Posted by Paul Graves, Founder of Elenthir AI

Over the past few months, I’ve been watching the AI space explode — and honestly, it’s been giving me serious 1996 vibes.

Back then, the internet felt new and terrifying. People compared it to Skynet. It was exciting, mystifying, and overwhelming all at once.

Now it’s happening again.

AI is the new frontier. Smart coders are having a party. Investors are falling over themselves to fund the next big thing. Some AI startups are hitting $1M+ ARR in just a few months. That traction is real, but it’s also fragile.

That’s why I entered this industry: not just to build cool things, but to help businesses cut through the noise and avoid what’s coming next for some.

Because if we’re being honest… this is starting to feel a lot like the dot-com bubble.

Remember Pets.com?

Back in the dot-com days, Pets.com ecame an overnight success — clever branding, fast growth, tons of attention. But the economics were flawed. Low-margin products, expensive logistics, limited retention. The company went public in early 2000 and filed for bankruptcy by November.

It became the case study for what happens when growth is prioritized over fundamentals.

Today, we’re seeing echoes of that in the AI space.

The Experimental Spending Era

A recent report from Menlo Ventures revealed that 60% of GenAI spend is still considered experimental. That’s a big red flag.

If you’ve got early revenue but no signal of proper retention or value realization, you don’t have product-market fit — you’ve got a false positive.

This is precisely what Mark Roberge (former CRO at HubSpot and founder at Stage 2 Capital) pointed out in a post that stuck with me. He emphasized the need to track Leading Indicators of Retention (LIR) — not just ARR or user growth.

What are LIRs?

Leading Indicators of Retention are real-time metrics that show whether users are truly getting value. Roberge recommends:

• A binary check on Ideal Customer Profile (ICP) fit

• A usage metric that correlates with customer-perceived value

From there, track usage on a monthly basis, not just at renewal windows.

Common LIR examples include:

• Static thresholds (e.g., 5+ key actions per month)

• % of predicted usage

• Progressive milestones (e.g., 10% in month 1, 25% in month 2)

These aren’t perfect, but they’re directional. And way better than getting surprised by churn after the hype fades.

What Gives Me Hope

The internet didn’t get its best strategists until after the bubble burst. But AI already has an army of sharp minds stepping up. Builders, thinkers, and evangelists are calling for balance now — before the crash.

That’s what gives me hope.

It’s also why I launched Elenthir AI: to help business owners and operators make smart, grounded decisions about AI adoption. Not every shiny tool is worth your time. And not every use case should be automated.

But the ones that are? Those can change everything.

Need help navigating AI without the hype?

We offer AI strategy & implementation for owners and operators who want real results. Book a free strategy call or shoot us a message.