Lessons from the Dot-Com Era

Do you remember the dot-com boom?

Maybe you lived it, maybe you only read about it. Back then, you could slap a “.com” at the end of your name, raise millions, and get listed on Nasdaq—no product, no profits, just hype. Fast forward to 2025, and you swap that suffix for “AI.” The vibe feels oddly familiar, right?

MIT’s GenAI Divide report is basically the bucket of cold water. It says 95% of enterprise AI pilots delivered no measurable impact, despite billions poured into them. That’s the Pets.com moment of this generation. But let’s not get too gloomy—because if history rhymes, it rarely repeats exactly. The internet didn’t die in 2000; it regrouped, matured, and eventually ate the world. AI is on the same path, only the tempo is faster and the ingredients more mature.

The Divide Right Now

Here’s the crux: a small group of companies is extracting value from AI, but most are spinning wheels. The real bottleneck isn’t hardware or regulation—it’s learning. Most AI systems don’t get better in the flow of work. They don’t adapt, don’t remember, don’t capture feedback. People try them once, shrug, and go back to the old ways. Call it the graveyard of pilots.

Meanwhile, outside the official projects, employees are already using AI tools on their own. That “shadow AI” is spreading like wildfire. Think back to how Zoom spread in 2020: first individuals, then teams, and finally entire companies. That’s the same bottom-up pressure AI is seeing today.

Echoes of the Dot-Com Days

• Then (late ’90s internet): billions of dollars spent, few real users, bandwidth too slow, infra too immature.
• Now (mid-2020s AI): billions spent, shiny demos everywhere, but most workflows don’t stick. Memory, evals, governance still shaky.

The internet bubble popped, but after broadband, better infra, and cheaper compute came along, the second wave gave us Amazon, Google, Salesforce. The AI cycle looks similar, but faster. Why? Because unlike 1999, workflows are already digitized, cloud infra is everywhere, and consumers are already pulling AI into their daily lives. Everyone with a phone has touched AI—there’s no going back.

How Adoption Really Happens

1. Bottom-up: Task Copilots.
   AI slips into everyday tasks: writing drafts, checking QA, helping with analysis, answering customer questions. It saves time, cuts costs, improves quality—slowly but surely. Not evenly, not everywhere, but enough to stick. Employees get comfortable with AI before anyone makes it “official.”
2. Top-down: Strategy-Coated Business.
   Once those small wins accumulate, leadership steps in. They start layering AI across functions—sales, service, ops, planning. Governance and monitoring turn scattered usage into strategy. And because the workforce already trusts the tools, adoption feels natural.

Both these paths build something deeper: the Infinite Knowledge Legacy (IKL). That’s every corrected answer, every edited contract, every labeled piece of data feeding back into a company’s knowledge base. Over time, that data wealth becomes as important as revenue—it’s company wisdom on autopilot.

What’s Next

• Short term (1–2 years): Pilots stop failing at such high rates. Tooling for memory, feedback, and monitoring matures. Copilots become reliable in narrow lanes. Companies formalize shadow AI into official policies.
• Medium term (3–5 years): Task copilots grow into function-level strategies. CFOs stop asking for “demos” and start asking for ROI. Clear winners begin to emerge.
• Long term (5–10 years): Just like the internet era, most vendors disappear while a handful dominate. Companies with IKL sprint ahead; laggards lose options and agility.

Wild card? Agentic supply chains. Imagine AI agents transacting with each other—negotiating contracts, reconciling invoices, running workflows end-to-end. Sounds crazy now, but so did “buying books online” in 1997.

The Takeaway

History rhymes. AI hype will crash, but the wave that follows will be the one to watch. This time, the mix is richer: digitization already done, pandemic habits in place, consumer pull undeniable. AI isn’t waiting for company approval—it’s already in people’s hands.

For businesses, the play is simple: don’t wait. Start small and bottom-up, track impact, learn fast. Then coat the business top-down, building your IKL as you go. The companies that learn—not just spend—will be the ones shaping the next 20 years of digital history.

And if you’re a veteran of the dot-com days?

You’ve seen this movie before.

Only this time, it’s running on Nx speed.