I sit in many vendor booths and conference keynotes. After enough of them, the slides fade out. AI powered router. AI-powered insights. AI-powered content. With the third statement, you cannot say that you are being sold the same product four times or four products once. The term is broad enough to cover anything, which is a problem. It doesn’t tell you anything about what you’re buying.
I don’t think we’ll be talking about AI as much within a few years, at least not as much as we do now. Jay Pattisall and Mike Proulx at Forrester made this point recently: It’s the same arc as electricity. Everything was electric [whatever] until it was gone. We put food in the fridge, not the electric fridge.
There is something useful in clarifying why you are withdrawing this word because it will not disappear gracefully on its own. Electricity left us talking because we stopped paying attention to it. It just 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.
More AI systems will always be possible. Instead of the light bulb failing to turn on, the AI will fail silently, and the wrong overconfidence answers look like the right overconfidence answers unless you do rigorous testing and management. AI will begin to be removed from low-level ambient applications and will stick to those where there is real money on the line for fault: diagnosis, credit scoring, and liability.
Four skills, not one class
Branding isn’t just about avoiding buzzword overuse. That’s more important than the conversations that don’t happen when we use an umbrella term that covers overlapping but separate activities and actions. Under “AI” sit four different things a marketing, CX, or service system can do. Say which one you’re buying, and your decisions, ownership lines, and processes become sharper. Ultimately, so is customer experience.
10X your SEO with Semrush for Enterprise.
The world’s most powerful SEO platform, built for the Enterprise.
Request a demo
Stop worrying about what AI for a specific platform means. Divide it into four bundles: generation, augmentation, detail, and orchestration.
1. Generation
When a machine produces an artifact, it is a product, and humans do not need to be the author. Millions of emails, each one for one person. Synthetic test data that could not be collected in the real world. Designs pile up into millions upon millions of possibilities that no one can draw by hand.
2. Raising
A second set of human hands still controls the workflow. Real-time assistant including service response. It’s a tool that helps you quickly prototype your analytical models. Augmentation consists of four main sections. It silently devours most of what people say is AI intelligence. Augmentation is less technical than vertical. The work is still done by a person. They can do more.
3. Understanding
This skill gives a decision. Sometimes at the front of the process (predictable: propensity to score, churn risk), sometimes at the back (analysis of work already done). What comes out is understanding, and someone still has to do something about it. When that person is another plan, give the fourth bucket.
4. Orchestration
This capability covers operations across systems, tools, and agents. It is achieved through a chain of supervision: all steps are authorized by one person in the extreme, which has full control over the other. Autonomy is just the far end of the dial, not something you can buy yourself.
Most real deployments are two or three of these skills at once, and that’s fine. The next best practice, for example, is a data model that feeds the orchestration engine with intent. The point of naming them is not purity. The salesperson who sells you the AI gets to get away with what he’s really doing. A salesperson selling you a “data model that feeds your best action engine” can’t.
You’ll notice that this doesn’t say anything about search AI, search engine optimization (GEO), or AEO, on purpose, despite those being hot topics for many marketers. That’s customer behavior: how people find you through an answer engine. The AI umbrella I’m talking about covers the tools your teams use, not the ones your customers do: different problem, different article.
Where the four skills overlap
Two skills that may be combined or mixed are nurture and generation because they both produce something. I use one query to edit them, and they work for anything:
Remove the AI. Can a skilled person still do this, be it less, less, or worse?
Yes, it means being raised. “Write this in my voice.” “Pass the contract.” Someone who has the ability to do both without assistance. AI raised author. There is no single generation: a thing exists only because it is a machine operating at a scale or dimension that no human loop of approval can reach. A million individually tailored emails. Training data no one has ever collected. No one would write those by hand.
The line was composition. A skilled person could write the first set. No one would write a second time. Addition is about who is in the chair. Generation is about what is produced. They always pile up. Every day “AI made me something” is usually a generation working for a human writer, and the test still tells you what conversation you’re in.
How to decompose AI into meaningful tasks
It doesn’t take a very professional person to do this well. Take AI-powered subject lines, for example. It is divided into three ways:
- Skills: The generation is writing, working to raise. The advertiser still owns the campaign and authorizes the submission.
- Method: A generative model, sitting where a rule-based template or empty A/B test used to sit. (The process is the mechanism behind the ability: rules, prediction, generation, or agent. Whether this one gets its seat is to fight for the next part.)
- Cost of wrongdoing: Does anyone read the difference before reaching a customer? A failure in this case is a clever subject line that is off the mark or completely wrong, out of scale because it’s readable at a glance.
That’s one feature and three different conversations, and none of them happen while it’s AI-powered.
What you can do next
Start by applying the framework to the AI programs and tools already in place in your organization.
Re-tap all current AI efforts with one of four skills
Double money appears immediately when the lines stop all learning AI. I’ve watched a stack test reveal a bunch of tools that solve the same problem, a few that were decommissioned months ago but are still charging. You can’t get that when every line says AI.
Use the names to assign the owners
“Who owns the pipeline of content being produced?” you have an answer. “Who owns AI?” has a committee. Accurate language is a measure of accountability and a budget line your CFO can protect. If you don’t own the entire stack, that’s okay. He owns the name, and it is what forces every organization to show its work.
Next time the slider says AI-powered, don’t ask what it’s powered by. Instead, ask which of the four does it and how much it costs you if it’s wrong. Marketers benefit from a broader AI label. You benefit from knowing exactly what the technology does and what happens when it fails.
