The State of Decision Intelligence

Walk into three firms in the same quarter. A design agency in one city, a mid-market consultancy in another, a business process outsourcer running client programmes across two continents. They do not know each other. They do not compete. They have never shared a supplier, a conference, or a consultant. And yet each of them, in the last year or two, has quietly grown the same strange new habit.

At the agency it is a document. Nobody was told to keep it, and it has no official owner, but somewhere on a shared drive there is a running file where a few senior people note the calls that mattered: why they walked away from a pitch, why they restructured a retainer, why they moved their best creative director onto an account that was not, on paper, the most valuable one. At the consultancy it is a meeting. It looks like an operating review, and it sits on the calendar as one, but if you attend you notice that the numbers are only the opening act. The real content is a sequence of decisions: what to do about the engagement that is slipping, whether to price the extension honestly or defensively, who to promise to the client they are courting. At the BPO it is a job title that did not exist there five years ago, a small team sitting between sales and delivery whose remit nobody can define in a single sentence but everybody can point to, because it is the team that catches the decisions that used to fall down the gap.

None of the three has a name for what they are building. If you told them they were early practitioners of an emerging discipline, they would look at you strangely and go back to work. But they are. Convergent evolution, in business as in biology, is the surest sign that something real is being solved. When unrelated organisms, or unrelated firms, arrive independently at the same structure, it is because the same pressure is bearing on all of them, and the structure is the answer.

The turn

This chapter is written as a state-of-the-field report, of the kind an industry publishes about itself each year. Those reports have a familiar shape. They open with an adoption curve, a maturity model, a market size and a growth rate, a quadrant with vendors sorted into leaders and laggards.

This report has none of that, and the absence is deliberate, but it needs explaining, because Decision Intelligence is not a new phrase and this book is not pretending to have coined it. The term has been used for years across analytics, decision science, and artificial intelligence, and analysts have published market numbers under the name. What those numbers tend to measure, though, is a mixture of analytics, decision automation, and AI-assisted recommendation. They say something real about those adjacent technologies, and very little about the thing this book means by Decision Intelligence: a widely adopted organisational practice for capturing the decision itself, its owner, its evidence, its alternatives, and its outcome, and learning from it over time. That practice is what remains unsettled, and it is what has no honest chart yet.

Which means the real question is not the one a market analyst asks, how big is this and how fast is it growing. It is the question a naturalist asks in the field: is this real, is it spreading, and how would you know before there is a number to point to? The three firms above are the beginning of an answer.

Why the pieces finally fit

Every idea has a moment when it becomes not merely good but possible, and the two are often decades apart. The idea of a single shared record of a company's transactions is old, but enterprise resource planning had to wait for the affordable database and the reliable server before a firm could actually keep that record at scale. Decision Intelligence is at exactly this kind of threshold now. The idea, that a firm should capture its decisions and learn from them, is not new. A thoughtful managing partner in 1985 would have nodded along and then explained, correctly, why it was impractical. What has changed is not the wisdom of the idea. It is that three things which were missing have quietly arrived at once, and together they remove the practical barrier that made the partner shrug.

The first is that almost every act of work now leaves a digital trace. A generation ago, the context around a decision, the conversation, the meeting, the back-and-forth in a corridor, happened in rooms and on paper and vanished as it occurred; to capture it you would have needed a clerk in the corner writing everything down. That is no longer the shape of work. Meetings are transcribed, the arguments that used to happen in a corridor now happen in a searchable channel, systems expose their state through interfaces, and the raw material of a decision, what was known, what was said, what was chosen, increasingly exists in a retrievable form as a by-product of working at all. Storage did get cheap, and that helped, but a structured decision record was always tiny; a firm could have kept millions of them in 1995 without noticing the cost. The barrier was never the price of the disk. It was that the context around the decision was not being captured in the first place. Now, more and more of it is, whether or not anyone has decided to use it.

The second is that data infrastructure has matured. For thirty years firms accumulated systems faster than they could connect them, and the result was the estate this book has spent two parts describing: nine sources of record and no source of decision. But the plumbing has genuinely improved. The modern practice of pulling scattered records into a common place, warehouses and pipelines and the unglamorous discipline of making data usable, is now widespread and reasonably well understood. This matters for Decision Intelligence because a decision is worth little in isolation. Its value comes from being connected to the evidence it rested on and the outcome it produced. That connective substrate, the assembled record a decision can be anchored to, is finally something most firms possess. They built it to make dashboards. It happens to be the foundation the next layer needs.

The third change is the most recent, and to my mind the most important of the three. Capable language models can now help structure and summarise reasoning. The hardest part of capturing a decision was never the storing of it. It was the writing of it down. Asking a busy senior person to stop after every judgement call and compose a paragraph explaining what they chose and why was a tax nobody was ever going to pay, which is the real reason decision capture stayed a good idea that never happened. Models change the economics of that tax. They can draft the record from a meeting transcript, extract the actual decision buried in a long thread, summarise the arguments on each side, and turn a five-minute conversation into a legible entry a human need only correct rather than compose. This is not the model making the decision. It is the model lowering the cost of remembering that the decision was made.

None of the three, on its own, is sufficient. Cheap storage with no way to connect records is a bigger junk drawer. Mature pipelines with no decision to anchor is the dashboard trap by another name. Capable models with nothing true to stand on is the blind cockpit of an earlier chapter, fluency without grounding. It is the arrival of all three together that matters, because together they turn decision capture from a virtuous impossibility into an ordinary thing a firm can actually do. The barrier that has fallen is the technical one. The barrier that remains, the harder one, is organisational. The pieces now fit. Most firms have not yet picked them up.

The idea that a firm should learn from its decisions is old. What is new is that it finally became practical, and the firms that noticed are already behaving differently.

The signs are behaviours, not numbers

Because there is no honest chart, you read the state of this field the way a field naturalist reads a landscape, by tracks and habits rather than by census. And the tracks are everywhere once you know their shape.

Start with the document at the agency, because it is the plainest sign of all. Across a surprising range of firms, people have quietly started keeping decision logs. They rarely call them that. It is a running note, a channel, a standing item in a shared file, a habit one leader started and a few others copied. But its function is unmistakable: it exists to record not what happened, which the other systems already capture, but what was chosen and why. This is not an invention of the last few years. Software engineering has kept architecture decision records for a long time, precisely because engineers learnt the hard way that a codebase remembers what was built and forgets why, and that the why is what you desperately need when you return to change it. What is new is that the same instinct is now surfacing outside engineering, in the commercial and delivery cores of ordinary firms, arrived at independently by people who have never heard the phrase.

Then the meeting. Look closely at the operating reviews that have proliferated across well-run firms, the monthlies and the quarterly business reviews and the pipeline calls, and notice how many of them have quietly stopped being reporting meetings and become something else. A reporting meeting exists to transmit the past: here is what the numbers did. But the meetings that leaders actually value, the ones they protect on the calendar, have quietly turned into something closer to decision rituals. The numbers are the first ten minutes. The real work is the argument that follows: given this, what do we do about it, who owns the move, what did we decide last time and did it work. The best of these reviews are, functionally, the firm gathering to make and record its decisions in one room on a regular cadence, even though nobody wrote that on the invitation. The genre label says review. The behaviour says decision.

A meeting where executives decide things is nothing new, and not yet the discipline. What is new, and what actually marks the shift, is when the review starts to leave a trace: when the decisions it produces get an owner, a reason, and a place to live, and when the next review opens by asking how the last round of decisions turned out. The line between the two is the whole of the difference, and most operating reviews have not yet crossed it. What is worth noticing is how many are drifting toward it.

And then the request, the one every finance leader and every chief executive eventually makes, usually in a tone of mild exasperation. Give me one number I can trust. It sounds like a request for a dashboard, and firms treat it as one, and it never quite works. The team builds the number. The number gets contested. Someone finds a flaw in how it was assembled, the definition shifts, two departments produce two versions, and the leader is back where they started, holding a figure they cannot fully believe. What is happening, underneath, is that the leader has asked for a trustworthy output while the inputs remain a pile of unrecorded judgement calls. The forecast is only as trustworthy as the decisions about which deals are real. The margin number is only as trustworthy as the decisions about scope and staffing that no system holds. Chase the request for one trustworthy number to its root and you find it was never a request for a number at all. It was a request for decisions you can trust, and the number is only their shadow.

The request for one number you can trust is, underneath, a request for decisions you can trust. The number is only ever as good as the reasoning it stands on.

Three behaviours, then, none of which appears in any market survey, all of which point the same way. This is what the early adoption of a discipline looks like from the inside, before it has a name to announce itself with.

The functions with no good name

There is a second class of evidence, and it hides in the org chart rather than the calendar. Watch the new functions that firms have been standing up over the past several years, and watch especially the ones whose job is genuinely hard to describe.

Revenue operations is the clearest case. A decade ago the function barely existed, and now a great many firms have one, or are building one, or feel guilty that they have not. But ask five leaders what it does and you will get five answers, and underneath the variety one theme keeps recurring. Revenue operations tends to become the de facto owner of the decisions that fall between sales, marketing, and finance, the decisions that no single one of those departments could make alone and that all of them were quietly making badly. How do we define a qualified opportunity. What does the forecast actually mean and who is allowed to move a number in it. When two teams disagree about a deal, who decides. These masquerade as reporting problems. They are decision problems that live in the seams, and the function that grew up in the seam keeps finding itself holding them.

Delivery operations, or delivery excellence, or a transformation office, whatever a given firm has chosen to call it, is the same instinct appearing on the other side of the business, later and less consistently named. But in firm after firm some small team has appeared to hold the decisions that fall between sales and delivery: what did we actually promise, can we actually staff it, when an engagement starts to slip who has the authority to change course, how does the reality of delivery get carried back to the people making the next promise. Often it is assembled by a leader who could not name what they were solving but knew that work kept falling through the same gap and that somebody had to catch it.

Look at what these functions have in common and a pattern shows through. They are not defined by a system, the way the sales team owns the CRM or finance owns the ledger. They are defined by a boundary. They live in the spaces between departments, and whatever title the firm chose, they keep being pulled toward the same work: owning the decisions that cross those spaces and belong to no single system. This is not proof that they are secretly Decision Intelligence departments. Most of them have never heard the term, and they do plenty of other work besides. It is something more useful as evidence. It shows that when decisions routinely fall between departments, firms feel the need for someone to own them, and they will grow an organ to do it before they have a name for the job, in the same way that a body under a new strain builds muscle where the load falls. The awkwardness of the titles is not confusion. It is newness. The discipline is what names the need and makes the response intentional.

An honest maturity read

Now the part of a state-of-the-field report: the honest assessment of how far the thing has actually come. The answer is not far, and saying so plainly makes the argument stronger rather than weaker, because the alternative is to pretend a young thing is mature and to be caught out by the first reader who works in the field.

Decision Intelligence today is early, uneven, mostly manual, and mostly invisible. It is early in that the practices described in this chapter are recent, thin, and concentrated in the firms that happen to have a leader who cares. It is uneven in that a firm may run a superb decision ritual in its finance function and nothing at all in delivery, or keep an immaculate log for one client and none for the rest; maturity is patchy within a single firm, let alone across a sector. It is mostly manual in that almost all of it runs on improvised tooling, a text file, a spreadsheet, a channel, a meeting, a person who remembers to ask the right question, rather than on anything purpose-built. And it is mostly invisible in that it happens under other names, inside meetings called something else, performed by functions with borrowed titles, by people who would not recognise the term for what they are doing.

Put those together and you get the truest single description of the field's current stage: it is a discipline that practitioners are inventing locally, without shared language. Each of the three firms in this chapter's opening has built a version of the same thing, and each has its own private vocabulary for it, and none of them can learn from the others, because none of them knows the others exist as fellow travellers rather than unrelated firms with an idiosyncratic habit. Every mature discipline passed through this stage. It is not a failure. It is a phase, and it is the phase this one is in.

Decision Intelligence is not a solved, packaged, sold category that you can go and procure. The reason to write it down carefully anyway, and the reason a body of state-of-the-field thinking has begun to appear under the Optimal Nexus name, is that reading a discipline correctly while it is still forming is precisely what lets a firm act before the crowd does. You do not get to be early twice.

Five signs you are already doing it

The honest maturity read raises a practical question: how would you know where your own firm sits? Not by whether you own the word, which is worth nothing, but by whether you exhibit the behaviour. There are five signs, and they double as a rising scale.

• Important decisions leave a structured record, not just an outcome buried in a system.
• Those decisions have named owners.
• The evidence and the alternatives are preserved, not only the choice that won.
• Outcomes are connected back, later, to the decisions that caused them.
• The next similar choice begins from that memory rather than from scratch.

Read them in order and you can feel the discipline deepening down the list. The first sign is a firm that writes its decisions down. The last is a firm that compounds them, where a hard call in a room draws on the firm's own record of the last ten calls like it. A firm doing all five is practising Decision Intelligence, whatever it calls the activity, and whether or not a single model is anywhere in sight. A firm doing none is deciding in the dark, however luminous its dashboards. Most firms today do one or two by accident and none on purpose, and that gap is the whole distance this book is asking a leader to travel.

Why people businesses feel it first and hardest

Return to the three firms in the opening, and notice the thing they had in common that has gone unremarked until now. They were all people businesses. The agency, the consultancy, the BPO. That was not a convenience of the example. It is the heart of the argument, and it explains why the discipline is surfacing here first and biting here hardest.

A company that makes a product also lives or dies by its decisions. Pricing, the roadmap, inventory, safety, supply chain, capital allocation: a single bad call in any of these can destroy enormous value, fast. The difference is not that product firms are insulated from judgement. It is where the judgement goes. In a product business, a great deal of judgement is embodied, folded into a repeatable product, a process, an asset, a design that carries the decision forward and applies it a thousand times without anyone re-deciding. The judgement is banked in the thing, and the thing keeps paying it out. In a people business there is far less to bank it in. The product is judgement, delivered fresh, in real time, by individuals: who does what work, for which client, at what price, to what standard. The decision is not folded into a durable object that will apply it again next week. It is made, delivered, and gone. There is no product holding the reasoning in place after the person who reasoned it has moved on.

Follow that through and you can see why the absence of a decision layer is felt so directly here, along three edges at once. It is felt in margin, because a people business earns its money in the gap between what was sold and what it costs to deliver, and that gap is nothing but a running total of decisions, priced, scoped, staffed, escalated, each one small, each one unrecorded, the whole leak invisible until it surfaces in a month-end number nobody can explain. It is felt in quality, because the client experiences your judgement without any product wrapped around it, and a bad call is not hidden inside a manufactured good but delivered straight to them as the service itself. And it is felt in trust, which an earlier chapter argued is nothing more than a client's prediction of your future decisions, extrapolated from the record of your past ones, so that a firm which cannot see its own decisions is a firm that cannot prove its own trustworthiness at the exact moment it is asked to.

The pressure that produces the discipline is highest where the product is judgement, because there the missing layer is not an abstraction. It is the difference between a good year and a bad one, felt on the skin.

But wouldn't the big platforms already own it?

Here is the strongest objection, and it deserves the strongest form. If this were as real as the chapter claims, it would already have a market. The large platform vendors would have named it, packaged it, and started selling it. There would be a tidy category with an agreed definition, a row of logos, an analyst's quadrant sorting the leaders from the challengers, a hype cycle chart telling you where on the curve to worry. The very fact that Decision Intelligence has none of that settled apparatus is surely evidence that it is not a real market at all, just a consultant's coinage dressed up as a movement. Where is the industry, if the thing is so inevitable?

Take the objection seriously, because its premise is sound and its conclusion is simply backwards. The premise is that mature markets have tidy categories. True. The error is inferring that a thing without a tidy category must therefore be unreal. Get the sequence right and the inference collapses, because categories are not how disciplines begin. They are how disciplines end up. Every discipline that now has a quadrant spent years first as scattered local practice, invented independently by practitioners who had no shared name for it, exactly as described in this chapter. A generation of leaders can remember when the phrase data scientist did not exist, when the people doing that work were statisticians and analysts improvising with whatever tools they had, long before there were job titles, degree programmes, tooling categories, and conferences to gather them. In every case the order was the same: the behaviour came first, then the shared name, then the tooling, and only last of all the quadrant. The category is the fossil a living practice leaves once it has stopped moving fast enough to be caught and labelled.

So the absence of a tidy category is not evidence that the thing is unreal. But it is not proof that you are early and inevitably right, either. A missing category can mean a field is early, or that it is fragmented, or poorly defined, or that demand is weak, or that adjacent categories have already absorbed it. On its own, the absence tells you only that the language, the boundaries, and the market structure are still unsettled, which is a fact about the map and not the territory. The positive evidence, the reason to believe something real is forming, is not the empty space where a category will be. It is the behaviour filling it: the logs, the decision rituals, the boundary-spanning functions, the same structure reinvented independently by firms that never compared notes.

A missing category proves nothing on its own. What tells you something real is forming is the same structure, reinvented independently by firms that never compared notes.

There is a second reason the big platforms are unlikely to have owned this first, and it is structural rather than a matter of timing. Recall the diagnosis of Part II: every major system was built around a noun, a record, a customer or a transaction or a ticket, and the decision falls between them, owned by none. A platform vendor's product is one of those systems. Its natural instinct is to pull the decision layer inside its own walls, to make the decision a feature of the CRM or the ERP or the analytics suite. But the decision layer, by its nature, cuts across all of them; it lives precisely in the seams that no single system can see, which is exactly why firms are growing cross-boundary functions to hold it. A discipline whose defining trait is that it spans systems is the least likely thing to be originated cleanly by the owner of any one system. It tends instead to emerge where the objection said to look and found nothing tidy: in the practice of firms, in the gaps, under improvised names, ahead of the market that will eventually form around it.

Which leaves the practical point, the one a commercial leader should actually take from this. The firms formalising the discipline now, before the category arrives, are not early for the sake of being fashionable. They are acquiring the advantage of the well-prepared. When the shared language and the tooling do arrive, and they will, those firms will already have the operating muscle to use them well, while their competitors are still reading the brochure and mistaking a purchase for a practice. You cannot buy the years of habit in a hurry when the quadrant finally publishes. That is the whole of the advantage, and it is only available before the thing is obvious.

The bridge

So that is the state of the field, read honestly. The idea has become practical because the missing pieces have arrived together. The behaviours are spreading, visible in the logs firms keep, the reviews that have quietly become decision rituals, the functions grown to own the gaps, and the leaders discovering that the one number they wanted is only ever as good as the decisions beneath it. And the pressure is highest, the payoff soonest, in the people businesses where the decisions are not inputs to the product but are the product itself.

All of which sharpens a single question and hands it to the next part of the book. If Decision Intelligence is a real discipline, arriving now, and felt first by firms whose trade is judgement, then it is not enough to recognise it in the wild. A leader needs a coherent way to practise it on purpose, a way of running the whole firm as one connected decision loop rather than a set of departments each keeping its own private half of the story. That way of thinking exists, it has been worked out most fully for exactly these businesses, and it has a name. The next part gives it one.