The market had a fantasy. Platform vendors sold it hard. Business owners bought it.
Push a button. Solve the problems. Let a few technical people program the thing and walk away. AI as the employee who never needs managing, never needs context, never needs direction. Just results.
What they actually built was closer to hiring a sharp, capable person, putting them in an empty office with no resources, no onboarding, no brief, and expecting everything to happen correctly.
It didn't. That's why AI was a wasted resource for so many businesses last year. Not because the technology failed. Because the direction never existed.
AI is a sentence.
Someone directing. Something to do. A skill required. Context to draw from. Resources that inform it.
In December 2025, Anthropic's Barry Zhang and Mahesh Murag stood up at the AI Engineering Code Summit and said something worth paying attention to.
Don't build agents. Build skills instead.
They're right, but there's a little more to it. The skill is the verb.
Not the platform.
Not the tool.
Not the prompt template.
The verb — the action you're directing, the decision you're encoding, the outcome you're accountable for.
Disappointing for so many.
Everyone wanted to ship a sentence.
Billions of dollars dressing up something that was always just a sentence.
But wasn't that the point of AI after all?
Some of us had the wisdom to see it early. Not because we were technical. Because we understood communication. We'd spent careers directing work — people, markets, organizations — and we recognized the structure underneath the hype.
Every direction is a decision encoded in language.
- Who?
- Why?
- Where?
- When?
- What good looks like.
- What happens if it doesn't go as planned.?
The nouns carry the who and what. The verbs carry the action. The adjectives carry the quality standard. The conjunctions carry the logic. The punctuation carries the weight.
A well-constructed direction isn't a command. It's a decision tree expressed as language.
The people who understand this, AI that is, instinctively aren't always the most technical people in the room. They also aren' t always the youngest.
They come from different backgrounds - security, communications, scientific research, librarianship — fields built on navigating complexity, translating across domains, and directing action under uncertainty.
Sentence structure is the workflow.
The scientific method is already the AI workflow. Hypothesis. Methodology. Tools. Experiment. Observation. Results. Conclusion. The SMART framework is already the AI framework. Specific. Measurable. Achievable. Relevant. Time-bound.
None of these are new.
They're human systems built long before AI existed. What they share — and what good AI direction shares — is the demand to commit to an outcome before you know the result. To hold complexity without collapsing it into a single lane.
That's the opposite of siloed thinking. Linear thinking. Compartmental thinking. Those are useful for executing inside known systems. Not for directing unknown ones.
AI has three real capabilities.
- Automation — AI executes repetitive defined tasks.
- Augmentation — AI enhances human judgment and capability.
- Acceleration through Orchestration — human and AI working together across skills, context, and scale to reach outcomes neither could reach alone.
We haven't reached true agentic AI. Many serious researchers argue LLMs are structurally incapable of it. What we have is powerful, human-directed, scalable.
That's enough. That's more than enough.
Clarity. Ownership. A single point of accountability. Not consensus. Progress requires iteration. Iteration requires imperfect results.
Skills over agents. The skill is the verb. Always was.
When people tried to box up those sentences, package them, and ship them as products — they lost their own vernacular. The sentence was never the product. Your story is.
If you want to learn more about how to utilize AI to your best advantage inside of your business, get in touch here