lineage
A Love Letter to AI Skeptics
In Praise of Skeptics
I love skeptics. I’m a fan of the Challenger Sales (red) and Challenger Customer (blue) books. Applied their techniques before they were books, at VMware and Cisco.
In a nutshell: when we were selling new tech we SOUGHT the informed skeptic. They had their company’s best interests — and their career — at heart. They were well read. They had great questions. Everyone else in the company listened to them. If you won the skeptic, you won the deal.
Skepticism is the immune system of good thinking. It’s the thing that stops us buying snake oil, joining cults, and deploying to production on a Friday afternoon. Every hype cycle needs skeptics. They’re the ones who ask “but does it actually work?” while everyone else is pre-ordering.
The AI hype cycle needs skeptics more than most. There is real nonsense out there. Vendors promising magic. LinkedIn influencers who discovered prompting six months ago telling you it will replace your entire team. Executives who’ve never touched the tools making sweeping pronouncements about transformation. Products that are just a thin wrapper around an API call with a 400% markup. If your skepticism alarm is ringing, good. It should be.
So this isn’t a takedown of skepticism. It’s a love letter because I love skeptics. But only informed skeptics. Not blockers. Not talkers. And love letters are allowed to be honest, and there’s something honest I need to say.
Most AI skepticism I encounter isn’t earned. It’s inherited. Borrowed from headlines, absorbed from social media, reinforced by people who are also borrowing it. And it has a specific, identifiable shape: confidence without contact. Strong opinions, no hands on the wheel.
That’s not skepticism. That’s something else wearing skepticism’s clothes.
The Paradox
Here’s the thing I keep running into, and I can’t logic my way around it.
To be a credible skeptic of anything, you need to understand it deeply. You wouldn’t take seriously a restaurant critic who’d never eaten there. You wouldn’t trust a book review from someone who’d read the blurb. Credible criticism requires contact with the thing you’re criticising. This isn’t a high bar — it’s the minimum.
With AI, that minimum is harder to reach than people think. You can’t understand what these tools do by reading about them. You can’t understand them by watching YouTube videos. You can’t understand them by sitting in a conference session where someone demos Claude or ChatGPT for twenty minutes. Those things give you awareness. They don’t give you understanding.
AI is an experiential technology. It’s like cooking: you can watch every series of MasterChef and still not know what it feels like when the pan is too hot. You can read every article about prompt engineering and still not know the difference between a prompt that works and one that nearly works. The gap between reading about AI and using AI is not a small gap. It’s a canyon.
Which brings us to the paradox.
If you use AI deeply enough to understand it — if you push through the initial awkwardness, build real workflows, learn where it breaks, discover where it shines — you almost certainly won’t remain a skeptic. Not because you’ve been brainwashed. Because you’ve seen what it can do, and you’ve felt the shift in how you work. You might still have reservations (good — you should). You might still see limitations clearly (better — that’s expertise). But wholesale dismissal? That tends to evaporate on contact with reality.
So we have a paradox: the only way to earn your skepticism is to do the thing that would likely dissolve it.
I’m not trying to win an argument here. I’m describing a pattern I see over and over. The people with the strongest anti-AI positions are almost always the people with the least hands-on experience. And the people with deep experience almost always have a nuanced, qualified, “it depends” view — which is the opposite of skepticism. It’s informed engagement.
What Fluency Actually Looks Like
If you think I’m being vague about what “deep experience” means, fair enough. Let me make it concrete.
Zapier published their AI Fluency Rubric in March 2026 — a framework for evaluating how meaningfully people are using AI across six job functions: Engineering, Product, Support, Marketing, Sales, and People. It draws on Anthropic’s work on AI fluency, and it’s one of the most honest assessments I’ve seen of what competence actually looks like in practice.
The rubric defines four levels: Unacceptable, Capable, Adoptive, and Transformative. The labels alone tell you something — “Unacceptable” isn’t “beginner.” It’s the level where you’re using AI as “a lightweight assist inside a mostly unchanged workflow” and you “cannot clearly explain tool or model choices, limitations, or how your usage has evolved.” That’s not a starting point. That’s a plateau you’ve chosen to stay on.
Capable — one level up — requires that you “show real tool and model literacy: can explain why you use different tools for different tasks, where those tools break down, and how you’ve refined your workflows over time.” That’s not casual use. That’s deliberate practice. You can’t get there from YouTube.
But the line that stopped me cold is in the People row — the HR and leadership function. Under “Unacceptable,” it reads:
“Skepticism is untested, not informed. Actively blocks their team from experimenting with AI. No AI Builder setup, no enablement participation. Becomes a bottleneck for transformation work.”
Read that again. Zapier — a company that processes billions of automated tasks — looked at the landscape and decided that uninformed skepticism isn’t just unhelpful. It’s unacceptable. Not as a value judgment on the person. As an assessment of professional competence. In 2026, blocking your team from experimenting because of opinions you formed from reading articles is a failure to do your job.
That’s not me saying it. That’s an AI fluency framework, built on Anthropic’s research, from a company with a front-row seat to how work is actually changing.
The Canyon Between Reading and Doing
I keep coming back to why the paradox holds, and it’s this: AI is a tool, and tools are understood through use.
You can read every article about woodworking. You can watch a hundred videos of someone using a hand plane. You will still produce a rough, uneven surface the first fifty times you try it yourself. The knowledge isn’t in the article. It’s in the feedback loop between your hands and the wood.
AI is the same. The knowledge isn’t in the concept. It’s in the loop. It’s in knowing that your prompt didn’t work and feeling why. It’s in discovering that giving the model more context made it worse, not better. It’s in the moment when you restructure a workflow around AI capabilities you didn’t know existed three weeks ago — because you were using it every day and the possibilities became obvious through practice, not study.
The rubric measures this. It doesn’t ask “do you know what AI can do?” It asks “can you demonstrate that your work has changed?” Can you point to before-and-after? Can you show where AI is load-bearing in your process — not decorative? That’s fluency. And fluency comes from immersion, not observation.
This is why reading about AI and using AI produce fundamentally different conclusions. Readers see the failures, the hallucinations, the hype — because that’s what gets written about. Users see those too, but they also see the texture. The 80% of interactions where it genuinely helps. The workflow that went from four hours to forty minutes. The draft that came out better than they expected. The question they asked at 11pm that saved them a day of work.
Readers get the highlight reel of failure. Users get the daily reality of utility. No wonder they reach different conclusions.
The Invitation
So here’s my love letter, skeptics: I think you might be right. About some of it. The hype is real. The overselling is real. The risk of doing AI badly is absolutely real.
But you can’t know which parts you’re right about until you’ve done the work. And the work isn’t reading. It’s using. Deliberately, critically, with your skeptic’s eye wide open.
BECAUSED YOU MIGHT FIND A NEW THING BECAUSE OF YOUR UNIQUE DOMAIN EXPERIENCE.
Because that’s what AI does. It unleashes you, it you let it.
Look at the Zapier rubric. Find your role. Read the “Capable” column — not “Transformative,” not “Adoptive,” just “Capable.” Then ask yourself honestly: am I there? Have I put in the hours to have an informed view? Or am I forming conclusions from a distance?
- This is not a punitive system, it’s a “setting you free motherfucker” system.
If you do the work and you still have reservations — real, specific, experience-based reservations — I want to hear them. That’s the skepticism the world needs. Grounded, earned, and useful.
But if you haven’t done the work, your skepticism isn’t protecting you. It’s limiting you. And worse, if you’re in a position of influence, it might be limiting everyone around you.
The paradox isn’t a trap. It’s an invitation. Use the tools. Build the fluency. Earn the right to your opinion.
And if you come out the other side still skeptical? You’ll be the most valuable voice in the room.
Steve Chambers is the founder of Viewyonder, a GenAI consultancy pioneering people with AI at work. He’s been through enough hype cycles to respect the skeptics — he just wants them to be the informed kind. If your team is navigating AI adoption and the skepticism is slowing you down, get in touch.