The Word "AI" Is a Marketing Scam That Lets Vendors Charge You 3x
By Don Ho, Co-Founder & CEO, Kaizen AI Lab
Published: January 28, 2026
TL;DR: Vendors slap "AI-powered" on products and charge 20-40% premiums for technology that would have been called "automation" three years ago. I've evaluated dozens of these tools for clients. Most are basic rules engines with a chatbot glued on top. The definitional ambiguity around "AI" isn't just a philosophical problem. It's a pricing scam, a governance gap, and a regulatory minefield. Here's how to stop paying the AI tax.
---
I Evaluated an "AI-Powered" Contract Review Tool Last Month. It Was an If-Then Statement.
A client asked me to evaluate an "AI-powered contract analysis platform" they were considering. The vendor's pitch was impressive. Machine learning. Natural language processing. Intelligent risk scoring. The demo looked sharp. The sales team used the word "AI" fourteen times in a 30-minute call. I counted.
We got access to the trial environment. My team spent two days testing it against contracts with known issues, documents we'd already reviewed manually so we had a baseline.
The "intelligent risk scoring" was a keyword matching system. If the contract contained the phrase "indemnification" or "limitation of liability," the system flagged it. If those phrases were absent, the system missed the risk entirely. No contextual understanding. No analysis of what the indemnification clause actually said. Just pattern matching on specific strings.
The "natural language processing" was a search function. You could type questions in plain English, and the system would find the paragraph containing the most relevant keywords. Useful, sure. Also available in every PDF reader released since 2019.
The "machine learning" component? A categorization model that sorted contracts into types (NDA, MSA, SOW) based on header text and document length. We tested it with a non-standard contract format. It classified an equipment lease as a non-disclosure agreement.
The vendor wanted $4,499 per month. A comparable document management system without the AI branding runs about $1,200.
That's a 275% premium for a keyword matcher with a chatbot interface. And the client was about to sign the annual contract.
John McCarthy Predicted This 70 Years Ago
John McCarthy coined the term "artificial intelligence" in 1956. He later made an observation that perfectly describes the 2026 enterprise software market:
"As soon as it works, no one calls it AI anymore."
The flip side is equally true: as soon as you call it AI, you can charge more for it.
McCarthy was describing the "AI effect," the tendency for technologies to lose the AI label once they become commonplace. Spell check was AI. Voice recognition was AI. Spam filtering was AI. Route optimization was AI. Search ranking was AI. All of these felt like breakthroughs when they were introduced. Now they're features. Nobody pays a premium for email spam filtering because it's "AI-powered."
But the cycle resets with every wave of AI hype. In 2026, we're at peak label inflation. Vendors who were selling "automation platforms" in 2023 are selling "AI-powered intelligent automation" in 2026. Same product. New label. Higher price.
I've evaluated AI tools for clients across lending, legal, real estate, food and beverage, and professional services. The ratio of genuine AI capability to marketing AI is roughly 1 to 4. For every tool that genuinely uses large language models, neural networks, or advanced machine learning in ways that create real value, four tools are using the word "AI" to describe what software engineers would call business logic, rules engines, or basic statistical models.
The AI Tax: What You're Actually Paying For
The premium for "AI-powered" enterprise software in 2026 ranges from 20% to 40% over functionally equivalent products without the label. On some categories, it's higher.
Here are products I've personally evaluated for clients where the "AI" label was doing marketing work, not technical work:
"AI-Powered Customer Insights Dashboard." A business intelligence tool with pre-built report templates and a chatbot that converts natural language queries into SQL. The chatbot is a thin wrapper around an API call to a general-purpose language model. The actual analytics are standard aggregation queries. The non-AI equivalent costs about 40% less.
"AI-Driven Workflow Automation." A rules engine. If condition X, then action Y. The "AI" part is a natural language interface for defining the rules. You type "when a new lead comes in, send an email and create a task" instead of clicking through a visual workflow builder. The underlying automation is identical to tools that have existed since 2018. The pricing is 2026.
"AI Compliance Monitoring System." Keyword matching on regulatory databases with a classification model for sorting alerts into categories. The classification model is genuinely machine learning. The overall system is roughly as sophisticated as a well-configured Google Alert with a spreadsheet. The vendor charges $2,999 per month. A competent analyst with RSS feeds could match the output.
"AI-Enhanced CRM." The AI enhancement is lead scoring based on demographic data and engagement metrics. Lead scoring has existed in CRM software for over a decade. The model underneath is a logistic regression. Not bad. Also not what most people picture when they hear "AI." The premium over the non-AI tier is $45 per user per month.
I'm not saying these tools are worthless. Some of them are genuinely useful. The keyword matching contract tool saves time, even if it's not what the marketing promises. The lead scoring CRM helps prioritize outreach. The workflow automation tool works.
But useful and "AI" are different claims. The tools are useful because they're decent software. The "AI" label justifies the pricing premium. The premium isn't buying AI capability. It's buying AI marketing.
The Governance Problem: You Can't Govern What You Can't Define
The pricing scam is expensive. The governance gap is dangerous.
Ask a business owner how many AI tools their company uses. Most will name one or two: ChatGPT, maybe Copilot. The actual count is usually ten to twenty.
Their CRM's lead scoring? AI. Their email platform's send-time optimization? AI. Their accounting software's anomaly detection? AI. Their e-commerce recommendation engine? AI. Their phone system's transcription? AI. Their scheduling tool's optimization algorithm? AI.
These features don't market themselves as AI because they crossed McCarthy's line years ago. They work, so nobody thinks of them as AI anymore. But under regulations like Colorado's AI Act, many of these embedded features qualify as "high-risk AI systems" because they influence consequential decisions about people.
The company that builds its AI governance framework around ChatGPT and Copilot is leaving a dozen other AI systems completely ungoverned. The lead scoring algorithm that decides which prospects get attention. The anomaly detection model that flags financial transactions. The recommendation engine that shapes purchasing decisions. All operating without oversight because nobody calls them "AI."
When Colorado starts enforcing its AI Act in June 2026, the definition of "AI system" in the statute is broad enough to capture most modern software that makes predictions or recommendations. Companies that haven't audited by function rather than label will discover they have regulatory obligations they didn't know existed, for tools they didn't know qualified.
The Regulatory Minefield
Every AI law needs to define AI before it can regulate it. Those definitions are either too narrow (missing things that should be covered) or too broad (capturing every piece of software with an if-then statement).
The EU AI Act defines an "AI system" as "a machine-based system designed to operate with varying levels of autonomy that may exhibit adaptiveness after deployment and that infers how to generate outputs such as predictions, content, recommendations, or decisions." That definition covers ChatGPT. It also arguably covers Excel's FORECAST function.
Colorado's definition is broad enough to capture most modern software that makes predictions or recommendations. California's emerging frameworks use different language with different boundaries. A tool that qualifies as "AI" in one jurisdiction might not qualify in another, which means your compliance obligations change depending on which state's definition you apply.
For businesses operating across state lines, this definitional chaos means you can't even determine which of your tools fall under which state's regulations without a legal analysis of each tool against each jurisdiction's definition.
How to Stop Paying the AI Tax
Evaluate by Capability, Not Label
When a vendor says "AI-powered," ask five questions:
1. What specific algorithm or model powers this feature?
2. What data does it train on?
3. What are the known limitations and failure modes?
4. Can you provide documentation sufficient for a regulatory impact assessment?
5. What does this product do that a non-AI alternative doesn't?
The answers separate genuine AI capability from marketing. If the vendor can't articulate what model powers the feature, you're buying a label.
Audit by Function, Not Nomenclature
Don't ask "what AI tools do we use?" Ask "what software in our organization makes predictions, recommendations, classifications, or automated decisions?" The second question catches the tools that crossed McCarthy's line and no longer feel like AI, plus the tools that claim to be AI but aren't.
Build your inventory based on what the software actually does, not what the vendor calls it.
Benchmark Against Non-AI Alternatives
Before committing to an "AI-powered" tool, identify the non-AI equivalent and compare. What does the AI version do that the standard version doesn't? Is the delta worth the premium? Sometimes it is. A genuine large language model integration that processes unstructured text is a real capability. A chatbot skin on a search function is not.
Govern by Risk, Not by Label
Your governance framework should trigger on impact, not on branding. A lead scoring algorithm that determines which prospects get human attention carries more risk than a chatbot that helps draft internal memos. But the chatbot is called "AI" and the lead scoring algorithm is called a "feature." Under a risk-based framework, both get appropriate governance based on what they actually do.
Why This Matters Now
Three years ago, the definition of AI was a philosophical curiosity. In 2026, with real laws carrying real penalties, the definition is a financial and legal question.
If your company uses software that predicts, recommends, classifies, generates, or decides, you're using AI. Regardless of what the vendor calls it. Regardless of whether it "feels" like AI. And in an increasing number of jurisdictions, that means you have compliance obligations, documentation requirements, and potential liability you haven't accounted for.
The vendors charging you the AI tax are betting you won't look under the hood. The regulators writing AI laws are betting you will.
Somebody's going to be right about that bet. Make sure it's not the vendors.
---
Kaizen AI Lab evaluates AI tools for clients, separating genuine capability from marketing premium. We build governance frameworks that cover the full technology footprint, not just the tools with "AI" on the label.
Take the AI Compliance Readiness Assessment: acra.kaizenailab.com
Learn more: kaizenailab.com
Book a call: cal.com/dhoesq/kaizen