Most people’s first instinct with AI is to use it the way they use a search engine. Type in a question. Get an answer. Move on. It’s a natural starting point because it’s the interaction model everyone already knows, and because the early, casual use of AI tools tends to reinforce it. Ask something simple, get something useful, feel like you understand how this works.

That instinct is also the single biggest thing limiting how much value most business owners get from AI. The search engine model produces search engine results. It doesn’t produce the kind of output that changes how a business operates.

The Vending Machine Problem

Henry Hays has a way of describing this that makes the problem concrete. When someone goes to a vending machine, pushes a button, and gets their snack, that’s the full extent of the interaction. The machine didn’t need to understand anything about the person. The person didn’t need to communicate anything beyond which button to push. The transaction is complete, and both parties move on.

That’s the relationship most people have with AI when they start. They push a button, metaphorically, by typing a question or a request. They get an output. They evaluate whether the output is what they wanted. If it isn’t, they either accept it anyway or they give up and do the task themselves. The idea that they could push back, redirect, add context, ask the AI to try a different approach, or build on what it produced doesn’t occur to them, because that’s not how vending machines work.

The problem is that AI isn’t a vending machine. It’s something closer to what Hays describes as a Tony Stark and Jarvis relationship: an ongoing conversation with an assistant that gets more useful the more it understands about what you’re actually trying to accomplish. The output from the first exchange is a starting point, not a finished product. The value comes from what happens in the back and forth after that.

What a Real AI Interaction Looks Like

The shift from vending machine to real assistant happens when you start treating AI outputs as drafts rather than answers.

A draft is something you react to. You read it, you identify what’s right and what’s off, you tell the AI what to keep and what to change, and you push it toward something better. That iterative process is where the real capability of AI shows up. The first output is rarely the best output. The fifth or sixth iteration, shaped by specific feedback along the way, usually is.

This requires a different kind of engagement than most people bring to AI in the early stages. It requires being specific about what isn’t working and why. It requires giving the AI enough context about the situation, the audience, the goal, and the constraints so that it can produce something genuinely tailored rather than generically useful. It requires patience with the process, because the iteration takes longer than accepting the first output, but produces something significantly better.

Hays puts it in direct terms: the relationship with AI needs to be a conversation, not a transaction. That means treating it like a capable assistant you’re working with, not a tool you’re operating.

Why Context Changes Everything

One of the most consistent differences between people who get a lot of value from AI and people who don’t is how much context they provide before asking for anything.

The business owner who types “write me a proposal for a new client” gets a generic proposal that sounds like every other generic proposal. The business owner who explains who the client is, what problem they’re trying to solve, what the business relationship has looked like so far, what tone is appropriate, what the most important thing to communicate is, and what a successful proposal would accomplish, gets something that requires far less revision and is far more likely to actually work.

The AI doesn’t know your business, your clients, your voice, or your goals unless you tell it. And telling it is the part most people skip, because telling it takes effort, and the whole point of AI is supposed to be that it reduces effort. But the effort invested in context is the thing that makes the output worth using. It’s not overhead. It’s the work.

This is also why the leaders who get the most from AI tend to be the ones who have used it long enough to develop a feel for what kind of context matters and how to provide it efficiently. That feel develops through practice. The first few interactions with any AI tool tend to produce middling results because the user hasn’t yet learned how to work with it. The interactions six months in tend to be significantly more productive, because the user has learned what the tool needs to perform well.

The Prompting Skill Nobody Talks About Enough

The way you ask AI for something, the prompt, determines most of what you get back. This is not a controversial point, but it’s one that doesn’t get enough practical attention in most conversations about AI adoption.

A good prompt does several things. It describes the task clearly and specifically. It provides relevant context. It specifies the format or structure of the output if that matters. It identifies constraints: what to avoid, what tone to take, what the output will be used for. And it leaves room for the AI to do something useful rather than just filling in a template.

Learning to prompt well is a skill, and like most skills it develops through practice and feedback. The feedback in this case comes from the outputs: when a prompt produces something close to what you wanted, you learn something about what worked. When it produces something off-base, you learn something about what was missing or unclear.

Hays makes the point that the best resource for learning how to use AI better is the AI itself. If you’re not sure how to ask for something, ask the AI how to ask for it. If the output isn’t right, ask the AI what additional information would help it do better. The tool is not only capable of doing the work. It’s capable of helping you understand how to get more from it.

Building the Habit Before You Need It

The business owners who are best positioned to lead their organizations through AI adoption are the ones who have already developed a personal AI practice before they need to talk about it with their teams.

That practice doesn’t have to be elaborate. It means using AI consistently on real work, pushing past the first output when it isn’t quite right, building the habit of providing context and iterating rather than accepting the first thing that comes back. Over time, that habit produces a level of comfort and competence with AI that is genuinely difficult to develop any other way.

The leaders who have that practice can do something their counterparts can’t: they can show their teams what good AI use looks like, not just tell them. They can answer the questions that come up when employees get stuck, because they’ve been stuck in the same places and figured out how to move through them. They can evaluate AI outputs critically because they understand what good outputs look like.

That practical knowledge, built through use rather than study, is what makes the difference between leading an AI initiative and sponsoring one. And it starts with the decision to stop treating AI like a vending machine.

Listen to the full podcast episode here.

Related Resources

• The Boring, Mundane, and Tedious: Where AI Actually Delivers for Growing Businesses: Knowing how to use AI is most valuable when it’s applied to the right tasks. This post identifies the workflows where growing businesses see the fastest, clearest returns from AI.
• Is It Too Late to Start Using AI in Your Business?: The best time to build this skill is before you feel like you need it. This post covers where most businesses actually are in the AI adoption curve and why starting now still makes sense.
• Why AI Adoption Stalls at the Leadership Level (And What to Do About It): Personal AI practice is the foundation of leadership credibility on AI. This post covers what it looks like when leaders model AI adoption and why it matters for the rest of the organization.
• The AI Execution Problem: Why Most Growing Businesses Can’t Get Past Experimenting: Individual AI skill is one piece of a larger organizational challenge. This post covers the full picture of what it takes to move from experimentation to real results across a growing business.