#28: Enterprise AI Adoption Tips from An English Literature Prof
Enterprise AI Field Notes: Prof. Geoffrey Moore wasn’t an MBA, wasn’t a Valley insider, wasn’t a Drucker-or-Porter grand theorist. Yet, he said - way back then - something useful to apply in the AI Ad
Geoffrey Moore was an English professor who taught medieval and Renaissance literature before wandering into Silicon Valley and rewriting Tech’s go-to-market (GTM).
And yet, he handed Tech its most useful metaphors: the chasm, the bowling alley, the tornado - way back in 1991. Principles that apply to today’s enterprise technology world.
I happened to re-read his seminal book, Crossing the Chasm, over this weekend, and the uncanny resemblance to our goings-on on Enterprise AI adoption is hard to miss. And a more recent 2021 book too.
Here is a 1991 classic that reframed adoption as a chasm (not a slope, we’d intuitively think it is a slope!) between early adopters and the early majority.
“Products die there in the chasm because pragmatists demand proof, not promise.”
Reading this pull quote, you might even say, as I did, “..well, obvious, isn’t it?” but you have to just count the bodies to see why it is not that obvious.
Some field notes follow.
[Ps - For the future, I may also draw on Rogers (diffusion), Kotter (change), Deming (quality), Christensen (disruption, JTBD), McGrath (discovery-driven strategy), Rackham (SPIN), Fogg (behavior), Eyal (habits), etc., to draw lessons from them towards AI Adoption inside enterprises. Why? Very simply, a useful non-insider lense helps us stay above the AI coalface and still see the whole mountain.]
1) Whole product beats clever tech
Moore (1991): Pragmatists don’t buy parts; they buy a whole product. Things such as core tech plus everything that makes it safe and useful (integrations, support promise, standards, references etc.). That’s how you cross.
2025 Enterprise AI: For us enterprise AI practitioners, it would mean treating “complete” as being model + workflow wiring + data contracts + RBAC + monitoring + runbook + ROI reporting. If Legal can indeed wave an audit pack and Ops has a playbook, you’re in pragmatist territory.
2) Beachhead → Bowling Alley
Moore: Win one niche with overwhelming force, then topple adjacent pins. Horizontal Day 1? Chances are you fall into the chasm.
2025 Enterprise AI: Pick one named workflow (claims triage, AML alert disposition, tier-1 contact deflection, such..). Give that 90–180 days, CFO-grade metric, owner on the hook. Adjacent use cases only after the first pin falls.
3) Evidence beats evangelism
Moore: Pragmatists copy peers. Share References > Talk rhetoric.
2025 Enterprise AI: Publish before/after, uptime, exceptions, savings. Land 2–3 named references in the same vertical. Internal wikis with gritty numbers beat glossy decks.
4) Positioning shift: revolutionary → reliable
Moore: Early adopters buy vision; early majority buys risk reduction, ROI, and fit.
2025 Enterprise AI: Speak P&L. Define baseline → identify the delta → show ROI impact. “Rework −20%, backlog −40%, payback in 2 quarters.” No “SOTA*,” no “tokens per sec” BS. Pragmatists buy outcomes they can defend. (*fancy 4-letter for state of the art)
5) Risk is the real chasm
Moore: The gap exists because pragmatists fear breakage, lock-in, and blame.
2025 Enterprise AI: Make governance a feature: lineage, red-team tests, bias checks, HITL gates, drift alarms, rollback drills. Risk signs off before go-live, not after. That’s the ticket across.
6) Over-resource the crossing
Moore: Crossing costs more than expected. Especially in the messy middle of enterprise AI adoption.
2025 Enterprise AI: Budget for process redesign, L&D, role remapping, MLOps hardening. The chasm taxes people and processes more than GPUs. Starve this and you stall.
7) Post-chasm behavior matters
Moore (tornado era): After you cross, scale operations and service quality; keep references compounding.
2025 Enterprise AI: Weekly Run Review (incidents, guardrail hits, savings realized). Keep proof fresh; widen the bowling alley with adjacent, similarly governed workflows.
Well, so finally.
Moore’s “obvious” only looks obvious until you count the bodies.
Crossing the AI chasm = industrialize context around the model, win one boring, referenceable workflow, then let evidence do the scaling.
Share which of these 7 points make more sense in your enterprise context!
Till next week, happy reflecting :)