Memory before Intelligence
Field notes from an operator who built his own company's executive operating system — and discovered, along the way, that it was suitable for any function.
1. The scene
I sit in the meeting and someone proposes, with conviction, exactly the decision that was discarded two years ago. It is not bad faith. It’s the new person in the job. Look at the numbers, do the calculation that anyone would, arrive at the obvious path. The obvious way that we've already tested, that was expensive, that we've recorded somewhere — and no one in the room remembers where.
I vaguely remember. I remember there was a why. I remember that it involved a client, a deadline, a risk that nobody saw at the time. I don’t remember enough to defend the position. I look to the side and realize that I am the oldest memory in the room. And even I have nowhere to point. The collection is scattered among emails, hallway conversations, minutes that no one reread, a document that perhaps exists in a folder from another team.
We approve the decision again. In eighteen months from now we will discover the same problem. Let's call this learning.
This paragraph isn't a scene, it's a pattern. It repeats in pricing meetings, in supplier committees, in process choices. A tribal criterion becomes a rule, becomes habit, becomes norm — and no one can anymore reconstruct the reason. When someone asks “why?”, the best answer in the house is “it’s always been that way.” The second best is “I think it has to do with that client from 2022, let me confirm”. Nobody confirms.
Each meeting of these is expensive in three ways. It costs the hour spent redoing analysis that already existed. It costs the wrong decision when the why is left out of the equation. And it costs, above all, the trust of new people, who realize that entering this house means guessing — not learning. The invisible cost of oral memory is this: the company repeats itself without realizing, and blames the market for the learning curve of each new person.
I started taking notes. At the beginning it was just me, at the end of the day: what I decided, based on what, versus what alternative. It wasn’t a project, it wasn’t a method. It was afraid of forgetting. Every week, I saw that half of my questions were questions that I had already answered three months before — and I didn't know I had answered.
Then I started to note it down before the decision, not after. What options did I consider. What evidence I looked at. Which hypothesis did I reject and why? This became a habit before it became a discipline. Within a few months, I had a place to point to when someone asked “why did we decide this?” Not the memory of a person — a quotable trace, with date, with context, with the client of the time, with the discarded alternative and the reason for discarding.
The intention was never to build product. The intention was to stop paying twice for the same learning.
I started on my shop floor — the commercial side, which is where I operate. Customer, supplier, quote, negotiation, decision. The area that I know from bottom to top, where I am the guy who gets into the fight and comes out with a bruise. This was where the pain was most obvious to me.
In about three months I noticed something that I hadn't anticipated. The same architecture that was preserving commercial memory was used for any area that made a recurring decision. Shopping had the same problem. Operations had the same problem. Financial had the same problem, on another scale. HR and legal had discreet versions of the same problem. The pain was not specific to sales. The pain was of an executive function — any function in which today's why becomes tomorrow's symptom.
It was not an intellectual discovery. It was a scare.
I'll tell you what I've learned in these eighteen months. But the story isn't about technology, and I'm not going to try to sell you tools. It's about a habit reform that changed the way I make, record and retrieve decisions. Four pieces support each other — memory, politics, decision, cadence. Each one alone is insufficient. Together, they form what came to be, without my planning, an executive operating system.
Before I get to these four pieces, I need to show you the diagnosis that forced me to look differently. Five pains that appear in any area of any serious company — and that no one names out loud because they seem cost of existing. They are not.
2. The diagnosis
Five pains. I looked at them in the commercial first, then I saw the same figure drawn in purchasing, in operations, in finance. I changed the vocabulary, I changed the badge of the person who was complaining. The geometry was identical. Every house that's grown beyond the size-of-a-founder lives with these five — and each one is quiet enough to seem a natural cost of business.
No it is. Each one charges a price every month, but the invoice does not come with the name of the pain. It comes disguised as churn, as a bad margin, as people leaving, as an audit that doesn’t close, as a customer that disappears without warning. I will describe the five as they appear on the ground, without sparing on concreteness. If you recognize at least three of them in your own day, you already know why we're talking.
2.1. Critical knowledge lives in heads
The question that opens up this pain is always a “why” question, asked by someone in good faith, at a time when no one foresaw it being asked. “Why did we stop working with that supplier back in 2023?” — silence, three people look at each other, someone takes a guess. “Why did this item explode last year?” — you open the report, you do a new analysis, you spend a week to arrive at an answer that already existed somewhere. “Why does this safety rule exist?” — nobody knows, but everyone follows, and when someone proposes to change, the house gets tense without being able to explain the discomfort.
The critical point is not individual memory. It's that institutional memory is the sum of individual memories that no one has consolidated. The commercial loses a large client in 2023 — three people involved lives experienced the pain, knew the causes, and none wrote why in any quotable place. Two have changed areas since then. The third comes out tomorrow. In eighteen months, “why did we lose this client?” becomes a question without an owner.
Shopping has a harder version. The supplier panel seems stable until someone asks “why did Manufacturer X leave the preferred list?” and no one can rebuild. It may have been quality. It may have been a deadline episode. It may have been a political conflict on the canal. Without a trace, the new buyer forwards quote to the same manufacturer six months later, and the house rediscovers the problem the expensive way.
Finance lives this at another pace. Every item that came out last year has already been explained — in some call, by someone who understood The context at the time. But explained is different than recorded. When the CFO comes in November with the inevitable question of the budget, the explanation It's not spread across three heads and four email boxes. Operations is where the pain gets most dangerous: SOP written in 2019, updated a v ez in 2021, with a weird clause that no one dares to remove because “there was a reason”. No one remembers which.
The whole house operates like an archive where the documents were, but the index was lost. There is a record of what happened it happened. There is no record of why it happened. And why is what decides what you do tomorrow.
2.2. The decisions become phantom rules
Every mature house has unspoken rules stronger than the written ones. Pricing criteria that no one has formalized but everyone respects. Limits of approval that appear only when someone exceeds. Qualification standards that are learned by observing, not reading. They work — until they stop working. In commercial, the classic case is discount. The house has a logic of tribal pricing: such customer receives X, such segment accepts Y, above xix of Z we don't go. Those who have been there for five years know how to read the game. Those who joined this month, don't. the unspoken house standard was 6%, and nobody notices until the portfolio margin starts bleeding three quarters later. When the question, “What is our discount policy, anyway?”, the honest answer is: It depends on who you ask.
Purchasing has the same problem disguised as “supplier qualification criteria”. There is a list. The list is partial. The real disapprovals live in “we don’t work with this guy because…” — and the why is a story that He lives with a senior buyer. When this buyer takes holiday, the house quotes with a prohibited supplier, discovers on delivery that it was prohibited, and no one can explain in writing what everyone knew orally.
In finance, the ghost rule is usually the approval limit. The CFO approves X up to such value, requests a committee above that, but the “such value” changes It changes with the category, it changes with the supplier, it changes with the quarter. No one noted the exceptions. emanates, not out of ill will, but because the real rule was never written. In operations, it's the SOP that everyone follows in a slightly different way than it is on paper — the way it works is oral, the way it is written is outdated, and the audit is the one who charges.
Ghost rule is not a lack of rule. It's an excess of unwritten rule. And unwritten rule only exists as long as the person who carries it exists in the company.
2.3. The report turned into theater
Every household has a reporting ritual. Pipeline review on Monday. Supplier dashboard on Tuesday. Monthly closing at the end of the month. Shift meeting at the factory ica. The meeting is held. The slide is looked at. It is ended. The review pipeline is the most visible example because it is, in general, performative by design. The seller enters with the opportunity at the stage and m that he thinks will sound good. The manager questions with vigor calibrated to the size of the audience. “Follow-up pending” is written and closes the screen We left the room without any new commitments that didn't exist before the room — but with the comforting feeling of having “reviewed the pipeline.” Next, the same deal comes back. The same “pending follow-up” reappears. No one asks what has been done since the last time. Purchasing lives this on the supplier dashboard. The dashboard exists. It's up-to-date. No one opens it outside of the monthly meeting, and within the monthly meeting it is looked at but not used — it turns out in the background for a conversation that would happen anyway. Financial has the monthly closing: you run the number, you present itself to the committee, you celebrate or justify yourself, and the next time when that the return number is at the next closing. There is no accountability on what each manager promised to correct between one closing and another. t happening on the floor — the number is visible, the deviation is visible, and even then the corrective action depends on someone being in a good mood to start the conversation.
The theatre works because it fulfills a real need — that of appearing as one is looking. But looking is not deciding, deciding is not registering, and registering is not charging. Without the four pa That's it, the report becomes liturgy. The rite is fulfilled, the appearance is maintained, and the house remains blind to the same signs that it insists on projecting on the wall every week.
2.4. The onboarding is bleeding
Ask any senior manager how long it takes for a new hire to become productive in the job, and the answer will come down out of courtesy. Ask again, sincerely, and the number doubles. Commercial it takes six months for a salesperson to understand the portfolio to the point of not giving a flag in the first conversations with a large customer. Purchasing takes four months for a buyer to understand the supplier panel — who is reliable , who is the historical problem, who you negotiate with, and who you accept. Financial takes three months to a year list decipher the DRE of the house, because the DRE of the house has its own names, historical exceptions, mirror accounts that They survived a 2020 reorganization. Operations is the worst: nine months for a technical leader to master critical processes, and even at nine months there are still decisions that he avoids making alone.
These deadlines are not for technical learning. They are for absorption of unwritten context. The employment contract says that the A person starts on Day 1. Real productivity starts when she already knows the names of the ghosts — which customers have history, which most suppliers are not touched, which rules apply when, which SOP is truly followed, and which is the audit piece. Learning happens through tribal absorption: hallway, lunch, coffee, “let me tell you a story.”
Houses that grow fast pay for this bleeding double. Each key departure resets the clock for whoever takes over. Each new hire consumes time of senior people who could be generating return accepting or avoiding risk. And the learning curve never ends, because the context one absorbs doesn't stay anywhere — it goes along with the person when they leave.
The cost of hemorrhage is rarely in the spreadsheet. It appears disguised: high turnover in the first six months, new salesperson losing deals that the senior would have closed, new buyer accepting a term that the senior would have ne gocido, new operator driving shift the way he understood, not the way the house needs. Each of these losses is treated as individual.
2.5. The leader becomes cognitive bottleneck
The four previous pains are supported in a fifth, which is the heart of the problem. When critical knowledge lives in heads, when decisions become phantom rules, when the re lab becomes theater and onboarding bleeds for months, someone needs to hold on the memory of the house. That someone becomes the leader. The VP of Sales is the one who remembers everything. He remembers the criteria with which the house closed that contract in 2022, he remembers the discount that was given conceded in exchange for what counterpart, remember the client who came in through the back door and why. Without him, the team runs limply. him, the team runs — but any substantial questions go through one head. ensives. When he leaves, the successor takes a year to reconstruct the map that lived between the previous's ears.
Purchasing has the senior buyer — the guy who knows which manufacturer raises his voice with and which he doesn't, which rep agrees to renegotiate and which hangs up the phone. No one lists reviewed it. that that loss was acknowledged in that way in that specific situation. l who hasn't been able to take a real holiday for two years, because any relevant stop on any line will eventually hit his cellphone.
The house looks efficient from the outside. Seen from the inside, it depends on four to six heads who carry the why. Each of these heads is a single point of failure, and no one dares to name the problem because to name it would be insecure. sult who is holding the house. The result is an organization that confuses leadership with compulsory retention of context — and that pays the price for this confusion in the first output that wasn't in the plan.
This is where oral memory shows its nature: not a diffuse cost, but an existential risk concentrated in specific people. And this is where the question that carries the rest of this text arises. depending on people's memory, and if the dominant reflex today is to “buy an AI tool that resolves this,” pre We need to look carefully at what AI actually solves — and at what it only amplifies.
3. Why AI alone doesn't solve
The first reaction of almost every executive who recognizes these five pains is the same. “So it’s a case of implementing AI inside.” It's not bad faith, and it's not intellectual laziness. It's the reflex trained by the last three years of public discourse — if the problem is memory, context, repeated decision, response that takes time, then the way out is technology that learns, that remembers, that summarizes, that converses. Buy a tool, integrate with the systems, train the team, wait for the gain.
The reflection is correct in the direction and wrong in the object. The AI has something to do with the output — but not in the form of product you buy. The problem Of the five pains it's not a lack of artificial intelligence. It's a lack of substrate. The confusion is of category, not of execution.
Karpathy — who ran AI at Tesla for five years and had been at OpenAI since its founding — was the one who named the turnaround with the most pr ecision. In September 2023, it published a sentence that was widely circulated among those operating on this border:
“LLMs, not as a chatbot, but the kernel process of a new Operating System.”
It's worth stopping at this word. Kernel is the core of the operating system — the computational heart that receives instructions, orchestra It processes, performs calculations, returns results. It is the most sophisticated part of the computer. Kernel without memory can't store what it processed yesterday. Kernel without disk has nowhere to store what it learned. Ke Kernel without a bus doesn't talk to the rest of the world. Kernel without a filesystem doesn't know where to find what it needs as you need. The kernel processes what it receives on the moment, in the order it receives, with no trace of what it processed before. Do it today, with brilliant, on top of what you throw in the context window — and forget everything the next moment.
The computer you use every day isn't the kernel. It's the whole operating system — kernel plus memory plus disk plus more files s bus plus protocol. The kernel is the noble piece. The system is what makes the noble piece use for something. The turn that matters for your company is this. The model is the kernel. The model is ever better, and will keep improving regardless mind what you do. But your company doesn't need a better kernel. It needs the operating system around it.
The practical consequence of this category error is already visible in those who have tried the shortcut. Executive opens the Cha tGPT, pastes three emails, a spreadsheet, and the minutes of the last meeting, and asks: “what’s the best next action with this customer?”. You receive back a plausible, well written, safe in tone — response — and without any connection with what happened in the previous twelve months with this client, with the unspoken criterion of the house for that segment, with the decision. tale that was granted to this type of account last time, with the story of the senior contact that this executive doesn't know. The answer sounds good because the kernel is good. ivo reads, finds it interesting, closes the tab, and acts on the instinct he had before asking. That was it. of the art of “AI in the enterprise” for most homes today — a beautiful summary of nothing.
The second mistake comes next. Recognizing that the first attempt didn't go ahead, the executive escalates to purchase a tool — a copilot, an integrated assistant, an internal chatbot fed with documents of the company. The premise changes from “I’m going to use public AI” to “I’m going to implement AI inside home. The reason is simple and has already been experienced by any house that implemented ERP in the 2000s or BI in the 2010s. Tool is muscle. Architecture is system Nervous. Buying more muscle in a company without a nervous system only amplifies what it was already doing badly — faster, in more places, with more relation shine Houses that buy eight distinct SaaS and don’t have the nerve to coordinate between them end up exactly where houses that buy “AI” without the four pillars. end up: with new spending, old pain, and the confused feeling that the technology has failed. It has landed on ground without foundation.
The jump that matters isn't in the model. It's in what's around it.
It's in the place where the memory of the house gets cured — not crammed in Drive, not fragmented among ten systems, not residual in the heads of four key people. It's in the way policies are written, versioned, retrievable. It's in the protocol by which a decision becomes an auditable event, with a date, with a why, with a discarded alternative. Four pieces. Four architectural decisions. Each alone alleviates a pain. Together, they form the substrate on top of which any kernel — ess and, the next, whatever comes in two years' time — starts to use. Without them, the best model in the world remains an engine without a chassis.
That's what we're going to talk about now.
4. The four pillars
The pillars are four: memory, politics, decision, cadence. There are not four features, there are not four products, there are not four projects that an area owner touches on. parallel while the others pretend they exist. They are an architecture — and an architecture only works if the four pieces support each other. a dead accumulation, bloated folder that no one consults. Politics without a decision is a poster on the wall. It matters. Cadence without memory is theatre that is repeated the next week, with the same slide and the same feeling of having already had this conversation.
I thought of them separately at the beginning, in the fright of day to day. I resolved one pain at a time, thinking that each piece was a piece. Only after a year did I realize that I had to password, without a name, a chassis. The four joints are what give the kernel — any kernel, this, the next one — the place to land. As each one appears in the routine of those who really decide something every day. And in each pillar, at the end, I leave a short sidenote about how this same problem is treated today in Technical vocabulary of applied AI — not to become a class, but because the translation helps those who need to explain it to a board or to a technical team.
4.1. Memory — the single truth
Memory, in the sense it matters here, is a specific thing: each entity that your company touches — customer, supplier, contract, decision, process, policy ca, person — has a canonical, living, indexed, quotable. One place, not five. This diagnosis is simple and uncomfortable: When someone asks “what do we know about this client?”, there is a single answer that anyone in the house can Open, read, and understand in three minutes — or are there eight partial answers, each in a head, in an email box, in an orphaned folder?
Today, the response of the majority of houses is the second. Important client has a record in the commercial system with old data, real history spread between the account manager, two emails from dir etor, a visit minute that no one has indexed, and a personal spreadsheet that someone keeps separately. Within the architecture, that same client has a live page — key people up to date, history of orders with decisions noted, context of each negotiation recorded on the day, policy applied for that segment referenced by link, open decisions marked as open. account comes out, the next person opens the page, reads, and starts from day 1 with what the predecessor took three years to accumulate. In the technical vocabulary of applied AI, this work is called “context engineering”: curating the context that feeds the system is more decisive than and choose the model. The business version of this idea is to build the institutional memory that feeds every decision in the house.
4.2. Policy — rules outside your head
Policy, here, is not a HR manual on the shelf. It is the set of de facto rules — pricing, discount, supplier qualification, common nication with client, risk escalation, confidentiality, succession, ethics in gray cases — written in prose, versioned with date, changeable with trace. Each rule has an identifiable author, a date of last revision, a reason justifying its existence, and a condition That authorizes his change. It’s not a ghost rule in the leader’s head. It’s a public rule, inside the house, with an address. The classic case is that of the discount policy that we saw in the diagnosis. Today, it is tribal: the senior salesperson knows, the new one discovers at the first slip, the margin of the portfolio responds by all. Within the architecture, the policy is written — who approves how much, with what counterpart, in which segment, under what condition. Explicitly decided, recorded, and becomes input for the next policy revision. on the pricing verse with the house behind him — not against him. The leader ceases to be permanent translator of the unspoken. In the vocabulary of autonomous agents, this is the “system prompt” or “steering” — how you direct the behavior of a system that makes decisions without you. alongside. Without explicit policy, each subordinate is an agent without instructions: improvises with good intention, and each to one side.
4.3. Decision — immutable event with why
Decision is the piece that most bothers those who see themselves for the first time within this architecture. Because it requires a discipline that no one has trained in: dealing each d ecision relevant as event — with date, context, alternatives considered, evidence looked at, why, and an explicit criteria of review. The event is immutable. A registered decision is not rewritten because someone changed their mind in May. the evidence, and marks the previous as revised. The history is intact. The reasoning is auditable. Today, a typical decision takes place in a meeting, with three people, and disappears in the next minute. It appears in the minutes as “decided: go forward with supplier B”. It doesn’t appear why, the discarded alternative A doesn’t appear, it doesn’t appear the evidence looked at, the sign that would justify going back doesn't appear. Six months later, when supplier B makes the first trouble, the house will not be able to rebuild if the decision was a good decision with a bad or bad result decision with predictable outcome. Within the architecture, the decision is in a canonical place, with all of it recorded ated on the day. When supplier B is ready, you open the record, read the why, compare the evidence of the time with the current evidence, and one decides: the decision was good and the result is noise (holds), or the evidence has changed (records the r eversion with why) In any scenario, the house learns. Today, the house just reacts.
Daniel Kahneman divided human thinking into two systems — the intuitive, fast (System 1) and the deliberate, slow (System 2). Applied AI research has been adopting the same vocabulary to describe how the Current models operate almost everything on System 1 — and that the leap is in forcing System 2. Companies have the same problem. Recording decision is the concrete act of forcing organizational System 2.
4.4. Cadence — the audible ritual
Cadence is the rhythm by which the house perceives, plans, acts and reflects. It's not the meeting. It's not the report. It's the cycle. Daily, someone realizes what changes I give the outside world — a customer who reacted, a number that went out of the curve, a risk that appeared. Based on what he perceived, he decides what is going to move on the day. At the end of the day or the week, look back and record: what happened, what you learned, what changed in the next week’s plan. Four steps. There is no plan. Without a plan, action is reactive. Without action, the plan is statement. Without reflection, action does not become learning — it becomes repetition. Today, in most houses, the cycle is broken in at least one point. There is perception without a plan (meeting that ends in “ok, let’s think”). There is a plan without action (calendar full of initiatives that no one pulls). There is action without reflection (sprint after sprint without retro). steps have a fixed place in the time of the week— not as bureaucracy, but as ritual. Whoever perceived, perceived in writing. Whoever planned, planned on what he perceived. year. Whoever reflected, recorded it on the same page where he planned. The next week, the same page is opened, and the cycle restarts on top of the previous one — not on top of forgetfulness.
The cycle of perceiving, planning, acting, and reflecting is what has been called, in the vocabulary of agents in applied AI, the “agentic loop”. This is what differentiates them. a system that only responds to a system that learns. Without this cycle, the agent doesn't evolve. Without this cycle, the company doesn't either.
The four pillars are the architecture. It's what exists around the kernel — the substrate that gives sense to any model that comes on top. But architecture, so called m, it is still an abstraction of spec. It does not answer the question that every board chair asks inside themselves when reading a text like this: “everything is okay, and what concretely changes for Me?". The owner who still operates. The shareholder who looks from the outside. The CFO who carries exceptions in his head. The COO who doesn't sleep. The commercial director who lost his senior. The director of purchases that still decide orally. It is this question that carries the next act.
5. What changes — perks per persona
Architecture, said like this, is still design. What it does for you depends on the chair in which you sit. The owner who still serves clients e. The counselor who sees the house from outside three times a year. The hired CEO who inherits a team that knows the company better than him. The one who carries exceptions in his head. The COO who doesn’t sleep. The VP of sales who lost his senior in January. ce your portfolio. The HR who does onboarding in a tribal place.For the owner not yet operational
You are the one who gets into the fight and comes out with a bruise. Your house works because you serve a difficult client, decide on an ambiguous quote, read the silence of Supplier before the bad email arrives. You know this is a problem — you just don't see how to get out of it without losing the thread. see outside your head, not as a frozen manual, but as a living trace of discretion. When you're not in the room, the house decides the way you do ecide — the why is in a quotable place. Succession is no longer about “finding another you.” It becomes “running a system that thinks like you. For the shareholder, advisor or board
You will never know the house as the operator does. It's not your role. Your role is to understand if the house is resilient without you becoming the operator. Today, the one The only way is to tour the factory, talk with three key people, and faith. What changes: the house is audible from outside. inspectable star — why, discarded alternative, evidence of the time. Departure of key executive ceases to be an existential threat. Institutional assets become an added asset, not discounted risk.
5.1 For the hired CEO
You don't have the free pass from the owner. You inherited culture, grudges, unspoken rules that nobody told you in the interview. Each decision is read by the person. home against a history that you didn't live. Delegating is the only way — and delegating today is faith, because the context is in heads that are not yours. What changes: delegation comes with a preserved context. When you send someone to resolve a situation, that person opens the track of what the house has already decided iu, read it in fifteen minutes, and act within the criterion — not improvising against it. You get out of the bottleneck without losing the thread.
5.2 For the CFO
In November, someone asks you why that item broke, and you redo the analysis for the third time in the year — for the third person difficult erent. The previous two were to the board and to the auditor. Each time spent two days reconstructing what had already rebuilt, and the reconstruction It dies after the meeting. What changes: each analysis done once becomes a reusable asset. The exceptions memorial — why this account was treated as yes, why that approval stood up — it gets out of your head and turns to the page with address. You respond by pointing. For the COO or industrial director
You live in a contradiction that no one taught you how to resolve. SOP is what the audit wants. SOP-with-why is what the factory needs when a new leader takes the shift. Today your rules are written but lame — they list the what, they hide the why. Someone proposes simplifying a weird clause and the house is tense a without explaining the discomfort. What changes: each rule gets the why attached. The new leader reads and understands, in twenty minutes, why the clause exists and what would change If she left. Technical onboarding falls by half. Human variability decreases without you auditing anymore — because the rule is no longer tribal.
5.3 For the VP of Sales
You know how much it costs to lose your senior. Not in salary — in an invisible wallet. The senior carries, in his head, the story of each client: who decides, which discount makes sense for which account, which old complaint needs warm cloth. When he leaves, the house discovers the wallet of the expensive way — losing deal which and it would close in sleep. What changes: pipeline and account history live outside of mind. Pricing and discounting have defensible track — you show the criterion used In twenty similar cases. Departure of key salesperson fails to reset the portfolio. The replacement enters on day 1 knowing where the ghosts are. For the director of purchasing or supply chain
You deal with people who know their portfolio better than you. The big manufacturer knows how much you bought last year, how much your competitor pa ga, and at which point in your schedule you give in. It has the panel. Each supplier has history Curated — price, quality, real episodes, reasons for level change. Each RFQ learns from the previous one, doesn’t start from zero. You sit down knowing what you've already decided about this supplier in three contexts, and why. He stops surprising you with his own wallet.
5.4 For the HR director
You deal with the problem that no other area names out loud. Onboarding is bleeding. Every key exit is cataclysm. And no one admits that the home and teaches by osmosis — coffee, hallway, “let me tell you a story.” What changes: the cognitive heritage becomes institutionalized, not tribal. iture — the new person opens the living pages of accounts, of policies, of decisions of the last twelve months, and gets to the first productive day in weeks, not in months. Key exit ceases to be cataclysm: the context that the person carried was already in a quotable place. For the CIO or CTO
You are charged by the board for an “AI strategy” and you know, better than anyone in the room, that the technology available is good. It’s missing what nobody We don't want to admit: substrate. Every pilot of the past two years has generated plausible summary and nothing actionable, because the model had nowhere to look for real context. of the house. What changes: the AI of the house stops operating in a vacuum. There is a curated collection where every initiative can land Pricing, decisions with why. You stop defending pilots who don't leave PowerPoint. All of this changes within a particular chair. But there is something that changes regardless of the chair you sit in— and That no one announced. When you install memory, policy, decision and cadence, you didn’t just do operational reform. er, it left standing a record that no workshop, manual or values poster at reception can produce. That's what separates this of any productivity project with AI — and which makes, by construction, inheritable what you've built.
5.5. The byproduct: institutionalized culture
You wanted operational efficiency. You got something bigger. You entered this reform to shorten onboarding, stop repeating analysis, get the leader out of the bottleneck — op output erational, visible gain, contract easy to defend. At some point in the second semester of this walk, someone from the outside — an adviser, a candidate in interview, a student Idleness in late afternoon chat — he'll ask you a specific thing and he'll leave the room with the feeling that he's understood your company in a way that no one has understood before . You will realize, with some delay, that you have documented your culture without ever calling it a culture project. Mission, vision, and values in PDF are theatre. Not because they are false — they are, generally, sincere on the day they were written. They are theatre because they describe aspiration, not behavior. The workshop of cul ture documents what the house would like to be on a Monday offsite, not what the house decides on a Tuesday with a tight budget. Dress, hours, holiday policy — not the tacit criteria with which this house grants discounts, chooses suppliers, closes customers. s six months following learning the real culture by absorption — coffee, hallway, “let me tell you a story.” The official document and the real culture live on different planets.
Real culture is the aggregate of decisions recorded with why, of policies explicit, of memory cured, and of cadence maintained. al evidence. It's the pricing rule that exists in writing and has been revised three times — with the date, the reason, the author. That every area fulfills because everyone has seen that it produces learning, not because it's on the calendar. In systems that learn from data, behavior emerges from what was recorded during training — and in organizations, Culture emerges from the traces that were left when nobody was calling it culture. Same physics, different scale. Reform the four pillars and culture appears as sediment — not as poster.
It's what's left when the leader goes away. It's the only operational definition of culture that matters for anyone thinking succession, valuation, M&A or simply Essentially durability of the house beyond a generation. Culture that lives in heads goes out the door along with the person. Recorded culture remains. it becomes an added asset, not discounted risk. In due diligence, it becomes trace reading, not a runner’s interview. to inherit, not personality to clone. Institutionalized culture is culture that survives the person — and is the only one that deserves the name.
6. How to get started
It's not an IT project. None of the four pieces live in a license, in a portal, in a supplier roadmap. It's not “implantation” — AI is the kernel; what is missing is the surrounding system, and that doesn’t come in the box. And, above all, it’s not a change of tools. It's habit reform. The tool you already have will probably survive this reform. and, how you record, how you review — and how your household stops paying twice for the same learning.
Habit reform doesn't start with the whole company. It starts with an area and a decisionmaker. From this text, the one who keeps up his sleep on a Tuesday with a tight budget. Going through the initial discomfort of writing the why when no one is still demanding the why written. I really go to that area, every day, and I want to stop carrying the why in my head. In twelve months, three things will be different in your house, and none of them needed a workshop. The decisions in the pilot area have an audit trace vel — you open and read. The unspoken rules have become written, revisable rules, with an author and date. Then, without ceremony, you will realize a fourth thing that no one announced in the contract: your company will have a documented culture without ever having done it. a culture workshop. What was written is what the house actually thinks — not what it would like to have thought on an offsite.
If this resonated with something you feel every day but never knew to name, let's talk. No agenda, no deck, no proposal — just a conversation between operators. The rest comes later, if it comes.
Memory before intelligence. That's the order.
You might also like
From literacy to execution: why Brazilian councils need to act in the next 90 days The curiosity phase is over. Between 2023 and 2025, the Brazilian market went through an intense cycle of “AI literacy” — workshops, lectures, exploratory pilots. This cycle fulfilled its role, but now it takes its toll. Boards, investors and clients want to see a return. The question is no longer “what can AI do?” and it became “where is the amount she promised?”
Read full article →Today's challenge is not to start using Artificial Intelligence, but rather to transform pilot projects into measurable value at the executive level. The reality is harsh: while 82% of companies report some positive ROI, 95% of pilots die before reaching production, creating the dreaded "Pilot Gap." The cause of this failure is not a lack of budget, but a failure of method: Lack of strategic alignment with business priorities. Fragmented or ungoverned data. Excessive focus on tools rather than solving real problems. The difference between the average and the leadership is stark: companies that scale AI with governance and method achieve $10.30 return for every $1 invested, almost triple the average of $3.70. For the Council, the success metric is clear. What defines who climbs: Focus on real business problems. Based on organized data (Data Readiness). AI KPIs tied to executive compensation and auditable governance from the outset. Before approving the next investment in technology, the Board's test must be one: the question that matters is not "are we using AI?". It's: "Are we getting a measurable return from AI — and do we have a plan to scale?" If the answer is not clear, your investment in innovation is at risk.
Read full article →Dear C-Level, If in my previous article I highlighted that Artificial Intelligence (AI) is essential for the survival of your business, now it's time to focus on what really matters: Return on Investment (ROI). The enchantment phase with Generative AI has passed; Now we need a practical and strategic approach. The market predicts that AI will add trillions to the economy, but many companies are still unhappy with the results. The problem is not technology, but the lack of a clear strategy and adequate data. As someone who has developed expertise in AI, from the use of no-code tools to the strategic use of OPEN AI's CODEX and multiple agents through orchestrator in N8N, I have learned that the difference between hype and real profit is in value-driven execution.
Read full article →