Want to Automate With AI? Build These 5 Habits First

I’ve noticed a widening gap between marketers who use AI to finish isolated tasks and those who use tools like Codex or Claude Code to coordinate entire projects with surprisingly little input.

The difference usually isn’t access. Most people can open the same tools. The difference is whether they’ve developed the habits required to delegate well, evaluate results, and think in processes rather than individual tasks.

If you want to move from “ask ChatGPT for a draft” to giving AI meaningful autonomy, start here: build five habits that improve your communication, judgment, and ability to see how work gets done.

1. Invite AI to ask questions

The first habit is simple: before AI starts working, ask it to clarify the assignment.

Try adding a line like this to your prompt:

Before you begin, ask me any questions you need answered so you can do this well.

This improves the output, but that’s only half the benefit. It also trains you to become a better delegator.

Most of us aren’t naturally great at explaining a complete project. We leave out the audience, constraints, examples, definitions, dependencies, and desired outcome because those details feel obvious in our heads. They aren’t obvious to anyone else—not to a teammate and certainly not to an AI system.

When AI asks questions, it exposes the missing context. Over time, you start anticipating those questions before they arrive. You begin thinking, “What information would someone need in order to do this correctly?”

That’s a valuable management skill, not merely a prompting trick.

Use questions to improve your briefs

When I delegate a project, I want the person—or the agent—doing the work to understand more than the immediate task. What is this for? Who will use the result? What does success look like? What should be avoided? What decisions are already made?

You don’t need to answer every possible question in advance. In fact, it’s often better to let the questions reveal what matters. The practice is to make clarification part of the workflow instead of treating it as a failure.

If you’re building a more advanced AI marketing workflow, this habit gives you a much stronger starting point.

2. Edit early and often

Most people follow the same pattern: write a prompt, read the response, notice it missed the point, then add another prompt explaining what they meant.

Sometimes that’s the right move. But often, it’s better to go back and edit the original prompt.

There’s a small edit arrow behind nearly every prompt. Use it.

Editing the original request helps AI avoid getting lost in a long chain of corrections. More importantly, it helps you improve the request itself. You’re not simply adding more instructions to a growing pile of context. You’re refining the assignment from the beginning.

Turn every disappointing output into feedback

When the result isn’t what you wanted, ask yourself what the response revealed:

• Did I leave out important context?
• Was the audience unclear?
• Did I describe the desired format?
• Did I provide an example of the quality or tone I wanted?
• Did I ask for something too broad?

Sometimes you don’t know what you want until you see the wrong version. That’s normal. AI can help you discover your preferences and clarify your thinking.

But once you understand what was missing, go back and improve the original instruction when possible. This creates a useful feedback loop: better prompts, better results, and a sharper understanding of how to communicate with the system.

Pair this habit with asking AI to ask questions, and you’ll become a much stronger communicator. You’ll also build the kind of precision required to delegate larger projects instead of just handing off tiny tasks.

3. Know what good looks like

This is the hardest habit because it depends on judgment. You need to know what a good result looks like before you can reliably ask AI—or another person—to produce one.

That sounds obvious, but it becomes difficult when you’re working in unfamiliar territory. Maybe you want to build a software product but have never used one like it. Maybe you want to automate a sales process but don’t understand how the process works manually. Maybe you’re asking AI to create a strategy in a category you’ve barely studied.

In each case, you’re asking for an outcome without having a strong baseline.

How will you know whether the result is excellent, acceptable, or completely misguided?

Build a baseline before you automate

You don’t necessarily need decades of experience. But you do need exposure to the work. That might mean:

• Doing part of the process manually
• Reading a respected book about the subject
• Studying several strong examples
• Interviewing someone who has done the work successfully
• Learning the common problems, tradeoffs, and failure modes

Experience gives you pattern recognition. It helps you notice when something is intuitive, awkward, incomplete, or likely to break.

Without that baseline, autonomous agents can stay busy without creating much value. They may produce work that looks plausible, and you may approve it simply because you don’t know what else to look for.

That’s one of the biggest risks of AI automation: automating toward an undefined destination.

Before you give an agent a larger assignment, define your quality bar. What would a strong result contain? What would make it useful? What would make it dangerous or wasteful? The clearer your standard, the more autonomy you can safely give the system.

This is also where the human advantage remains visible. As I’ve written about in the AI paradox, AI can generate and transform information quickly, but human judgment still determines whether the result deserves to move forward.

4. Think beyond the task

Knowing what good looks like at the big-picture level is essential. But you also need to think about the smaller output you’re requesting right now.

When AI is working, don’t immediately reach for your phone or switch tabs. Use the brief waiting period to predict what it’s going to return.

What should this specific output contain? What format would make it useful? What might go wrong? What would pleasantly surprise you?

Then compare your expectation with the actual response.

Practice prediction at the smallest level

This sounds like a minor exercise, but it builds an important skill: understanding what AI can realistically do in a single step.

Large automated workflows are made of small steps. If you can’t anticipate and evaluate the small steps, you’ll struggle to design a reliable sequence of them.

For example, imagine you want AI to turn a customer interview into a marketing brief. That larger project might require AI to:

1. Extract the customer’s stated problems
2. Separate direct quotes from interpretation
3. Identify recurring themes
4. Compare those themes with your existing positioning
5. Draft a brief for a specific audience
6. Flag claims that need verification

Each step has its own expected output. If the extraction step is sloppy, the final brief can’t rescue the process. Thinking beyond the task means considering how this small result will be used next.

Over time, your expectations become better calibrated. You learn what AI handles well, where it needs supervision, and which steps require your judgment. That calibration is what allows you to expand automation without blindly handing over an entire project.

5. Visualize the process

The fifth habit is to become a process nerd.

I started noticing business processes everywhere years ago. A coffee shop has a process for taking an order, preparing the drink, and getting it to the right person. An airline has a process for selling a ticket, checking a passenger in, boarding the plane, and handling exceptions. A software company has processes for marketing, sales, support, development, and billing.

Once you start looking, you see systems everywhere.

When I place an order, I find myself wondering what happens behind the scenes. How is the request routed? Which steps happen automatically? Where does a human intervene? What happens when something goes wrong?

You don’t need to be correct every time. The exercise is learning to visualize the sequence.

Map the work before you automate it

For any recurring marketing process, try sketching the path from beginning to end:

• What triggers the process?
• What information enters first?
• Which decisions need to be made?
• What tasks happen in sequence?
• Which tasks can happen at the same time?
• Where does a human need to review, approve, or add judgment?
• What is the final output?
• What happens when an input is missing or unusual?

This process map makes automation much easier to reason about. You can decide what AI should handle, what a person should handle, and where the two should collaborate.

It also reveals opportunities that task-based thinking misses. Instead of asking, “Can AI write this email?” you start asking, “Can AI monitor these inputs, identify the right opportunity, draft the message, route it for approval, and record what happened?”

That’s a much more useful question.

If you’re interested in how I think about the boundary between automation and human trust, AI can find your buyers, but only you can earn their trust.

How the five habits work together

These habits reinforce one another.

• Inviting questions improves the quality of your delegation.
• Editing early improves the quality of your instructions.
• Knowing what good looks like improves your judgment.
• Thinking beyond the task improves your ability to design reliable handoffs.
• Visualizing the process helps you scale from one task to an entire workflow.

None of this requires you to start with the most sophisticated agent platform. You can practice all five habits in a basic chatbot.

Ask it questions. Refine your original prompts. Study strong examples. Predict the output. Map the steps around the task. The tool can change later. The mental model is the part that compounds.

Build the autonomous mindset before the autonomous workflow

The long-term shift isn’t from one AI tool to another. It’s from thinking in tasks to thinking in systems.

People who make that shift can give AI more responsibility because they understand the work well enough to guide it. They know what to ask, what to inspect, what to automate, and what not to delegate.

That mindset also makes you better at leading people. Clear expectations, thoughtful questions, quality standards, process awareness, and useful feedback are just as valuable with a team as they are with an AI agent.

So don’t begin by looking for the most autonomous tool you can find. Begin by strengthening the habits that make autonomy safe and productive.

Today, take one recurring task and practice the sequence: ask for questions, edit your request before adding more context, define what good looks like, predict the next output, and sketch the full process. Do that consistently, and you’ll be much closer to automating the work that actually matters.

In this episode, Dan Sanchez discusses five essential habits that marketers need to adopt to effectively leverage AI tools for project automation and team collaboration. He emphasizes the importance of mindset, communication, and understanding processes to transition from basic AI usage to advanced automation. Takeaways Embrace AI for automation Develop strong communication skills Chapters 00:00 Bridging the AI Usage Gap 01:10 Habit 1: Inviting Questions 02:36 Habit 2: Edit Early and Often 04:03 Habit 3: Knowing What Good Looks Like 06:26 Habit 4: Thinking Beyond the Task 08:20 Habit 5: Visualizing the Process 10:44 The Path to Autonomous Mindset

Dan Sanchez: As I meet more marketers, I'm starting to find that there is a gap between marketers who are using AI as a one-off task completion tool, just using ChatGPT to get simple things done, and those who are using Codex or Claude Code to orchestrate all kinds of projects, often with little to no input. It's just automatically happening by itself. And that gap is widening every day. In this episode, I wanted to cover five different habits. that's keeping you from reaching that level. Cause sure you might have opened up codecs, but do you have the mindset to be able to leverage it the way the advanced marketers are? Hello and welcome back to the AI Driven Marketer. I'm Dan Sanchez, my friends call me Danchez. And today I wanted to cover those five different skills with you. Because with those five skills, not only can you level up, but you can hand this out to your team in order to help them level up from where they're at, just using chat copy and paste ChatGPT over into a sophisticated agent driven tool so that we can get more done as a team. So let's go through these skills one by one, starting with the Easiest ones, but it's often the easy ones that end up as you do them repetitively, end up having the biggest point of leverage. So let's start with number one: inviting questions. Yes, you need to ask AI to ask you questions regularly, often, frequently throughout a process. It's one of the things that can teach you how to use AI better and better. Because if you say, hey, go get this done, but before you get started, ask me any questions that you have. in order to clarify what we're trying to do, not only does it make that output better every time, but it also helps you to anticipate what it's going to ask you. Because let's be honest, we're not all the best communicators in the beginning. We have to be trained on how to communicate and what kind of context we need to deliver in order to better delegate to people, but certainly to delegate to AI by asking it questions frequently or asking it to ask you questions. You will not only become a better delegator to AI, but I think you'll find yourself becoming a stronger communicator in your in in in entirety, because you'll start thinking ahead of e everything that needs to be communicated in order to delegate a full task or maybe even a whole project. But you have to start getting in the habit of asking AI to ask you questions. Habit number two is to edit early and edit often. Believe it or not, there's a little edit arrow behind every prompt that you send because we all have the habit of sending a prompt, reading the response, and then realizing, huh, that's not exactly what I meant. Maybe you forgot to ask the question, to ask you questions, right? But we can learn a lot by reading the output. Because sometimes we don't, we don't even know what we want yet, right? We're just asking for ideas. We're asking for inspiration, and that's a good thing. That's not bad. But sometimes we find out only after the fact. Instead of asking it f and giving it more context and generating again, it is often much better to go back to the edit arrow and change your original request. This is a strong habit because it not only helps AI not get lost in all the context of redoos, but it helps you tailor your prompts and makes your prompting stronger. And if you pair this second habit of editing early and often with this habit of asking it to ask you questions, you'll all You'll find that you'll become a stronger and stronger prompter and communicator. Now, number three is difficult, and you kind of either have it or you don't, but it's something you can know that you're working towards over the long haul. And that's to know what good looks like. Oftentimes we're trying to build things or bring things into existence with AI that we've never seen before, that we've never done, we've never even touched. You're trying to build a SAS app and you've never think. thought about what a SAS app could even look like. You've never given it a lot of thought. You've never built an app before. Well, it could be really hard to build something good. Why? Because you've you don't you don't know what good looks like. You don't know the pains that people go through to make something easy to use, intuitive, let alone all the infrastructure and things that could go wrong underneath with bad code. And that could be a problem. And I'm not saying do not do those things, but if you're heading into new territory, it helps. It helps To have experience in the thing. So what do you do if you want to build something and you don't have any experience, right? I can have you asking all the people who are like over the age of 40 like, yeah, that's right, experience. But what if you're 22 and listening to the show right now and you're like, well then like, well, what do we what do you want me to do? Just wait till I'm 30 or 40 and I have the experience to know what good looks like for a certain thing? No. Not necessarily, though it certainly is a massive pie point of leverage for those who have been doing the thing manually for a long time. but it does give you an idea of the fact that you probably need to go out into the world and try to do some of it manually. At least go read a book about the topic, go and talk or have an interview with somebody who's done the thing. It helps a lot. Because unless you know what good looks like, how would you even know if you got there with AI? This is a huge missing piece. And it's one of the things that's holding back your automation. Is because you don't even know what you're automating towards. You don't even know if you're 50% of the way there or all the way there or now going above and beyond what is normal because you don't know what normal looks like. So you need to establish a baseline of what good looks like. Otherwise, you can't send autonomous agents out to do things. They could be doing things, and you're like, I guess that's good. But if you don't know what good is, then those autonomous agents are just gonna be doing busy work for nothing. I mean, I think a lot of us probably feel like that in companies where our boss doesn't know what good looks like, and that could be painful. But it's even worse when it's you and all your agents, and it's gonna be nobody but your fault. So, number three, you have to know what good looks like. Go and get that experience. Go read some books, go interview some people who know what good look good looks like. If you do it aggressively, you can find out rather quickly. But you have to go and figure it out first. Now, number four is almost a subpart of number three, but it's even it's so much more small and micro that it's worth giving its own habit. Is you need to think beyond. The task. Oftentimes we give something for AI to do, but we don't think like what do we what do we what's what's the output supposed to look like? Before I was talking about it in the the macro, like what do we want an overall thing to look like? What do we know from experience the process should look like so that you arrive at the ultimate end, the big goal? But here, just the little piece of trying to think through what is this little task supposed to render? And getting the habit of thinking about it before you actually get it back from AI. And then comparing what you hoped for with what it gave you. Getting in a habit of thinking about it as it's rendering, because it'll be thinking, it'll be thinking, it always takes a second. Take that moment to think about what it's going to give back. And then comparing what you got back with what you were expecting to get back. It's a little habit. It takes a little bit of effort and most people will skip it. But it's that mental exercise you can do in that moment you have to wait. Put that moment to work. And think about how you can get closer and closer to what you were hoping for, or to be pleasantly surprised by what you get back. Because I find that those are the little things that you need in order to automate a sequence. In order to automate a bigger project, you need to be able to kind of reasonably understand what AI can do in the tiny little baby steps. Because if you can't do that, then you can't think through what the whole sequence should be unless you know what it can do on the baby step. So get good at anticipating what it's going to do so that over time. As you use it more and more, your expectations with what AI can do will align more and more closely, which will help you build better automated sequences or to give AI more autonomy over quite a bigger, a much larger range of a project. And then step five is visualizing the process. Visualizing the process. And I don't think this is again can be overlapped with the other two, but I promise it's something different. You need to get become like a process nerd in order to do this well. I have been like almost like a Pokemon trainer trying to collect them all, I swear, when it comes to business processes. I fell in love with it maybe ten years ago and then I just started hunting for them. Now I see them all over the place. It's hard not to, from a barista taking your order to coffee and how it goes through the Starbucks assembly line before your name's finally called at the other end, to when you buy tickets in an airline to where every business you interact with has a process and a system. Not just for taking orders and delivery, but for marketing and for sales. Everything's process driven out there in the the business world. I love to study them. I like to think and visualize the process. as I'm putting in my order, I wonder how it's taking place, how they're shipping it to me, what steps it can go. And it doesn't matter whether it's fully aligned with what it's actually doing, but it's starting to get an idea and visualizing different business processes that are all around you all the time. Even as they work on the internet and you're getting your podcast, like how is the app delivering this? What's the steps it's taking here? What's going on? Because if you can learn how to think and visualize the process, you'll find that you become it becomes easier to think about how it should work for you in your business or for you in your marketing department. But it takes a little bit of exercise here. That's why all these little habits kind of stack together so you can grow your like process development muscle. It doesn't happen overnight. It happens Through a bunch of little individual habits, a little individual behaviors that once you strengthen them over time, it becomes really easy to sit down with a complex project and think about, huh, I wonder how we can automate this whole process or how I can give just the right instructions to AI for it to autonomously take care of a huge portion of it. And I'll know which ones I need to do, it needs to do, and which ones we need to collaborate on before the whole thing's done. It's these little habits that will lead you and guide you from going from chat and use using a chat bot to get individual tap tasks done to actually pushing more and more of the work to a whole AI workforce. That's where we're going. And it's gonna get easier to get there. But right now, you can start building that habit and building that process muscle to lead AI so that it doesn't have to lead you. Believe it or not, this is actually really good help for leading people in an organization. This is often like the missing link to get from manager to director if you're not already a director and didn't know. So I'm going to be covering this process and more of these types of things in my new upcoming book called The Autonomous Mindset. Go to danchez.com/slash process, where you can get on an early launch list to get early access to this book. I'm so excited to be releasing it to the world very soon, but go to dance.com slash process. process where you can learn more about the four steps on how to develop awesome processes. It's coming soon. And I swear this book is something I've made for myself, for my team, because this is the missing link between learning how to just do tasks and getting way more work done effectively with AI.