Your company may be slow with AI, but you can’t afford to be

What can you do if you know AI could improve your marketing work, but your company is still stuck in committee about it? Start experimenting with personal projects that could also solve business problems.

AI has escaped containment in the business world and is now everywhere. At the grocery store, my app uses AI to suggest what I should buy. My dentist even told me about a new AI-powered app designed to help people clean their teeth. AI is suddenly everywhere.

I know some people are wary of the technology, and for good reason. AI is advancing faster than anything we’ve seen before. But this is an inflection point not just for email or digital marketing, but for society as a whole.

This feels like the early days of email, when we had to teach ourselves email skills and develop our own programs because companies didn’t believe in email yet. Once again, individuals have the power to create new things using new technology.

In my last MarTech column, I said that early innovation in the email space began in super-enterprise companies and trickled down because people at the largest companies were the ones most driven to use the new technology.

The opposite is happening with AI. Innovation and applications are bubbling up, thanks to people working on their own or at mid- to lower-market enterprises to explore and expand AI’s uses. 

I have several friends who have created apps or websites that incorporate AI to solve practical problems. Instead, the super-enterprise and enterprise companies are the laggards because restrictions on information security and privacy, or company inattention, constrain their creativity. 

How to learn AI right now

If you want to go full speed ahead on AI beyond surface uses like copywriting, but your company is still debating it, teach yourself. Don’t wait around for your company to catch up. Instead, be ready when it finally happens. 

This insight should be obvious, but it isn’t. I still talk to marketers who use AI like a search engine or have a limited or no understanding of its power and potential. That means leaning deep into AI and immersion. 

Now is the time to explore and learn on your own — without violating the company’s privacy or security policies. Learn about AI first, and then use what you have learned to solve practical problems, either at work or in your personal life. That’s what I did — and it all began in my wine cellar.

I test-drove my knowledge on a pressing personal issue: inventorying and managing my 300-bottle wine collection. It might sound a bit frivolous, but I’m a serious wine enthusiast with a collection built over more than 20 years. Keeping tabs on it has been a niggling problem for me.

Learn which tool to use

Knowing which AI tool can do what is a key lesson. ChatGPT, Claude, Google Gemini, Microsoft Copilot, and other large language models each have their own strengths and weaknesses. For example, Claude is terrible at image generation.

By using what I learned on this personal chore, and not on a business need, I was able to explore and learn more about each tool and later avoid a major pitfall in my professional life: failing because I used the wrong tool.

I needed to inventory my collection, but I didn’t want to spend all the time it would take to record each bottle. So I asked myself, “How can AI help me do this?”

I started with ChatGPT because I had already created several dashboards with it and was already familiar with it. It suggested that I take a picture of every wine label, upload the images, and then it would use pattern recognition to build my inventory.

Failure can teach you a lot

Sounded easy enough. So, I spent about half an hour shooting and uploading the pictures. But when I turned the app loose, it didn’t recognize many labels and returned incorrect information instead. For example, it told me I had a bottle of 1999 Screaming Eagle Cabernet Sauvignon worth $2,985. I wish. There were enough similar errors to make me doubt that any of ChatGPT’s inventory was accurate.

I was surprised. ChatGPT has been so good at business forecasting and modeling for my work projects. Why did it fail here? I still don’t have an answer. My assumption, however, is that I was asking ChatGPT to do something that it wasn’t designed to do. 

I moved on to Google’s Gemini, which I had also experimented with and generally found to be pretty useful and responsive.

I gave it the same context that I gave ChatGPT and the label photos I’d taken. The results were hit-or-miss. Gemini couldn’t process the pictures, but that wasn’t the worst part. Instead of returning an unsure response, it guessed at the labels, and it usually guessed wrong. 

I tried adding more context and correcting the errors, but Gemini didn’t produce a trustworthy inventory either.

My wife is a marketing executive and loves working with Claude. So, I gave it a try. This time, the results were a good surprise.  

It recognized most of the labels in the photos and flagged the images I needed to retake because they were too blurry or I had cut off important information. Instead of returning hallucinations, it returned helpful suggestions.

Once I retook the photos Claude needed, it quickly built me a 300-bottle inventory. It was exactly what I was looking for.

Why did Claude succeed where ChatGPT and Gemini failed?

I didn’t tell Claude anything I hadn’t told ChatGPT or Gemini. I used the same context, prompts, and processes. What emerged from this experiment was a clear distinction among the three models, driven by their respective infrastructures.

In short, each AI model differed in what it could do well and what it needed to be pushed to do. Does this mean that Claude is the right platform for you? Not necessarily.

My wine inventory showed me that I needed to understand the unique capabilities and quirks of each system. I’m using that knowledge now in other projects to better understand which AI model will best suit the task at hand. 

I know it would be easy to find infographics that explain the differences between the models, but until you experience using the models yourself, you can’t know or trust someone else’s graphics.

Knowing how these models differ gave me an advantage that I will continue to utilize when I start reviewing tech platforms that incorporate AI. You can achieve that same advantage when your company finally decides which direction it will take with AI if you’ve already test-driven the tech and have a firm understanding of what will fit your company’s specific needs best.

Building your career with AI knowledge

The education you attain while experimenting and working with the various AI platforms won’t just pay off now, but also later in your career.

AI know-how is the number one skillset marketing chiefs will be looking for when hiring for their teams, according to Litmus’ State of Email 2026 report. The time (and money) you spend now learning AI and experimenting with tools will make you a much more valuable prospect

For my AI model experimentation, I used the paid versions for my work and home projects. And, like signing up for a plethora of streaming TV services, the costs add up. Remember that this is an investment in your future.

If you’re just getting started, a free or lower-cost tier of each AI model can help you figure out the limits of each platform. Once you know that, it may make sense to move up to a more expensive tier — especially for the systems you find the most useful.

What AI tools mean for marketing

Some AI users are diving deep into AI models: coding, hosting Python servers, using GitHub, and getting ingrained in the process. Others are just using it like a search engine or creating pretty pictures and memes. 

We marketers need to be in the middle. We need to know enough to be useful, enough to organize and complete a task with the right AI tool, and enough to recognize when a human needs to step in to correct any errors.

The only way to get to that point is to start experimenting now. Try out all of the platforms. Make mistakes. Push the models to their limits because it’s important to know what those limits are.

We are at an inflection point in AI adoption, and that makes it the right time to start training on it. It would be nice if you could do it on the job. But sometimes the easiest way to learn is on a project you can apply in your personal life. 

Your company might be slow to adopt AI. But you don’t have to be. 

The post Your company may be slow with AI, but you can’t afford to be appeared first on MarTech.