AI vs. Automation: Where Each One Belongs in Your Business

Key takeaways

1. An automation is a set of rules. If you can explain a task as steps in order, it can run without AI.
2. Put AI where a decision gets made. Researching a new lead takes judgment. Copying it into your CRM does not.
3. Give AI context and instructions. Train it the way you would train a new employee: who you are, how you write, and where things go.
4. Get organized, then connect. Bad data in means bad answers out. Once your files are in order, AI becomes the way you ask questions of them.
5. Aim for minutes, not a 30-hour week. Five minutes saved 20 times a day is more than an hour, and anything you build needs maintenance.

The question we hear most about AI is some version of this: how do I get AI to automate everything for me? It sounds like one question. It is really two, because automation and AI do different jobs, and most of the frustration owners have with AI comes from treating them as the same thing.

This is how we explain the difference, where AI actually belongs in a workflow, and the order we tell owners to start in.

An automation is a set of rules

Describe a morning routine to a robot and it would sound something like this: get out of bed, brush your teeth, feed the cat, go for a walk. Simple steps, in the same order, every day. That is all an automation is. It is a set of rules: you do this, then this, in this order, and you call it a day.

Most businesses already run on routines like that. Invoicing is one. Look up the customer, look up what you did for them that month, make the invoice, send it. The instructions never change. You repeat them for every customer, every month, and anything you repeat that faithfully can be automated.

Notice what is missing from that list: AI. A lot of owners assume automation automatically means AI. It doesn’t. Most automations are plain rules, and they are the foundation everything else sits on. Think of them as Lego bricks. Step one, then step two, and eventually you have a whole house.

AI belongs where a decision gets made

Go back to the morning routine. Feeding the cat means walking to the pantry, looking around, and picking the right can. It feels like second nature, but it is a decision, and that is the difference. Brushing your teeth is a set of instructions. Finding the cat food takes a little thought. It is not like turning a doorknob.

That is where AI goes in a business: at the steps where something has to be figured out. Here is the simplest example we know. Someone fills out the form on your website and lands in your CRM. That is a plain automation, a copy and paste in one or two steps. Now add AI. Before you are even told about the lead, it looks up the person’s email and company domain, writes a short brief on the business, matches it to your services, and puts all of it in front of you. That part can’t be written as fixed rules, because it means going online, making sure you have the right company, and making judgment calls along the way. That is the job you give AI.

Why automating everything backfires

AI has become its own buzzword, and with it came a promise: hand everything over and step away. We have watched owners try.

People think it’s just like, let me use AI to automate my whole life. But you’ve seen where people attempt to do that, and it turns into an absolute disaster.

It usually goes wrong the same way. The owner never sets parameters, because they assume artificial intelligence won’t make mistakes. Then the automation breaks something. Email is the classic case. Tools that answer your email for you exist, but the replies may not sound like you, and the AI may answer emails you never wanted answered.

Each time, two things were missing: context and instructions.

Context is what the AI knows about you. The example we see most: someone asks for an Instagram caption for a video and gets the most generic caption imaginable, em dashes and all. The follow-up prompt is “make it sound more like me,” and nothing else. How would the AI know who you are? Do you write a certain way? Do you style your captions a certain way? Is there a brand message you are pushing? Without context, you will not get anything useful out of AI. Garbage in, garbage out, and the result is what everyone now calls AI slop. You can spot it all over LinkedIn: posts with no thought behind them beyond “write me a business idea about this.”

Instructions are how the AI should do the job. “Monitor my email” is not an instruction. The AI has no way of knowing that invoices go in one folder and a certain kind of request goes in another unless someone tells it. Treat it like an employee, and train it the way you would train a person.

It’s really weird to me that in 2026, AI is exposing the human skills that we should have before AI existed, meaning how to explain yourself and how to teach people stuff.

Teach it who you are

AI doesn’t arrive knowing who you are, how you write, or what your business is about. You have to teach it. The good news is that you don’t have to write a long description of yourself, because most of the material already exists.

1. Export your dataLinkedIn, YouTube, Facebook and Instagram all let you download your posts and engagement. It can take a few days to arrive. Feed it in and let the AI audit how you have actually written and acted.
2. Add writing from before AILetters, reports, school papers: anything you wrote yourself before AI existed. If you weren’t posting on social media before AI, older writing matters even more, because otherwise most of what the AI learns from may itself be AI-written.
3. Let it interview youWhen the ideas are in your head and won’t come out, ask the AI for a list of questions and answer them. You stop working on the questions and just answer.
4. Talk instead of typingDictate. A stream of consciousness is fine: no commas, no periods. The AI condenses it and organizes it.
5. End every prompt the same wayBefore you hit send, add one line: “If anything’s unclear, ask me questions to make it more clear.” It is the easiest way to get a better answer out of the same prompt.

Francis did this himself, with old letters and college reports. The way his AI writes to him now is significantly different from how it wrote at the start. Will it ever sound exactly like him? Probably not, because people keep changing. But it gets much closer. On a Business Collective call, the advice we gave one owner went a step further: ask the AI what it needs from you to do the job better.

Dictation unlocks a lot of people. Francis compares it to an exercise he learned in grade school called rocket writing: for a set amount of time, write whatever you are thinking, with no commas, no periods and no sentences, and see what comes out. We call it a brain dump. Everyone already has the content in their head. What you do every day is the content. The work is getting it out and shaping it, and that is where AI helps.

Where to start

So what should an owner do first? It depends on where you are.

If you’re not organized yet

Use AI to get organized. Plenty of businesses still run on pen and paper. Y Not You did too in the early days, and Francis still has the notebooks. If your work lives in Google Drive, connect it and have AI sort the jumble into folders, maybe by department: accounting, legal, marketing, taxes. That alone gives you a clear picture of what is actually in your business.

When you give bad data to AI, you’re just not going to get a good answer.

Once your files are in order, move on to email and communication, where a lot of owners are just as disorganized. An inbox full of unread notifications is time you can win back.

If you’re already organized

Use AI as the connector. Connect it to your Google Drive and let it learn what is in there, then ask it the questions you used to answer by hand. Based on last week’s expense report, what changed this week? It knows where the expense reports live and compares them for you. This is the logic layer: AI answering the questions that sit on top of your information.

We run the firm this way. Our business lives in Notion, and with the right prompt, the AI inside it can answer almost anything about the business. It can only do that because of the context and instructions we built up first.

Better prompts with CARE

The prompt is everything you type before you press enter, and the more detail you give, the better the answer. Francis’s example: if your video is about a trip to the grocery store, tell the AI what you bought, which brands, which store, the time of day, even the weather. Vague in, vague out. The structure we give clients is CARE.

Say you need an Instagram caption. Give the AI the context first. Then the action: write an Instagram caption, rather than leaving it to guess. Point it at a reference, like a caption whose format you want it to follow. Then set the expectation: this is for a short-form video, for this audience. You get a detailed result with far less back and forth.

One more thing people forget: social media isn’t one place. Instagram, TikTok, YouTube, LinkedIn and Facebook each work differently, so teach your AI the differences instead of assuming one caption fits all five.

Skills: rules with a brain

Once you are organized and your prompts are in shape, you can go further with what are called skills. A skill is an automation, a set of rules you follow every time, with AI handling the logic. The idea isn’t tied to one tool: it works across AI systems like ChatGPT and Claude.

We have one for this podcast. Type one short command, drop in the episode’s transcript, and it returns three YouTube titles to A/B test, a full description and the tags, all written to our brand guidelines. Nobody does anything else. Underneath, it is just rules: read the transcript, understand the business, apply this format, return these pieces. It’s the morning routine again, with a brain at the steps that need one.

Aim for minutes, not a work week

Some owners expect AI to hand them back 30 hours a week. That expectation sets them up for disappointment.

I say try to get back minutes. Literal minutes.

People lose five minutes copying and pasting an email from one place to another, and that is easy to automate. Five minutes here and there sounds small. Do it 20 times a day and it’s more than an hour, every day. Time is your biggest asset, and minutes are how you win it back.

The other trap is the do-it-yourself build. We talk to owners who want a whole AI automation built, then come back and say they just prompted a chatbot to build it instead. Our first question is what parameters they set. They put it live, it breaks, and now they have a problem they didn’t have before. The time they were trying to save turned into a new job.

Anything you build with AI needs looking after. We check ours consistently, with reminders.

There’s always maintenance. It’s like a car, right? It needs maintenance. It needs an oil check. It needs an oil change. It’s a tool.

Start with the rules

AI is a fantastic tool, but it is a tool, and the owners who get the most out of it are the ones who can explain how their business runs. Start with the routines you repeat every week and automate the ones that are just rules. Then put AI at the steps that need judgment, give it the context and instructions you would give a new hire, and go after minutes before hours.