{ “@context”: “https://schema.org”, “@type”: “AnalysisNewsArticle”, “mainEntityOfPage”: { “@type”: “WebPage”, “@id”: “https://martech.org/stop-buying-ai-and-start-buying-capabilities/” }, “headline”: “Stop buying AI and start buying capabilities”, “description”: “Organizations frequently invest in standalone AI tools without a clear operational goal, resulting in a tool. The real value comes from focusing more on building scalable marketing capabilities and fixing core processes rather than purchasing features.”, “datePublished”: “2026-07-20T08:00:00-05:00”, “dateModified”: “2026-07-20T08:00:00-05:00”, “author”: { “@type”: “Person”, “name”: “Greg”. Kihlström”, “jobTitle”: “MarTech & AI Transformation Advisor”, “sameAs”: “https://www.linkedin.com/in/gregkihlstrom/” }, “publisher”: { “@type”: “Organization”, “name”: “MarTech”, “url”: “https://martech.org”, “logo”: { “@type”: “ImageObject”, “url”: “https://martech.org/wp-content/uploads/2021/05/martech-logo.png” } }, “speakable”: { “@type”: “SpeakableSpecification”, “cssSelector”: ( “h1”, “.article-content p:first-of-type” ) }, “hasPart”: { “@type”: “WebPageElement”, “isAccessibleForFree”: “true”, “cssSelector”: “.article-content” }, “backstory”: “This commentary draws on field observations with enterprise marketing teams and industry consumer sentiment metrics, such as Gartner data showing consumer distrust of unverified AI results, to examine why tools fail without procedural frameworks.”
I attend many vendor presentations and conferences. After enough, the slides become blurry. AI-powered routing. AI-powered insights. AI-powered content. By the third speech, you can’t tell if you’re being sold the same product four times or four products once. The term is broad enough to cover anything, and that’s the problem. It doesn’t tell you anything about what you’re buying.
I don’t think we’ll be talking much about AI in a few years, at least not in the way we do now. Jay Pattisall and Mike Proulx of Forrester made this point recently: It is the same arc as the electric one. Everything was electric (whatever) until it wasn’t. We put the food in the refrigerator, not in the electric refrigerator.
There’s something useful about being specific about why you’re removing the term, because it’s not going to disappear gracefully on its own. Electricity left our vocabulary because we stopped paying attention to it. It works. You plug something into an electrical outlet and you get the same current whether you want it or not. If it doesn’t work, you know something is broken.
Many AI systems will remain probabilistic. Instead of a light bulb not turning on, the AI will fail silently, and overconfident incorrect answers will look like overconfident correct answers unless you have rigorous testing and governance. AI will begin to be phased out of low-stakes ambient uses and cling to those where there is real money at stake if a mistake is made: diagnostics, credit scoring, and legal liability.
Four abilities, not just one category
Removing the term isn’t just about avoiding the overuse of a buzzword. This matters much less than the conversations that don’t happen when we instead use an umbrella term that groups together overlapping but nuanced activities and actions. Under “AI” are four different things that a marketing, customer experience, or service system can actually do. Name the one you buy and your decisions, ownership boundaries and processes will be more precise. Finally, so does the customer experience.
10X your SEO with Semrush for business.
The world’s most powerful SEO platform, built specifically for businesses.
@media (maximum width: 768 px) { .headline-responsive { font-size: 30px !important; line-height: 1.3 !important; } }
Stop worrying about what certain platforms’ AI means. Break it down into four stacks: generation, augmentation, insights, and orchestration.
1. Generation
When the machine produces the artifact, which is the product, it is not necessary for humans to be the author. Emails by the millions, each personalized for one person. Synthetic test data that would be impossible to collect in the real world. Designs from billions of possibilities that no human could ever draw by hand.
2. Increase
The second pair of hands for a human who always controls the workflow. It is the real-time assistant that composes a service response. This is the tool that helps your analyst prototype working models faster. Augmentation contains the largest volume of the four categories. It’s quietly devouring most of what people claim is AI creativity. Augmentation is less of a technology and more of a posture. Ultimately, the work is still done by one person. They can simply do more.
3. Previews
This ability fuels a decision. Sometimes upstream of a process (predictive: propensity score, risk of churn), sometimes backwards (analysis of work already carried out). The result is understanding, and someone still has to act on it. When that person is another system, you have moved to the fourth compartment.
4. Orchestration
This functionality coordinates activities between systems, tools and agents. It exists on a spectrum of supervision: all steps individually approved by a human at one extreme, fully autonomous at the other. Battery life is just the tip of the dial, not something to buy in and of itself.
Most real-world deployments have two or three of these capabilities at a time, and that’s fine. The next best action, for example, is an information model that intentionally feeds an orchestration engine. The point of naming them is not cleanliness. A salesman who sells you AI can avoid the one he actually does. A vendor who sells you “an information model that powers your most powerful action engine” can’t.
You’ll notice that this doesn’t say anything about AI search, Generative Engine Optimization (GEO), or AEO, on purpose, although these are pretty hot topics for many marketers. These are customer behaviors: how people find you through a response engine. The AI umbrella I’m talking about covers the tools your own teams use, not the ones your customers use: different problem, different article.
Where the four abilities overlap
The two abilities most likely to be confused or mixed up are Augment and Generate, as both produce something. I use a question to sort them, and it works on anything:
Remove the AI. Could a trained person still achieve this, just slower, smaller, or rougher?
Yes means increase. “Write this in my voice.” “Write the contract.” A competent human does both without help. AI amplified an author. This does not mean generation by itself: the thing only exists because a machine has operated on a scale or in a dimension that no loop of human creation can reach. A million personalized emails. Training data that no one has ever collected. No one was going to write them manually.
The line is paternity. A competent human could have written the first set. No one would ever write the second. The increase concerns the person occupying the seat. Generation is about what has been produced. They are constantly piling up. The daily “AI made me a thing” is usually a generation working for a human author, and the test always tells you what conversation you’re in.
How to break down AI into more meaningful functions
You don’t have to be a highly technical person to do this well. Take AI-powered subject lines for example. Disassemble it in three ways:
- Abilities: The generation writes, in the service of growth. A marketer always owns the campaign and approves the sending.
- The mechanism: A generative model, located where a rules-based model or simple A/B testing used to be. (Mechanism is the mechanism behind the ability: rules, prediction, generation, or agent. Whether that one gets its seat is a fight for the next thing.)
- The cost of being wrong: Does anyone read the variations before they reach a customer? The failure in this case is a clever subject line that is subtly off-brand or downright wrong, distributed widely because it reads well at a glance.
This is one feature and three separate conversations, and none of them occur when filed under AI.
What you can do next
Start by applying the framework to AI initiatives and tools already in your organization.
Re-label each current AI initiative by one of four capabilities
Duplicate spending surfaces almost immediately once the lines stop reading the AI. I’ve seen stack audits reveal groups of tools solving the same problem, a few of which had been disabled months earlier but were still incurring charges. You can’t understand this when every line says AI.
Use names to assign owners
“Who owns the generative content pipeline?” » has an answer. “Who owns the AI?” » has a committee. Precise language is the prerequisite for accountability and a budget line that your CFO can actually defend. If you don’t have the whole stack, that’s no problem. You own the name, and that’s what compels the rest of the organization to show their work.
The next time a slide says “Powered by AI,” don’t ask what it’s powered by. Instead, ask yourself which of the four works and what it costs you if something goes wrong. Suppliers benefit from the broad AI label. You benefit from knowing exactly what the technology does and what happens if it fails.
The position Stop buying AI and start buying capabilities appeared first on MarTech.
