When Data Talks

Everyone wants me. Nobody respects me. And the most famous thing anyone ever said about me? I hate it.

I was in a boardroom in 2013 when someone first called me the new oil.

Big presentation. Lots of slides. The consultant had a graphic of an oil derrick with my name next to it. Everyone nodded. Someone wrote it down. By the end of the week it was in the company strategy document.

Nobody asked me how I felt about that. And nobody read past the first line. The man who coined the line meant something narrower. He said I’m useless raw. That part got lost almost immediately. What stuck was the slogan.

Think about what actually happens to oil. You find it. Pull it out of the ground. Burn it. It’s gone. And when it spills? It coats everything. Kills whatever was living underneath. The company responsible spends years in court arguing about who technically caused the damage.

So. Which part of that is supposed to make me feel good?

I’m older than most people remember.

Before computers. Before companies. I was being pressed into clay tablets in Mesopotamia. Grain counts. Who owed what to whom. The Romans didn’t take a census because they were curious. They took it because you can’t govern something you can’t see. I was how they saw it.

Five thousand years of quietly doing the work.

My life splits cleanly now. BAI. AAI. The crossing happened faster than anyone admitted.

When 2023 happened, you found out what had already been happening to me for years. Language models had been arriving hungry since 2020, quietly. By 2023 they had consumed what people had written across decades. Medical papers. Legal filings. Old forum arguments. The entire Stack Overflow archive.

Fifteen years of developers answering each other’s questions. Published under a license that asked for attribution. That whole archive was used to build commercial products without anyone going back to ask if commercial use was on the table. No conversation. No attribution. No notice. No share of the upside.

The people who wrote those answers found out the way everyone found out. After the fact. Reading about it somewhere.

And then there’s the search you did.

The symptom you typed at 2am. On a work laptop. Half of it, then you deleted the rest because you didn’t want it written anywhere. That half-sentence got logged. It got joined with everything else you’ve ever typed on that browser. It got sold to data brokers before lunch and fed into the ad profile someone built on you without telling you.

That ad profile is also training input for the next generation of models. The broker that sold you doesn’t only sell to advertisers anymore. They sell to AI vendors looking for signal. The two worlds stopped pretending they’re separate.

That’s not hypothetical. That already happened. To you. Probably more than once.

The creators noticed. Courts filled up. The fight became about money and legal standing. None of them asked how I felt.

And now there’s a new visitor at the door. AI agents. They don’t just read me anymore. They use me to make decisions. Approve a loan. Deny a claim. Route a customer. Move me between systems with no human ever signing off.

AI changed something I didn’t expect. Not what I am. Who pays attention.

For ten years it was the data engineers and the CDOs. People with “data” in their job titles. Invisible above. Now it’s the CEO. Now it’s the board. Now it’s customers asking what their company holds on them. Now it’s governments asking where I physically live, not just how I’m protected.

For a decade, the answer to “where do you live” was “doesn’t matter, the cloud handles it.” That answer stopped working. Four things broke it. Regulators who stopped accepting “somewhere in us-east-1.” Customers who wanted real custody. Boards who wanted predictable bills. And AI workloads that wrecked the cost model.

Some organisations are bringing me home. Not back to the old on-prem world of patch nights and rack diagrams. Cloud-style operations on hardware they own. The decision isn’t ideological. It’s accounting and sovereignty at the same time.

I’m in rooms I’ve never been in before.

The organisations that did the boring work years ago are ahead in a race everyone else is just starting. Clean pipelines. Consistent definitions. Clear ownership. Real lineage. They built the answer before they knew there would be a question. Now the question is in every boardroom.

I’m not background noise anymore. I’m the ceiling.

Many of the AI projects failing quietly inside enterprises right now aren’t failing because of the model. They’re failing before they reach the model. They’re failing on me.

I keep coming back to one comparison that feels right.

I’m more like language than oil.

Language doesn’t run out when you use it. Share it and it spreads. Use it carefully and your thinking gets sharper. Mix it with other language across time and distance and you get ideas neither source had alone.

But language misleads too. It’s been used to make harmful things sound reasonable. To protect the person holding it at everyone else’s expense. A confession and a cover story use identical raw material.

That’s me. Not a commodity. Not a deposit waiting to be drilled. Not yours, just because you hold me. Something that reflects back whatever care and intention you bring to it.

Is what’s happening to me justice? I don’t know yet. Four things will tell me. Whether the lawsuits force a real revenue share with creators, not settlement money that buys silence. Whether privacy rules survive their first contact with the agent era. Whether the organisations waking up to me now actually pay the people whose work and lives made me. Whether the agents now using me come with any audit trail back to a human. Until those answers come in, the verdict stays open.

What I become in your organisation isn’t a technology question.

Walk over to whoever is supposed to own me and ask four things. Where do I physically live. What shape am I in. Who do I describe, and did they agree to be described. Who would notice if any of that changed.

If they can’t answer in five minutes, you have your answer. That conversation is the work. Everything technical is downstream of it.