Why “data is the new oil” falls apart in creative work
Founders love to say data is the new oil and the proprietary dataset wins. For infrastructure, sure. But in brand and creative work, more data just buys you more average. The real asset is a strategist willing to say the obvious thing nobody else will.
TL;DR
- “Data is the moat” genuinely holds for infrastructure. Mastercard, Plaid and the like own datasets no LLM can recreate.
- It breaks down in brand and creative work, where a company’s “proprietary data” is mostly old decks, an unread tone-of-voice doc, and a few years of campaign reports.
- Feed all of that to a model and the output is still average. Data volume doesn’t manufacture a point of view.
- The winning move is a strategist saying the obvious, slightly uncomfortable thing nobody else would. That judgment is the real asset.
- If your pitch is “we have more data,” you’re selling fuel in a race that was always about the driver.
Every founder I talk to lately is reciting the same catechism. Data is the moat. Data is the new oil. Whoever owns the proprietary dataset wins. It’s said with the confidence of something already settled, usually right before they ask why their shiny new AI features feel exactly like everyone else’s.
Here’s the uncomfortable part: the line is half right. And the half that’s wrong is the half most people are betting their entire differentiation on.
Where the data moat is real
For infrastructure, the cliché holds up fine. Mastercard sees a meaningful slice of the world’s card transactions. Plaid sits between thousands of apps and the banking system. That data is proprietary, it compounds, and no amount of model cleverness conjures it out of thin air. You either have the pipes or you don’t. An LLM can’t hallucinate ten years of real settlement data. So yes: in those businesses the dataset is the moat, and it’s a deep one.
If that’s the company you’re building, stop reading and go protect your data. The cliché was written for you.
Where it quietly falls apart
Then the exact same logic gets cut and pasted into brand and creative work, and it stops describing reality.
Be honest about what a brand’s “proprietary data” actually is. It’s a SharePoint folder of old pitch decks. A tone-of-voice document nobody has opened since onboarding. Three years of campaign performance reports that mostly confirm the obvious. Maybe a brand book and a Figma file with forty logo variations in it.
You can feed every byte of that to a model. Go ahead. The output will be competent, on-brand, and completely average, a confident synthesis of everything you’ve already done. Which is exactly the problem. Averaging your own past work is a very efficient way to produce more of your past work. It is the opposite of an edge.
The asset isn’t in the dataset
The call that wins a brand isn’t buried somewhere in the performance reports. It’s the strategist in the room saying the obvious thing nobody wanted to say out loud: the uncomfortable truth about who the company is really for, or the position everyone in the building sensed but kept hedging around.
That move doesn’t come from more data. It comes from judgment, taste, and the nerve to commit. You can’t fine-tune nerve. A model trained on a thousand safe decks will hand you a slightly smoother safe deck. It will never tell a founder the thing the founder needs to hear and is quietly paying, whether they’d admit it or not, to be told.
More data gets you a better average. The work that matters is precisely the part that refuses to be average.
Fuel versus driver
Which brings us back to the oil. “Data is the new oil” is a great line for a reason. Oil is valuable, it powers everything, you’d be a fool not to want it. But in creative work, data is fuel, and fuel has never won a race on its own. Everyone on the grid has fuel. Some of it’s better than others. None of it is the reason a particular car crosses the line first.
The driver is. The instinct to brake later than feels safe, to take the line nobody else sees, to commit when committing is the frightening option. That’s not in the tank. It’s in the seat.
So if your whole pitch is “we have more data,” be careful what you think you’re selling. In infrastructure, you might be selling the moat itself. In brand and creative work, you’re selling fuel in a race that was always about the driver, and the teams that win it are the ones who still have someone in the room willing to make the call. If that’s the kind of work you’re after, let’s talk.
Frequently asked questions
- Is data really a moat?
- It depends on the business. For infrastructure companies that accumulate proprietary, hard-to-replicate data, such as payments, banking connectivity, or logistics, data is a deep and durable moat. For brand and creative work it usually isn’t: the "data" is mostly past output, and more of it just produces more average work rather than a defensible edge.
- What does “data is the new oil” actually mean?
- It’s shorthand for the idea that proprietary data is the scarce, valuable resource that powers modern businesses, and that whoever owns the biggest dataset wins. The metaphor fits resource-and-infrastructure businesses well. It misleads in creative work, where judgment and taste, not raw data, decide the outcome.
- Can AI replace a brand strategist if you feed it all your data?
- No. A model trained on a brand’s old decks, guidelines, and campaign reports will return a competent synthesis of work that already exists, a smoother version of the average. It won’t make the uncomfortable, committed call that defines great positioning, because that comes from judgment and nerve, not from pattern-matching the past.
- What actually creates a competitive advantage in brand and creative work?
- A clear, committed point of view: the strategist willing to say the obvious thing nobody wanted to say, name who the company is really for, and commit to a position competitors can’t honestly copy. Taste, judgment, and the nerve to commit are the durable assets, not the size of the dataset.
- When does a data moat genuinely matter?
- When the data is proprietary, compounding, and impossible to synthesize from scratch: real transaction histories, network behavior, sensor data, or anything protected by network effects and direct access. In those businesses the dataset is the moat. In creative work, it’s fuel, not the finish line.
