Telephones let demons into your home. Trains caused “railway madness.” Every transformative technology arrives wrapped in a moral panic — and AI is the latest entry on the list.
But this panic is different. You can’t wait it out. It’s sitting on your desk as nine decisions you have to make this quarter: a pilot pitch from the CTO, a vendor demo your CFO loved, a board member forwarding a Substack about agents, three separate teams asking for OpenAI seats. You don’t have time to read another 3,000-word post telling you AI will either save the company or end it.
What you need is a triage heuristic — a five-minute read on any proposal that tells you whether it’s a reversible experiment, a strategic commitment, or a distraction dressed up as urgency. Here’s one. Use it, fork it, throw it out. But stop reading hot takes and start filtering.
Your problem isn’t AI. It’s decision cadence.
Start by seeing the noise for what it is.
Open any feed, any newsletter, any think-tank brief, and you’ll find two AI narratives published in the same week. One ends scarcity, cures disease, and unlocks a decade of compressed scientific progress. The other ends democracy, employment, and possibly the species. An executive’s morning reading looks like this:
Post one: “AI just changed everything about marketing forever. If you’re not deploying agents this quarter, you’re already extinct.”
Post two: “Generative AI will hollow out 40% of knowledge work by 2027. The reckoning is here.”
I pulled 84 pieces of major AI commentary published between May 2024 and May 2026 and sorted them by how they framed the technology. You got me — I didn’t. I asked Perplexity to do it. Here’s what it found: 39 percent treated AI as a destroyer, 25 percent treated it as a savior.
That split is the latest verse of a very old song. New technology, supernatural suspicion:
Photography was feared as a way to steal part of a person’s essence.
Comic books, in the mid-20th century, were blamed for juvenile delinquency and moral decay.
Rock music was denounced, by turns, as satanic, rebellious, and psychologically corrupting.
Video games were charged with causing violence, addiction, and social decline.
Social media triggered fears of addiction, brain damage, and the breakdown of social order.
Ok, that last one turned out to be mostly true. But it’s yesterday’s panic. AI is today’s — and the breathless-versus-apocalyptic split is the noise you’re triaging against.
Here’s the thing: neither narrative helps you say yes or no by Thursday. “AI changes everything” and “AI is a bubble that will end careers” are equally useless at 9 a.m. with a contract in your inbox. That’s the actual problem. Not AI. The decision cadence.
The four-quadrant triage
So here’s the filter. Score every AI decision on four axes, 1–5, in under five minutes. The axes aren’t arbitrary — each one targets a specific failure mode that sinks AI investments: irreversibility, slow feedback, organizational resistance, and false urgency.
Add the four scores. That’s the entire framework.
16–20: Green-light it today. High reversibility, fast signal, team wants it, real competitive pressure. These are your no-brainers. You should already be saying yes faster than this.
11–15: Run it as a bounded experiment. Define what success looks like, set a kill date, name the owner. Most AI decisions live here.
6–10: Slow down. Something is off — usually org-readiness or reversibility. Don’t kill it; redesign it. Can you shrink the scope until it scores higher?
4–5: Decline, in writing. Document why. You’ll see this proposal again in three months from someone else, and you want the receipt.
The scoring is coarse on purpose. A 1–5 scale forces a judgment call instead of analysis paralysis, and the additive math means no single axis can hijack the decision. A proposal that scores a 5 on competitive necessity but a 1 on reversibility still lands in the “slow down” band — exactly where it belongs.
A worked example: the coding assistant
Theory is cheap. Run the framework on a real decision.
Your CTO wants to deploy an AI coding assistant org-wide: annual contract, $40 per seat, 400 engineers. Score it:
Reversibility: 3. You can cancel the contract, but you’ll have engineers who’ve built habits around it and code that was written with it. Not catastrophic, not clean.
Time-to-value: 4. You’ll see commit-velocity changes inside a month.
Org-readiness: 5. Engineers are already paying for it personally. They’re asking.
Competitive necessity: 3. Competitors are doing it. Not a moat — table stakes.
Total: 15. Bounded experiment. Roll it out to 80 engineers across three teams for 90 days. Define the kill metric up front — no measurable velocity change, or a security incident — and decide on day 91.
Now run the same five minutes on the vendor demo your CFO loved — the one that promised to “transform” your finance close with agents. It scores 7. Reversibility is low: the data integration is permanent. Time-to-value is six months. No one in finance has asked for it. Your competitors aren’t doing it either. The CFO loved the demo; the decision still scores 7.
That’s the framework working. It overrides enthusiasm — yours and your team’s — with structure.
Why this beats the takes
The framework isn’t just better than the hot takes. It’s the only thing that survives the meeting.
About 48% of executives describe their AI experiences as disappointing, according to Infor’s Enterprise AI Adoption Impact Index. That number didn’t come from the people who said no. It came from the people who said yes without a filter. They didn’t lack ambition — they lacked a way to tell a pilot from a commitment before signing the contract.
The deliverer-versus-doom binary you see on LinkedIn isn’t an information problem. It’s a decision-architecture problem. Hot takes don’t travel into board decks; frameworks do. The CTO who points to a scoring rubric on a slide wins the meeting against the peer quoting social media.
And the ground is shifting under the takes themselves. LinkedIn is actively suppressing the recycled thought-leadership patterns — the “it’s not X, it’s Y” cadence — that dominated 2024. YouTube’s CEO just named “managing AI slop” a 2026 priority. The originality bar is rising on every surface that matters. A named, repeatable framework — even an imperfect one — is structurally more durable than another take about whether AI is overhyped.
The one rule that matters
If you take nothing else from this piece, take the calibration error underneath every bad AI decision.
The cost of a reversible AI mistake is almost always lower than the cost of waiting six months to decide.
Most executives have this backwards. They treat pilots like commitments and commitments like pilots — agonizing over a 90-day rollout to 80 engineers while rubber-stamping a multi-year platform migration. The framework above exists to flip that: to make reversibility a first-class input to the decision rather than an afterthought.
So score the next decision on your desk. If it takes longer than five minutes, you’re probably overthinking it. If it takes less than one, you’re underthinking it. Five minutes, four numbers, one decision.
Your next move: pull the three AI decisions sitting in your inbox right now and score them before your next meeting. The ones that clear 16 — send the approval today. The ones under 10 — send the decline today. The rest go on a one-page experiment brief by end of week.
You’ll get more done in the next hour than the last three weeks of reading about AI got you.