AI makes it easier to move fast. It makes it easier to look busy. It even makes it easier to ship something that kinda works.
None of that means you crossed the chasm.
If you've read Crossing the Chasm, you already know the general idea: innovators and early adopters will buy your weird, half-baked thing because they like weird, half-baked things - guilty! Pragmatists buy outcomes. They want less risk, fewer surprises, and they want to be ahead of their competitors without the work.
AI doesn't change that. It just helps you reach the same wall with more velocity.
The chasm is still there. AI just makes you faster.
I worked with a physical product company that ran a successful crowdfunding campaign. The product was real. The "genius dev" behind it rode the hype straight into a FAANG job, leaving a crater where the software team used to be. That's where I came in.
Crowdfunding backers were classic early adopters. They tolerated rough edges, forgave bugs, and some of them got off on the chaos because it made them feel like insiders.
Then the company tried to pivot into SaaS and sell to mainstream businesses. It couldn't. Not because SaaS is easy (it isn't), but because the buyer changed.
Regular businesses wanted:
- integrations
- uptime
- support
- predictable delivery
- a reason to trust us next quarter, not just today
Aside: physical products are expensive and really difficult to build. You learn that the first time someone says "we should pre-order parts" and you realize you're betting real money on a number someone pulled out of their ass. Contract manufacturing negotiations can take up to a year — legal purgatory while your cash burns.
Sometimes the chasm never even shows up. The company just ends.
Years before that crowdfunding gig, I'd been kicking around an idea I called FindMeNoms — a map of experiences for people who love to travel, the wanderlust crowd. A startup in NYC was chasing something close to it: one designer co-founder, one UX-expert co-founder, and enough overlap with my idea that I picked up and moved to Williamsburg, Brooklyn, to go work on it out of a two-bedroom apartment.
That stretch put me in the room with a lot of TechStars NYC people, including the Digital Ocean guys back when Digital Ocean was still just a bunch of guys. Early adopter energy, wall to wall. Everyone in that apartment and every meetup around it was the exact reader Crossing the Chasm describes — excited by the idea itself, not by whether it worked yet.
It ended abruptly. "Hey, we need you to leave." I don't know to this day if it was me, them, or the dog. I bought a ticket back to Texas. I know the exact date only because I searched my own flight history: August 2nd, 2012 — La Guardia to Baltimore to Austin. I already had a gig lined up, moved in with my cousin in Round Rock, and moved on.
That startup never got the chance to hit the wall this post is about. It lived and died entirely on the early-adopter side of the chasm — TechStars energy, Brooklyn apartment, excited co-founders — and evaporated before a single pragmatist ever had the option to say no. No integrations to build, no uptime to guarantee, no procurement checklist, because there was never a mainstream buyer in the room to ask for any of it. AI wouldn't have saved that one either. The problem wasn't speed. There was no whole product yet, and there was no time left to build one.
Early adopters want the advantage. Pragmatists want it without the pain.
Early adopters will buy change because they want to be first. They want the advantage. They'll put up with you rewriting the docs every other day.
Pragmatists want the advantage too. They just don't want to be your beta testers.
This is where founders blow it. They take early adopter excitement and mistake it for mainstream demand. Then they build a roadmap around whatever the loudest power users yell about and pretend that's product-market fit.
I've watched this happen inside companies with a CEO who had a full-on cult of personality. People knew the plan was wrong. They did it anyway. The social pressure was stronger than the facts. Ask too many questions and you're "negative," which means you're not drinking the Kool-Aid fast enough.
That dynamic turns magical thinking into company policy. If the leader believes we're already mainstream-ready, the team starts acting like implementation details are optional.
Those details are the product. Everything else is just a demo that expires the second someone tries to use it for real.
Where AI actually helps (and where it doesn't)
AI accelerates the work. It doesn't do the work for you.
Four ways I've used it without torching the go-to-market plan:
Segment maps
Use the model to map your current users onto the adoption curve. Not the fake version where everyone's an "innovator." The real version where you actually look at buying behavior: Do they tolerate missing features? Do they self-implement? Do they accept "email me and I'll fix it" as support?
If yes, you're still in early adopter land.
Message variants
Visionaries and pragmatists are compelled by totally different messages.
AI can generate variants fast, but you still have to pick the truth and stick to it.
Visionary pitch: "This changes how you operate."
Pragmatist pitch: "This reduces X cost by Y% and here's how we implement it without breaking Z."
If you try to sell pragmatists on "change the world," you'll get lip-service and zero signed contracts.
Customer call patterns
AI can summarize customer calls and spot patterns.
But you have to ask the right question: are users praising the outcome or just geeking out over the novelty?
Early adopters say things like: "This is sick, I can't believe it works."
Pragmatists say things like: "Who else is using it, and what happens when it fails?"
Objection libraries
Pragmatists ask predictable questions: What's the switching cost? What breaks during integration? How long does security review take? Who handles procurement? What happens when support gets slammed? Who takes the blame if this thing causes a mess? AI can draft responses and build you a checklist. It can also write that "here's how we implement this" doc you've been putting off.
It can't make procurement move faster or suddenly decide vendor risk doesn't matter.
Where you still need boring software
Every time I build an AI system for a client, the same thing happens: they want the model to be the system of record. Wrong.
Use normal software for the stuff that needs to be traceable. CRM records with real accounts, stages, notes, close dates, lost reasons. Pipeline reporting with opportunity aging, conversion rates, sales cycle length. Evidence repositories with call recordings, transcripts, contracts, support tickets. Segmentation analysis in SQL or spreadsheets where the math happens in daylight and you can actually audit it.
Without this, you're not data-driven. You're just making up stories about your pipeline and praying nobody asks to see the numbers.
The parts you can't automate (because they're human)
Strategic focus is still on you. You still have to pick which segment gets the next six months of your life. If you try to serve everyone, you'll end up with a product early adopters love and a sales pipeline full of "great call, let's circle back."
References don't come out of a model either. You earn them by delivering outcomes, repeatedly, with boring consistency.
And you have to read risk yourself. Pragmatists aren't afraid of your model architecture. They're afraid of being the person who bought the tool that caused a mess.
AI can help you write, analyze, prototype, and ship faster. But AI-powered productivity doesn't replace founder judgment when deciding which customers to serve or what proof they need.
If you're still selling to visionaries while pretending you're ready for pragmatists, you'll hit the chasm at full speed and wonder why nobody's buying.
-Sethers