The Amnesiac Enterprise

The contract runs to forty pages, and the delivery lead reads all of them twice before she finds the sentence that will define her year. It sits halfway down page nineteen, buried in a schedule of service levels, and it says that the client may request unlimited revisions to any deliverable within the sprint in which it was produced, at no additional charge. She reads it a third time. Then she goes looking for the person who agreed to it.

He is not there. He closed the deal eleven weeks ago and has since been promoted to a larger territory, or moved to a competitor, or is simply on a beach he has earned, unreachable. What she wants from him is not the clause. She has the clause. She wants the reason. Why unlimited revisions? Did the client demand it, or was it offered? Was it traded for something (a longer term, a faster payment schedule, a reference call), or was it given away in the last hour before quarter close to get ink on a page? Was it a blunder, or was it the price of the whole deal?

None of this is written anywhere. The clause is in the contract. The reasoning is in one man's head, and the man is gone. What she has inherited is a promise she must now honour without knowing what it cost or what it bought. For the next twelve months, every time the client asks for another revision, she will absorb the work, and she will never be certain whether she is being taken advantage of or simply delivering exactly what her own firm sold.

This is not a story about a bad clause. It is a story about a firm that cannot remember why it decided what it decided, told once, in miniature. Multiply it by every deal, every quarter, and every seam in the business, and you have the condition this chapter is about.

The turn

In the last chapter we walked the modern software estate and found the same hole in every system. The CRM records the pipeline, the ERP records the transactions, the HRIS records the people, and not one of them records the decision. We called this the decisionless stack: the difference between the record and the decision. The record is what happened. The decision is why it happened, and what else was considered, and by whom, and at what cost. Software captures the first with great fidelity and loses the second completely.

If the decision is never captured, the enterprise cannot remember it. And an organisation that cannot remember its decisions is condemned to make them again. Not the same decisions sharpened by experience, which is wisdom. The same decisions from a standing start, with the experience thrown in the bin, which is amnesia. The firm relearns, at full price, what it already knew and forgot. It does this constantly, in every corner of the business, and it does not even feel the loss, because you cannot grieve a memory you never knew you had.

The question this chapter answers is short and expensive. What does it cost a business to forget?

The contract without the reasoning

Begin at the handover, because that is where forgetting does its clearest damage.

When a salesperson wins a deal, they are holding far more than a signed document. They are holding a rich, private model of the whole thing. They know that the procurement lead did not care about price but cared enormously about the transition being invisible to her own boss. They know the price landed where it did because a competitor had quoted lower and the client needed a reason, any reason, to justify paying more. They know where the risk is buried, because they buried some of it themselves.

None of that model is in the contract. The contract is the residue of the decision, not the decision. What crosses the seam from selling to delivering is the residue: the scope, the price, the dates, the clauses. The model that produced them stays behind in the salesperson's head.

The contract is the residue of the decision, not the decision. What crosses the seam is the residue. The reasoning stays behind in one person's head, and that head has moved on.

So the delivery team does the only thing it can. It reconstructs. It reads the unlimited-revisions clause and, lacking any record of why it was agreed, either assumes the worst and staffs defensively against a risk it does not understand, or assumes the best and gets ambushed in month three. Either way it is spending real money re-deriving a judgement that another employee of the same firm made in full, with better information, eleven weeks ago. The firm is paying twice for one piece of thinking, and it has the second payment on no ledger.

Even the firms that take the handover seriously feel this. They hold a proper meeting, both sides in the room, and the salesperson walks the delivery team through the account. It helps, and it is not enough, because a handover meeting transfers facts fluently and reasoning badly. The salesperson conveys what is true about the client and forgets to convey what they decided and why, partly because their own reasoning has worn smooth into a set of givens, invisible even to them. An hour later the delivery team has better notes and the same core problem: they hold the conclusions and not the thinking, and the thinking is what they will need the first time reality departs from the plan.

This is not anyone's fault, which is exactly why it persists. The salesperson is not withholding the reasoning; they are simply not asked for it, not paid for it, and not equipped with anywhere to put it. Their incentive ends at signature, and the handover, where it exists, is treated as a transfer of documents, not decisions. The folder moves. The thinking that gives the folder meaning does not, because no system was ever built to hold it.

After the deal is won

Look at where a firm spends its attention and almost all of it points at the moment before signature. Everything up to the win is instrumented: the pipeline has stages, the forecast has a rhythm, there are qualification frameworks and review meetings and a small ceremony when a deal closes, a message in a channel.

And then, at the instant of the win, the instruments go dark. The gong sounds, and the deal drops off the edge of the observed world into the place where the work actually happens. What comes after the deal is won is the least watched, least recorded stretch of the entire business, and it is the stretch where the money is genuinely made or lost, because the margin does not live in the sale. It lives in the distance between what was sold and what it costs to deliver, and that distance is travelled in the dark.

This is the great blind spot, and it is structural: the seam between the promise made to win the work and the work that must honour it. On one side sits a set of decisions dressed as commitments, this price, this scope, this timeline, this assumption about how many people it will take and how senior they must be. On the other sits a delivery organisation that must make those commitments true but never sees them as decisions. It sees them as constraints that arrived from nowhere, because the reasoning behind them was never captured. The half that promises and the half that performs are run as separate worlds joined by a document.

Unpack a single promise and you can count the decisions folded inside it. A price is a decision about how many hours the work will take and how expensive those hours must be; a timeline, a decision about how many people can be pulled onto it and how fast; a scope boundary, a hundred small judgements about what falls inside the line. In a people business these are the entire economics of the engagement, set during the sale by someone whose job was to win it. Delivery receives them as fixed quantities and cannot tell a deliberate stretch from a careless one: whether the margin is thin because the client was strategic and worth it, or thin because someone guessed the effort wrong. Both look identical from the delivery side. A hard number, no reason, honour it.

You can watch the effect in the language people use. Sales talks about what it won. Delivery talks about what it was handed. The passive voice is the tell. When a delivery director says the account was priced badly, or the scope was never realistic, they are describing decisions made by their own colleagues as though they were weather. They cannot argue with the reasoning because they cannot see the reasoning. They can only live with the result and quietly resent the seam.

And the forgetting runs both ways, which is the proof that this is not a sales problem but a memory problem. Delivery learns things every day that sales never receives. The team on the ground discovers that this client will never buy the premium tier, whatever the account plan says; that the real expansion is not the German office everyone is chasing but the Italian one nobody is. Each of these is hard-won intelligence, worth more than most of what the CRM holds, and each of them dies inside the delivery team, because there is no channel and no record to carry it back to the people who will price and pitch the next deal. Sales forgets what it decided and delivery forgets what it learned, and the two are the same failure: a firm with no memory for the judgement that moves between its halves.

Memory that walks out of the door

Ask a firm where its institutional memory lives, and if it answers honestly, it will point not at a system but at a person. It lives in the account manager who has run the client for six years and simply knows why things are the way they are. It lives in the founder, who carries an unwritten map of every important decision the company has ever made and is the reason so many meetings end with someone saying, let us just ask her.

This feels like strength. A firm full of experienced people who carry the history in their heads has a kind of deep knowledge that no document could match. But knowledge that lives only in people is not an asset the firm owns. It is an asset the firm rents, from individuals, on a lease that can be cancelled without notice.

Every resignation is a data deletion with no backup. When the six-year account manager leaves, the client relationship does not transfer to her successor. The record of the account transfers: the contract, the invoices, the ticket history. The memory does not. Why the firm gave this client a break on rates two years ago, which of their people are difficult and which are decisive, what was promised verbally in a meeting that no minutes captured, which past decision must never be reopened because it nearly ended the relationship, all of it walks out of the building in her head and does not come back. Her successor begins from a signed contract and a warm handshake, and relearns over the next eighteen months, through her own expensive mistakes, what her predecessor already knew.

Reorganisations do the same damage more quietly. Nobody leaves, but the memory is shuffled into new boxes where it can no longer be found. The person who knew why the pricing model worked that way is now running a different region and is no longer consulted. The team that carried the lesson is broken up and its members distributed, and the lesson, which lived in the space between them, in the shared understanding rather than any single head, simply ceases to exist.

The cruelty of it is that nobody signs off on the loss. A firm that deletes a database has an incident. Someone is accountable, there is a review, the backups are checked, questions are asked about how it was allowed to happen. A firm that loses a decade of pricing judgement to a departure or a reorganisation has a leaving party and a note about backfilling the role. The two events destroy comparable value. Only one is treated as a failure of stewardship, because only one is visible, and it is visible only because the database sat in a system and the judgement never did.

The compound interest of forgetting

The cost of amnesia would be tolerable if it were paid once. It is not. It compounds, and you can feel it in three patterns that repeat in every firm that cannot remember its decisions.

The first is the argument you have already had. A genuine question gets raised, debated properly, and settled: should we take fixed-price work in this category, given how badly the last one went? The room reaches a good answer. Then the answer is never recorded as a decision, only as an outcome that happens to hold for a while. Six months later a new hire, or a new head of sales, or simply a different set of people in a different room, raises the same question in good faith, and the whole debate runs again from the beginning. The same evidence is dug up. The same conclusion is reached, or worse, a different one, because the reasoning that produced the first answer was never available to challenge. The expensive part was never the conclusion. It was the thinking, and the firm pays for it again every time, because it kept no receipt. The second is the concession that failed before. A client asks for something: an extended payment term, a discount tied to volume that never materialises, a clause that quietly transfers risk. The firm tried exactly this once, with another client, and learnt the hard way that it was a loss. But that lesson was never captured as a decision with a reason attached. It dissolved into the general fog of things that went badly. So the concession is granted again, by someone who was not there the first time, and it fails again, in the same way, for the same reason. The firm does not recognise the repetition, because to recognise it you would need a memory of the first time framed as a decision, and there is none. It files the second failure under bad luck and prepares, unknowingly, to grant the concession a third time.

The third is the mistake that scales. One account is underpriced, not by accident but because of a specific wrong assumption: that a certain kind of work would take a certain amount of effort, when in truth it takes far more. The reasoning behind the price was never written down, so nobody can point to the flawed assumption and correct it. But the price itself lives on, in the CRM, in the templates, in the sense of what this kind of deal goes for. When the next three accounts of the same shape come along, the same assumption is applied, silently, three more times. The firm has taken its single most expensive error and, precisely because it could not see the reasoning inside it, replicated it across the portfolio. It scaled its mistake faster than it scaled anything it did well.

A firm that cannot remember its decisions does not simply fail to learn. It industrialises its errors, applying yesterday's flawed reasoning to tomorrow's accounts at speed, because the reasoning was never visible enough to be caught.

Put the three together and you see why forgetting is not a fixed cost but an interest payment. The re-derivation tax is invisible on any statement, because it hides inside things that look like normal work: another strategy discussion, another discount, another account that came in under margin. But it grows with the firm, because a bigger firm has more seams across which the same forgetting can happen at once.

Why the software forgets

Here is the plainest fact in this chapter. Software should remember decisions, and almost none of it does.

We know how to build memory. We built it for code. When an engineer changes a system, a version control tool captures not just the change but a note explaining why the change was made, attached permanently to the change itself, so that years later anyone can read the history of the reasoning. The capture is a by-product of the work; you cannot really do the work without it. We built memory for money, too. The ledger, and double-entry before it, is a system whose entire purpose is to remember every movement of value in a form that survives the people who recorded it. These are systems of record for a verb: for changing, for transacting. They hold the act, not just the artefact. Engineers would never accept a system in which nobody could explain, six months later, why a change was made. Executives accept exactly that for million-pound commercial decisions every day.

We never built the equivalent for deciding. Enterprise software was architected around nouns, around the customer and the transaction and the employee and the ticket, and a decision is a verb. But the deeper reason is that capturing a decision means capturing it at the moment it is made, and the moment a decision is made is exactly the moment nobody wants to stop and write anything down. The choice happens in a meeting, on a call, in a corridor, in the pressured minutes before a deadline. It is felt as an interruption to ask for the reasoning right then. And a decision documented later is not the decision. It is a reconstruction, cleaned up, missing the alternatives that were rejected and the discomfort that was overruled, which is to say missing everything that made it a decision rather than a foregone conclusion.

Notice what those two working systems have in common. Neither of them relies on anyone finding the time to write things up well after the fact. The capture is fused to the act, which makes it cheap, timely, and complete, and it does not have to compete with the real work for attention because it is part of the real work. That fusion is the whole trick, and it is the thing decisions lack. There is no natural moment in the making of a decision at which recording the reasoning is the obvious next keystroke, so the reasoning never gets recorded, and from the silence we conclude, wrongly, that decisions are simply too fluid to capture.

The systems in the estate happily record the results of decisions: the price that was set, the person who was staffed, the ticket that was escalated. Every one of those records is a decision with its reasoning surgically removed. The stack is not neutral about memory. It actively strips the why out and keeps the what, and it does this so smoothly that the firm mistakes the surviving records for the memory itself. A firm can have perfect records and total amnesia at the same time, and most do.

The counterpoint: but we have a wiki

The obvious objection is that this is a solved problem, and the solution is discipline. Good firms document. They keep a wiki, a knowledge base, a shared drive of playbooks. They have handover templates and account plans and a folder called, hopefully, Institutional Knowledge. If the enterprise forgets, the answer is surely to write things down properly and to enforce the habit.

It does not save them for four reasons.

First, documentation records conclusions, not live reasoning. By the time anyone writes the wiki page, they are writing the sanitised version: we decided to price it this way. The alternatives that were weighed and rejected, the tradeoff that was accepted, the person who disagreed and why, the specific fear that drove the concession, all of it is compressed out. But that compressed-out material is the reasoning, and the reasoning is the only part with any power to inform the next decision.

Second, it is written after the fact, by whoever has time, which usually means never, or thinly, or by someone who was not in the room. Documentation is a second job stacked on top of the first, and it loses every contest for attention with the actual work. The decisions that most need capturing, the fraught ones made under pressure, are precisely the ones nobody has the bandwidth to write up afterwards. So the wiki fills with the easy, low-stakes knowledge and stays silent on the expensive stuff.

Third, and most fatally, it rots. A wiki page is true on the day it is written and begins decaying immediately. The context shifts, the client changes, the pricing model moves, the assumption is quietly superseded, and the page does not know. It sits there, confidently asserting something that stopped being true months ago. Readers learn that the knowledge base cannot be trusted, that half of what it says is stale, and once a source cannot be trusted it stops being read, and once it stops being read it stops being updated, and the decay accelerates. A stale document is worse than an empty one, because an empty one at least does not lie to you with confidence.

Fourth, the page is not connected to anything. It sits in the wiki while the decision it describes lives somewhere else entirely: in a contract, a staffing plan, a forecast, an invoice. Nothing links the reasoning to the thing it produced. Even when the page is written, current, and true, the person facing the live decision does not know it exists, cannot find it in the moment they need it, and would have to already know the answer in order to search for it.

Add the four together and you see why the wiki was never going to work. What a firm needs is not a better library of pages written after the fact and hoped over. It needs a system of record for decisions, in which capture is a by-product of deciding rather than a chore that follows it. The model is the commit message, not the encyclopedia: the reasoning recorded in the flow of the work, at the moment of the choice, attached to the thing decided, carried forward with it so that whoever inherits the decision inherits the thinking too. Documentation is what you write when your systems cannot remember. The goal is systems that can.

The bridge

If the diagnosis so far is fragmentation and forgetting, the industry does have an answer ready, and it is delivered with great confidence. Connect the systems. Pipe the CRM into the ERP, the ERP into the finance suite, all of it into a warehouse, and put a dashboard on top so that at last the whole business can be seen on one screen. Integration, the story goes, will join what the decisionless stack broke apart, and the joined-up firm will finally remember itself.

It is a reasonable hope, and it is the subject of the next chapter, because it does not survive contact with the problem we have just described. Moving records between systems faster does not create a memory the systems never held. Integration connects what each system contains, and what each system contains is records, not decisions, so the more of them you connect, the more completely you can see what happened and the less able you are, still, to see why. The result is a business that looks joined up and remains amnesiac, which is in some ways a more dangerous place to be, because it feels like sight. The next chapter is about that illusion.