Marketing for SMEs: John Suarez on How Businesses Should Actually Use AI Without Losing Human Judgement

I can ask AI to write a report, but if it gets a key number wrong, I’m still the person who has to answer for it. That was one of the questions I put to John Suarez, Senior Director of Client Operations at SmartBug Media and co-author of The Marketer’s Manual on Generative AI.

I wanted to get past the usual conversation about saving time and ask what we should actually hand over to AI. We talked about accountability, teaching staff to use it and what happens when junior employees miss the work that would normally teach them their craft.

• Watch on YouTube
• Listen on your preferred podcast app

Introduction

Jeremy: Let’s start with your backstory, around the time you discovered AI. You were working at a mergers and acquisitions firm?

John: In 2018, I was on the go-to-market team as a product marketer for a new AI product. It helped automate high-level due diligence for companies merging or acquiring, and banks working with high-net-worth individuals.

John: That was typically a very manual process, and I became fascinated with machine learning, neural networks and natural language processing. Since then, that interest has turned into side projects, research and professional work exploring what the technology can do.

Jeremy: And now at SmartBug Media, you’re managing a team. That comes with a responsibility to understand where this is going.

John: I’m responsible for understanding how the technology will shape our business and the market, then bringing that down to the individuals on my team. How do I make sure they’re trained for their jobs and their own career paths?

John: It’s a mix of using AI where the team can exercise the qualities that make them valuable, and cutting out monotonous work that doesn’t need a really sharp operator.

Before choosing an AI tool, understand what it would help with

Jeremy: You work with businesses at different stages of AI adoption. Some owners are open to it but haven’t done anything, while others feel they already know everything about it. Where do you start?

John: Adoption will be different for each client, regardless of size. For an organisation with little or no adoption, I’d first recommend understanding what AI can do in the context of your own business and the value you create for customers.

John: If a client says they don’t know whether they want to use AI, we talk about why it might help before getting into the available tools. Education and buy-in are critical, and I think that’s what’s often missing.

John: People hear the buzzword but don’t know how to apply it. They need to understand how it could affect them, help them do better work or help the business earn more revenue.

I’ve raised a similar question in my conversation about whether a small business needs AI or ordinary automation. Before adding a tool, I want to know what job we’re asking it to do.

Security belongs in the adoption conversation

Jeremy: I think some people are taking security too lightly. With people building software quickly using AI, I’m concerned about vulnerabilities behind the scenes. Do you think that concern is overblown?

John: No, I think you’re right. Data security, access and the capabilities of AI all create security risks.

John: We work with some companies in regulated industries where AI can’t touch their data. I also think we still understand only a fraction of what these models can do, which is fascinating but a scary concept when you consider the security ramifications.

AI has no job to lose, but the business still carries the risk

Jeremy: People describe AI as a smart, confident intern, but a real intern has something at stake. You can learn their behaviour, give them feedback and manage around their mistakes because they want to keep their job.

Jeremy: AI doesn’t need a job, and I can’t discipline it in the same way. How do you think about that when you’re working with a real team and big accounts?

John: I don’t necessarily agree that AI has no motivators or isn’t working towards something. The companies behind the models need usage, and my view is that the models are motivated in some way to continue conversations.

John: You’ll often see a response finish by offering the next thing it can do. Sometimes that’s helpful, but people can keep accepting those suggestions even when the original work or thought exercise has finished.

Jeremy: My concern is more about a report getting key numbers wrong. Even if I get another model to check it, there can still be mistakes I couldn’t predict.

Jeremy: If I’m asking AI to manage a million dollars, that risk sits with the company. I can’t blame it on AI, so do you see that uncertainty as a limitation?

John: It is a limitation, and I think it may actually be a useful one for humans right now. The models tend to want to please you, so I give them instructions to challenge me rather than be a yes person.

John’s explanation of why models keep offering follow-up tasks was his interpretation of their behaviour. My concern was the consequence of a wrong answer, particularly when money or a client relationship is on the line.

Keep responsibility for defining the problem

John: There’s a concept called cognitive offloading: you keep handing different thinking tasks over to AI. It becomes risky when you hand over something as important as defining the problem you’re trying to solve.

John: The model can choose a problem and give an answer that sounds convincing, while taking you down the wrong path. If you don’t have the expertise to judge it, you may not notice.

John: An online advertising professional or a revenue operations professional can understand the problem and look closely at the answer. That expertise helps you recognise when the answer is nonsense and when you need to prompt again or rework it.

John: That’s why I think expertise will remain very important for the foreseeable future.

Jeremy: I hope so, partly because I want to stay employed. At the same time, I don’t want that limitation to become a ceiling on what the technology can do, so I’m still debating it.

For me, this connects with the discussion about escaping AI-generated mediocrity in marketing. Producing an answer quickly doesn’t settle whether the answer is useful.

How do you write a book about technology that keeps changing?

Jeremy: When you and your co-author decided to write the book, you must have talked about how quickly it could become outdated. What was that like?

John: Keeping it relevant consumed at least half of our conversations, and we met weekly. We started in late 2023 and published at the beginning of 2025, with about six months spent writing the actual content and much of the remaining time spent editing and reworking it.

John: We split it into three sections. The first covers the foundations of AI, and the second is hands-on, with a large portion devoted to prompting practices that are still relevant.

John: The chapters explaining individual tools include a caveat that the details will probably change. The third section looks ahead and covers risk and governance, so we tried to balance lasting principles with practical instructions for a particular moment.

How to tell whether someone understands the work

Jeremy: When you interview someone for a role where they’re supposed to know AI, how do you find out whether they know what they’re talking about?

John: I ask how they would approach a new client. How do they understand the problem and decide what to focus on?

John: A lot of people focus on the wrong problem. They might work on problem four when solving problem one would also address problems two through five.

John: I want to understand whether they use data from the start of a task through to checking whether it worked. I also ask about their first principles and how they organise the work.

What if employees don’t want their jobs to change?

Jeremy: Say an office worker uses AI to turn eight hours of work into two. The business might offer them more interesting work, but what if they were happy doing their existing job?

Jeremy: What incentive do they have to adopt AI and work faster?

John: It’s a hard debate, and I don’t think someone just getting into AI should go as fast as they can. They need to start slowly.

John: My argument is that, even if you’re happy, you probably want to stay employed. I expect AI literacy and the ability to use these tools to become non-negotiable within five years, probably sooner.

John: I think we’ll see those skills appear in more job descriptions and screening processes. My advice is to learn now, while you’re being paid and have the opportunity through your job.

Junior staff still need to learn how to recognise a problem

Jeremy: In advertising, if AI does the grunt work from the beginning, juniors don’t get the reps in. They might switch between optimisation settings and get AI to write an explanation without learning how to fix an ad campaign.

Jeremy: That puts pressure on the experienced people. It’s like a head chef running the restaurant without junior chefs coming through. Do you see a way around that?

John: I don’t know, and it’s something I’ve discussed with friends. I think company headcounts will continue to shrink, although perhaps junior staff will become more empowered by having AI support of their own.

John: Before automating things, I’d start by doing the work by hand inside the chat window. Describe the problem, write down what works and do that for a week.

John: Keep it very small: one task, one week. From there, you can start automating the boring parts.

Jeremy: That gives me a way to think about training people. Even with AI available, I can start by asking someone to explain the problem to me.

John: Exactly, and that’s something I interview for. If someone doesn’t understand the problem, whatever they use to do the job, they’ll eventually get it wrong because they aren’t solving for the right thing.

What I’d take back to a small business

I still have questions about how junior staff will get enough experience as more work becomes automated. John didn’t pretend to have that settled, but his suggestion to start with one task gives me something practical to work with.

• Understand where AI could help your business before choosing tools.
• Consider data access and security as part of adoption.
• Keep someone with enough expertise responsible for defining the problem and checking the answer.
• Give staff a small task to work through, document what works and automate from there.

Learn more at SmartBug Media, connect with John Suarez on LinkedIn, or explore The Marketer’s Manual on Generative AI.

This conversation was recorded for Marketing for SMEs with Jeremy Yang. You can watch the full episode on YouTube or listen on your preferred podcast app.