The One Who Knows Why That Field Is There Is About to Retire

At every company I've worked at there was a field in a table, in some database or some system, that nobody could explain.

It was there, it was used, and if you touched it something broke three modules over. And there was always one person, exactly one, who knew why it existed. Who remembered the 2011 project where somebody decided that for a reason that made sense at the time.

When that person left, the reason left with them. The field stayed.

That person, in the little world of SAP, is on average 47 years old today.

And everything the industry is building depends on getting what's in their head out of it before they go.

That, and something else, is what this edition is about.

The numbers nobody wants to look at

Let's start with two stories that connect to each other. One uncomfortable, and the other one too.

In January, PwC presented its annual survey of chief executives at Davos (things always happen at Davos): 4,454 CEOs from 95 countries. 56% said artificial intelligence brought them no significant financial benefit. Only 12% could show improvements in costs and revenue at the same time.

It's not an isolated data point. MIT found that 95% of generative AI pilots produce no measurable impact on the balance sheet. S&P Global reported that 42% of companies abandoned most of their AI projects during 2025, more than double the previous year.

And this isn't people who didn't adopt. It's people who adopted, spent, and can't find where the value went.

The blindfolded intern

The other day I read this and it stuck with me: using AI at the enterprise level today is like hiring the smartest intern that exists, then blindfolding them, dropping them alone on an island inside the company, and expecting results.

The problem isn't the model's intelligence. Two years ago these models couldn't pass the bar exam and today they'd be in the top 1% of scores. The intelligence comes by default.

What's missing is the context. When you name your best employee, you don't name the one who got the highest score on an exam. You name the one who understands how things get done here, who knows the customers, who knows why that field exists in that table.

That knowledge lives in people's heads. It almost never got written down. And you can't transfer it to a machine because it isn't anywhere.

The other side of the same problem

Now the second story. Last week I saw a video that landed in the developer community, a guy named Brett, developer, twenty years programming. He spent eighteen months using AI tools to write code, seriously, every day. And he says he decided to stop using them entirely. That he's betting on it and wants nothing more to do with the whole thing.

What he describes isn't what you'd expect. He doesn't say the code was bad. He says it was good, that it came out fast, that they shipped more tasks than ever.

What he describes is something else. That he stopped understanding the system he was working in. That reviewing thousands of lines a day written by someone else is impossible to do well. That he started feeling apathy toward code he hadn't written. And that after eighteen months he realized he'd stalled his own growth, because instead of learning he'd been ticking boxes and hitting next. He felt he was vegetalizing.

His key line: you drift away from the code, and when you're far away, you don't even know where to look for what you need.

One story comes from the business side and says AI doesn't pay off because nobody wrote down how this company works. The other comes from the trade side and says he used AI and stopped being the one who understands how this works.

It's the same problem seen from both ends.

The SAP data point that matters to you

I've been going through the latest The 2026 Developer Insights Survey: The Report that SAP put out recently and the numbers leave you thinking. I don't know whether for good or what, but thinking for sure.

The available data indicates that around 72% of ABAP developers are over 40, with an average age near 47. SAP's annual developer survey this year shows the same movement: the older age brackets grow, the 25 to 44 ones shrink. Young people show up, take a look, and leave.

And there's a detail from the same survey worth looking at next to that: 70% of developers already use AI to have concepts explained or to learn new things, and consulting a model shows up third among preferred resources when they have a problem. Which means whoever comes in today learns from the model, and the model learned from the forums and answers the old guard wrote over twenty years. If the one who knows stops writing in the forum because they ask the model instead, the source dries up from both ends at once.

At the same time, SAP's CEO said a few weeks ago that there's a possibility that in three or four years there won't be anyone developing software inside his company.

The easy reading of that, and the one I saw circulating a fair bit, is that SAP wants to do without consultants. That it knows something we don't and is laying the groundwork.

I don't think that's it, and the market numbers argue against it. Rates are firm, there are thousands of open positions (or at least that's how it looks), and consultancies have listed the lack of SAP talent as one of their main operational risks for years. 40% of ECC customers are still defining their migration plan with the deadline on top of them. Nobody wanting to get rid of consultants picks the moment of maximum scarcity to do it.

The more boring explanation is the more likely one:

They're not replacing people who exist. They're covering the hole left by people who never arrived.

When the CEO says programmers are disappearing, he isn't announcing a decision. He's describing something that already happened, years ago, and that he can't reverse by hiring.

The real race

Now put the three things together and look at what comes out.

Companies don't see returns from AI because their business context was never written down. That context lives in the heads of people averaging 47 years old. And what's being sold as the future, autonomous agents, the company that almost runs itself, needs exactly that context written down in order to work.

The race isn't between humans and machines. It's between writing down the context and the person holding it retiring.

And here's the uncomfortable part, the one I've barely seen written anywhere.

What companies are doing to fix the 56% problem is, basically, taking that context out of people's heads and writing it where the agents can read it. Documenting the ontology, mapping the processes, encoding the business rules.

It's a sensible idea, and probably inevitable.

It's also, exactly, what makes someone like Brett feel he's training his own replacement.

Both things are true at the same time. And anyone who tells you there's a comfortable answer to that is selling you something.

Before you agree with me

Two things that cut against what I just said.

The first is about one of my own data points. SAP's developer survey has 589 voluntary responses over a population of hundreds of thousands. Jelena Perfiljeva and Paul Modderman took it apart with class on their podcast: people answering anything out of fatigue, because the survey was endless and didn't show how much was left; architects and analysts answering a developer survey; questions worded in a way that pushes the answer. As a photograph of the ecosystem it's directional, nothing more. The 72% figure comes from elsewhere and is sturdier, but it's worth saying: if I lean on the survey, I'm leaning on something flimsy.

And the same with the age figure: it comes from estimates of professional profiles in the United States, not from a census of the ecosystem. It works as an indication, not as proof. What is supported by several separate sources is the direction: the ecosystem is aging and young people aren't coming in.

The second is about me: I am not neutral. I work in technology, I live off this, and I push AI adoption in my area. When I say the one who understands is still the one with value, I'm describing a world where my own trade still makes sense. Keep that in mind when you read me. Or at least that's what I believe as of today, because next year I don't know.

The 56% is not a model problem. It's the bill for years of not writing down why things are the way they are.

And the part you do control is which side of that equation you're standing on. If you're the one executing tasks, there's a tool that does them faster and cheaper than you, and it will improve every month.

If you're the one who knows why that field is in that table, you're the only source of something no company has written down, and that's about to walk out the side door.

The difference isn't how much you produce. It's how much you understand of what you produce.

Pablo / En Producción

P.S. If there's a field, a process or a rule at your company that only one person can explain, send this to that person. They're worth more than they're paid.

References:

• PwC, 29th Global CEO Survey, presented at Davos in January 2026.
• MIT (NANDA project): 95% of generative AI pilots with no measurable impact.
• Sankar, Prukalpa (Atlan). Interview on the enterprise context layer, 2026.
• Demographic data on the ABAP developer base (72% over 40, average age 47), via DevPro Journal, May 2026.
• SAP Developer Insights Survey 2026: 589 responses from 44 countries.