This is the lesson where we catch the overcorrection. Some people hear "outcome over technology" and proudly decide they never need to understand the tech. That's how you become the person who gets dazzled by any demo and bullied by any vendor. Outcome-first only works if you're also tech-smart.
The balance: data and judgment
Bezos is famously customer-obsessed — and he's just as insistent that you cannot ignore the technical, data side of the business:
"If you're in business and you're not looking at your data, trust me, your competitors are going to beat you. But if you're only looking at your data, you also will not win, or at least not win big."
— Jeff Bezos, Italian Tech Week 2025That's the whole module in one quote. Ignore the tech and data, you lose. Worship the tech and data with no human judgment, you also lose. The sweet spot — and your future job — is in the middle.
Know what the tools can actually do
You don't need to build AI models. You do need what YC's Charlie Warren calls model fluency: a current, realistic sense of what today's tools can and can't do.
"You need to know what frontier models can do today and design the product to ride the curve as they get better. There is no substitute for great tech here. People underestimate this."
— Charlie Warren, Y CombinatorHe even offers a sharp evaluation question he calls the "Sam Altman test": as the models get better, does your idea get stronger — or does the technology just swallow it? That single question separates a durable plan from a doomed one, and you can only ask it if you understand the technology's trajectory.
A peek under the hood: how modern AI actually "works"
Here's a mental model that will make you sound like you've been doing this for years. Most people picture AI as a magic chatbot — a "copilot" that makes you a bit faster. YC's Tom Blomfield says that picture is wrong:
"[Copilots are] basically just taking the old way of working and adding a more powerful engine onto it… instead, you can reimagine what a company is as a set of recursive, self-improving AI loops."
— Tom Blomfield, Y CombinatorYou don't need the math. You just need the shape of the loop, because almost every serious AI system is some version of it:
- Sense — it takes in information (emails, support tickets, sales data, a customer message).
- Decide — rules about what it's allowed to do on its own vs. what needs a human's okay.
- Use tools — it calls real things: a database, a calendar, a payment system.
- Check — a quality gate: safety filters, human review for anything high-stakes.
- Learn — it notices where it failed and feeds that back to the top, so next time is better.
Once you can see that loop, AI stops being magic and becomes a system you can reason about — and poke holes in. "What's the sensor? What can it do without a human? What happens at the quality gate when it's wrong?" Those are tech-smart questions, and you just learned to ask them.
"If it is recorded, it happened to the AI. If it did not get recorded, it did not happen to your intelligence."
— Tom Blomfield, on why AI is only as good as the data it can seeYou'll never out-code the engineers, and you don't need to. Your edge is a solid mental model of how systems work, so you can scope projects realistically, spot risks (bad data, hallucinations, privacy, security), and not get sold fairy tales. "Garbage in, garbage out" isn't a cliché — it's the #1 reason real AI projects fail. The MIS pro who understands data quality, the loop, and where humans must stay in the loop is worth their weight in gold.
A vendor demos a flawless AI tool. Which question shows you're tech-smart?
Telling real from hype
Bezos lived through the dot-com bubble and sees the same energy in AI. His take is wonderfully balanced: the hype is real and the value is real — your job is to separate them.
"Every experiment gets funded. Every company gets funded. The good ideas and the bad ideas… [but] AI is real, and it is going to change every industry."
— Jeff BezosAnd here's a tech-smart filter from Blomfield for deciding what's worth investing in: the flashy app on top is ephemeral — it can be rebuilt in a weekend as models improve. The durable, valuable asset is the data and know-how underneath. Being able to tell the throwaway layer from the lasting layer is exactly the kind of judgment that makes you useful.
The "it'll replace our whole support team" pitch
A startup promises their AI will fully replace your company's 12-person support team in 30 days, no humans needed. The demo is jaw-dropping. What's your tech-smart response?
You like the upside but you interrogate the loop. "Where does it get its answers — our help docs? How good and current are those?" (Garbage in, garbage out.) "What's the quality gate — what happens when a customer's question is rare, angry, or high-stakes?" "Who reviews and improves it?" Every credible expert in these talks keeps a human in the loop for the hard 10%. A claim of "zero humans in 30 days" isn't ambition — it's a red flag. You're not anti-AI; you're pro-reality.
Pick any AI tool you've used. Run it through the loop: what does it sense, what tools might it use, and where could it go wrong?
What's one piece of technology you've been treating as "magic" that you'd like to actually understand? What's stopping you?