I remember the early-2000s: I read about people in the US returning items they’d ordered online because the experience simply didn’t match expectations. At the time, the narrative was clear: “Online shopping won’t work — people want to touch, feel, try things.” Fast forward to today and we’re buying cars online.
The same happened with mobile-first applications: in the late 2000s people argued mobile hardware, screen size and connectivity made it impractical. Now you shoot movies on your phone. The cloud faced its critics — especially in regulated sectors like banking and healthcare, where security and compliance seemed insurmountable. Today it’s foundational.
I believe the Artificial Intelligence (AI) adoption curve is following the same path: a phase of intense hype → visible failures/pull‐backs → mainstream acceptance. And just as with e-commerce and cloud, it isn’t a switch from 0 to 1 but a gradual climb.
1. E-commerce: the early doubts and the slow build
When e-commerce first emerged, it faced two big issues: consumer trust and logistics. Online ordering around “2002” still had people reticent about paying without seeing the product, unclear return policies, slow shipping. That period coincides with the post-dot-com crash era when many online retailers had failed.
By 2000 the U.S. retail e-commerce sales were roughly US$27.6 billion, and by 2024 they’d grown to nearly US$1.2 trillion.
What changed?
Payment security improved (SSL, better gateways)
Logistics/methods matured
Consumer mindset shifted: “I can trust ordering online.”
Mobile and better connectivity kicked in.
Lesson for AI: early applications will disappoint because “the system” isn’t fully mature and user trust / expectations aren’t aligned.
2. Cloud computing: the regulated-industry hold-out
Cloud faced its own “won’t adopt” arguments — particularly in banking, healthcare, regulated industries. Concerns: data security, sovereignty, compliance, control of infrastructure. For example, research shows technical barriers like security and data portability can raise the likelihood of non-adoption by up to 26×.
Today: about 94 % of organisations globally use cloud computing in some form. And in healthcare and finance, adoption is accelerating.
So cloud moved from “we can’t” to “we must”. The shift happened not because the technology magically changed overnight — but because economics, operating model, regulation and appetite changed.
3. AI: the present cross-road
We’re where e-commerce was in the early 2000s or cloud in its early adoption phase. There’s massive excitement (and hype) about AI. But many projects will falter:
Expectations are too high: “We’ll replace 50% of jobs in 12 months” — unrealistic.
User readiness is uneven: organisations lack skills, cultural change, governance.
Economic/regulatory concerns: Who owns the model? Bias, ethics, data privacy.
The incremental nature gets overlooked: It’s not “all or nothing”.
Instead of “AI or bust”, we need to think “How do we adopt incrementally?”
4. How to move from hype to value (based on 25+ years building tech)
From my experience as a technologist and founder, here’s what helped:
Pick incremental use-cases: Choose “augment human work” (not replace) initially. Let’s get 10-20 % improvement, then build.
Lower expectation risk: Don’t promise 10× ROI from day 1. Frame as “we’ll reduce X-hours by Y%” then scale.
Build the capability & change the culture: Teams used to “build system, hand it over” must shift to “train, iterate, embed”.
Governance matters: Cloud taught us that compliance + security + operations must be baked in early. Same for AI.
Measure, learn, adjust: Monitor real uses, failures, feedback. Use that to create the next wave.
Mind the ecosystem: Just like e-commerce needed payments + logistics + devices, and cloud needed networks + services + trust — AI needs data, models, compute, usage flows, ethics.
Conclusion
Yes — AI will follow the same arc: intriguing novelty → sharp disappointments → business-as-usual.
If you’re working in software, architecture, or leading teams (like I have for two decades), the practical advice is: treat AI like a journey, not a flip-switch. Start small, prove value, build capabilities, invest in readiness (people + process), and don’t chase the “moonshot” as your first step.
Here’s a question to invite engagement:
What incremental AI use-case would you pick in your current organisation that’s low-risk but meaningful? I’d love to hear what others are trying—and how they’re treating the “small wins” mindset.
First published on LinkedIn.