What Twenty Years of Marketing Taught Me About AI Hype

I started in marketing in 2006, working on product lines at GNC that had to survive on a shelf next to a hundred competitors.

Since then I have been told, with total confidence and roughly every three years, that everything was about to change. Here is the actual list, in order, with what each one promised and what it delivered.

AI marketing is not overhyped in what it can do, but the hype misidentifies what changed. Previous marketing technology cycles delivered real capability while failing organizationally, because businesses bought tools without changing process. AI follows the same pattern with one genuine difference: it collapsed the cost of production labor, which changes what is possible for small organizations specifically.

The pattern underneath all of it is more useful than any individual prediction.

The cycles, in order

Search and the early web, roughly 2006 to 2010

The promise: every business needs a website and a search presence, and the ones who move first will own their categories.

What happened: broadly true, and the first movers did benefit lastingly. It also produced a decade of businesses paying for websites that were brochures nobody visited, because having a site and being findable turned out to be different problems.

What it taught: presence is not the same as visibility. That distinction still catches businesses out today.

Social media, roughly 2008 to 2014

The promise: direct relationships with your audience, unmediated, free.

What happened: it worked genuinely well for a while. During my independent consulting years around 2012 to 2015 I grew social content views for clients by figures that sound absurd now, because organic reach was real and the space was uncrowded. Then the platforms did what platforms do. Reach was throttled, audiences you built became audiences you rented, and the free channel became a paid one.

What it taught: never build your business on land you do not own. This is why email lists still matter, and why they will still matter in 2036.

Marketing automation, roughly 2012 to 2018

The promise: nurture leads automatically, score them, and hand sales only the ones worth calling.

What happened: I lived this one closely. At Simio starting in 2015 I implemented Dynamics, Salesforce, Marketo, and Act-On, building the company's first real digital marketing foundation. The technology worked exactly as advertised. Across the industry it mostly failed anyway. Companies bought platforms and never built the content to put in them, never agreed with sales on what a qualified lead was, and never assigned anyone to own it. The software sat there generating a monthly invoice.

What it taught: the tool was never the constraint. The process was. This is the single most repeated lesson on this list.

Content marketing and inbound, roughly 2013 to 2019

The promise: publish useful content, earn attention, stop interrupting people.

What happened: correct in principle and completely swamped in practice. Everyone published, volume exploded, quality collapsed, and the businesses that won were the ones who committed for years rather than months.

What it taught: the strategy was right and the execution horizon was wrong. Most businesses quit at month four, which is precisely when nothing has happened yet and everything is about to. If that sounds familiar, it should. It is the same shape as what is currently happening with AI-generated content, at higher speed.

Martech consolidation, roughly 2016 to 2022

The promise: integrate everything, get a single view of the customer, make decisions from real data.

What happened: the stack got enormous and mostly did not integrate. At Simio during my VP years and later at Orange Logic and TROY Group, a meaningful share of my job was untangling systems that had been bought individually and never designed to work together.

What it taught: every tool you add has a maintenance cost nobody accounts for at purchase. Small businesses feel this most acutely, because there is no operations person absorbing it.

Account-based marketing, roughly 2017 to 2022

The promise: stop chasing volume, target the specific accounts worth winning.

What happened: genuinely effective when done properly. We used ABM at Simio to get onto enterprise software standards lists at major organizations, which is the sort of thing that does not happen by casting a wide net. It also became a label attached to ordinary outbound, which diluted the term until it stopped meaning anything.

What it taught: good strategies get commoditized by vendors, and once a term becomes a product category you have to look past the label at what is actually being done.

AI-driven advertising, roughly 2022 to 2024

The promise: hand the machine your budget and your creative, and it will allocate better than you can.

What happened: at TROY Group, restructuring paid search toward Google Performance Max meaningfully reduced customer acquisition cost. It worked. It also removed visibility. You get results and less understanding of why, which is a real trade rather than a free win.

What it taught: automation that improves outcomes while reducing insight is worth taking, and worth taking with your eyes open.

The pattern

Look at that list and the same shape appears in almost every entry.

The technology usually worked. In nearly twenty years I can think of very few tools that failed to do what they claimed. Marketing automation nurtured leads. Social reached people. Content earned attention. ABM landed accounts.

The failure was organizational. Companies bought the tool and did not change the process, did not assign an owner, did not build the inputs it needed, or quit before the horizon the strategy required.

The winners were rarely the first buyers. They were the ones who restructured how work happened to match what the tool made possible. That is a much less exciting sentence than the ones on vendor websites, which is precisely why it keeps being ignored.

Every cycle produced a category of casualties who concluded the technology was overhyped, when what actually happened was that they bought it and changed nothing.

I would bet a significant amount that in 2029 there will be a large group of businesses who tried AI, saw no return, and concluded it was a bubble. They will be describing their own implementation and attributing it to the technology.

So is AI overhyped?

The honest answer, in two parts.

The claims about what it can do are broadly accurate, and that is unusual. Most cycles oversold capability. This one, if anything, is being undersold in the specific areas where it is strongest, because the marketing focuses on flashy demonstrations rather than on the boring, high-volume, research-and-drafting work where it genuinely excels.

The claims about what it means are wrong. The framing is that AI changes strategy, that marketing is now fundamentally different, that everything you knew is obsolete. That is the part being oversold, and it is the same oversell as every previous cycle.

Positioning still matters. Knowing your customer still matters. Consistency still beats intensity. Measurement still separates the businesses that improve from the ones that guess. None of that has been repealed and none of it will be.

The one thing that is genuinely different

I want to be precise about this, because it is the part worth taking seriously.

AI collapsed the cost of production labor. That is the change. It is narrower than the hype suggests and more consequential than it sounds.

Consider what running a real marketing operation used to require: someone to write, someone to build and send, someone to design, someone to handle technical work, and someone to make sense of data. Even a lean version is three or four people. Three or four people costs more than most small businesses make in profit.

That is why small businesses never had real marketing systems. Not because the strategy was secret. It has been documented and taught for decades. The labor was unaffordable.

That constraint moved. One person with judgment and the right tools can now build and run what previously required six.

For a large enterprise this is an efficiency gain. Meaningful, not transformative. For a business with one marketing person or none, it is the difference between having a marketing function and not having one. That is why this cycle matters more for small businesses than for anyone else, and it is almost the opposite of how it is being discussed.

What to do with all of this

Assume the technology works and the implementation is the risk. Twenty years of evidence supports this. Spend your worry budget on process and ownership, not on tool selection.

Do not buy before you know what you are pointing it at. The most reliable predictor of a failed implementation, in every cycle, was buying first and deciding second.

Expect a longer horizon than you are promised. Every strategy on that list took longer than the pitch said.

Change the process, not just the tooling. This is the one lesson that appears in every single cycle, and it is the one most consistently ignored.

Be suspicious of the label. When a genuinely good idea becomes a product category, vendors attach the term to whatever they were already selling. This happened to inbound, to ABM, and it is happening to AI right now.

The part I would want a business owner to hold onto

I have watched a lot of confident people be wrong about what was about to change, and I have been one of them.

What I have never seen fail is this: a business that knows exactly who it serves, is findable when those people look, captures their information, follows up consistently, and measures honestly. That worked before any of these tools existed. It works with all of them. It will work after whatever comes next.

The tools change what is affordable. They have never changed what is true.

The five things, in order, are here. More about the reasoning behind this company is on the about page.