The Two Kinds of AI Hype You Should Ignore

Most small business owners are getting AI advice from two groups of people who have never run a small business. One group sells AI. The other group writes about AI. Both are loud, both are confident, and both will steer you into bad decisions if you let them.

The signal you actually need is buried under their noise. So let's name the two species of hype, learn to spot them, and talk about what to pay attention to instead.

Vendor hype: "our model is best"#

Vendor hype is the easier one to recognize because it has a price tag attached. Every AI company is currently claiming their model beats every other model on some benchmark you've never heard of. GPT beats Claude. Claude beats Gemini. Gemini beats both on a Tuesday. A new open-source model from a lab you can't pronounce just topped the leaderboard.

None of this matters to your business.

Benchmark performance is a measure of how well a model does on standardized tests. Your customer service queue is not a standardized test. Your invoicing workflow is not a standardized test. The difference between the #1 model and the #5 model on most benchmarks is a few percentage points on tasks that have nothing to do with what you'd use it for.

What vendor hype obscures: integration cost, reliability over time, how the model behaves on your specific data, what happens when it fails, and what you're actually locked into when you build on their API. A model that's 3% "smarter" but rate-limits you at peak hours is worse than a slightly dumber one that just works.

The tell for vendor hype is that it talks about capability in the abstract. "Our model can reason across 200 pages of context." Cool. Can it reliably extract line items from the specific PDF format your three biggest suppliers send you? That's the question. Nobody on a keynote stage is going to answer it.

Trend hype: "everyone is doing X"#

Trend hype is harder to spot because it doesn't come with an obvious sales pitch. It comes from articles, LinkedIn posts, conference talks, and a friend-of-a-friend who heard about a company that automated everything and now runs on four employees.

The pattern is always the same. A handful of impressive case studies — usually from venture-funded startups or Fortune 500 pilots — get aggregated into a trend piece. The trend piece gets quoted in five more articles. By the time it reaches you, "some companies are experimenting with autonomous agents" has become "every business needs an agent strategy by Q3."

The problem isn't that the case studies are fake. They're real. The problem is that they're not your business. A fintech with 80 engineers shipping an autonomous research agent tells you nothing about whether you, with a team of twelve and one part-time bookkeeper, should build one. A retailer with a million SKUs deploying personalization AI is not a template for your 400-product Shopify store.

Trend hype trades on FOMO. The implicit message is: if you're not doing this, you're falling behind. The honest version is: most of the businesses "doing this" are doing pilots that haven't proven out yet, and a lot of them will quietly kill those pilots in eighteen months when the bills come due.

The actual signal: what's shipping in production at companies your size#

Here's the filter that cuts through both kinds of hype. When you hear about an AI use case, ask two questions:

First, is this in production, or is it a demo? Production means it's been running for at least six months, handling real volume, with real consequences when it fails. Demos and pilots tell you almost nothing. Plenty of impressive demos never survive contact with messy real-world data.

Second, is the company that's running it shaped like mine? Same rough headcount, same rough revenue, same industry constraints. A 12-person professional services firm and a 12-person ecommerce brand have wildly different automation profiles. A 200-person manufacturer and a 200-person SaaS company might as well be on different planets.

This filter is brutal, and it's supposed to be. It eliminates probably 90% of the AI content you'll encounter. What's left is much more useful: actual peers, actually shipping, with actual results you can replicate or learn from.

Where do you find this signal? Industry-specific communities and forums, usually. Trade associations. Customers of the same vendors you're considering — call them and ask what's actually working. Operators who've been doing this for a couple of years and have nothing to sell you. None of it is glamorous, and none of it gets written up in Forbes.

The counterargument, and why it's wrong#

The pushback I get on this is reasonable on its face: "If I only pay attention to what's already proven at companies my size, I'll be permanently behind. I'll never be early on anything."

Fair. But here's the thing about being early: in AI specifically, early is usually expensive and almost never decisive. The companies that built custom GPT-3 integrations in 2021 mostly threw that work away when GPT-4 came out. The companies that built bespoke RAG pipelines in early 2023 watched off-the-shelf tools catch up and pass them by 2024. Early adopters in this space have been repeatedly punished, not rewarded.

The businesses winning with AI right now aren't the ones who moved first. They're the ones who waited until a use case was clearly stable, then implemented it cleanly and integrated it deeply. They picked their spots. They didn't try to be on the frontier of everything.

For a small business, the right posture is roughly 18 months behind the frontier. Let the venture-funded companies eat the cost of figuring out what works. Then copy them, cheaper, with better tools, on a clearer set of assumptions.

What to actually do this quarter#

Ignore the model benchmark wars. Pick whatever model your eventual implementation partner recommends and move on. The difference won't matter to your outcomes.

Ignore the trend pieces. If an article doesn't name specific companies of your size shipping specific things in production, it's not information — it's mood.

Instead: write down your three most expensive recurring workflows. The ones that eat the most hours, generate the most errors, or block the most growth. Then ask whether anyone shaped like you is already automating any of them successfully. If yes, copy them. If no, wait. There will be plenty of time.

If you want a sober second opinion on which of your workflows are actually ready for automation and which ones are still in hype territory, that's what our discovery process is for.