen, by whom, on what evidence and when it is next reviewed. That is the authorship register. It is a morning’s work for a company secretary and it is usually the first time a board has seen, on one page, how much of the organisation’s judgement has quietly changed hands.
Across the organisation, the share of decision classes with an authored allocation is the Drift Index: the one number that tells a board how much of its judgement it still owns.
Where to go next
Start wherever you are.
- For you: run the Ghost Test on three decisions this week.
- For your team: ask the author’s question at the next sign-off. If it lands, I can run a bespoke Drift versus Design session for your workplace: keynotes and workshops.
- For your board: the Authorship Audit builds your register, scores your Drift Index and shows the board where its judgement has changed hands. Advisory from £5k: board advisory.
- For the argument in full: the working paper, The Ghostwritten Decision, sets out the theory, the evidence and the research programme behind this page. Read the research.
- Every week: I have tracked what emerging technology does to human judgement in Box of Amazing since 2017. Drift versus Design grew out of that watching.
I have spoken on this to teams at Google, Barclays and other organisations, and it sits at the centre of SuperSkills: The Seven Human Skills for the Age of AI (Kogan Page, 2026). The skills are how you become a designer. This page is what you are designing against.