The Greatest Challenge to Unlocking the Power of AI: Choice

There is a baseline of this newfound efficiency that will lift all ships. We can all have an AI bot write a pretty darn good first draft of meta tags, product descriptions, blog posts and thank you notes. That is a sea change that will lift all ships.

But ultimately, our goal should be to leverage AI to gain competitive advantage for our businesses. And that does not come from using the same tools as everyone else in the same ways they use them. You don’t see companies earning market share because they use email.

The opportunity to unlock business advantage will come from using AI better than your competitors use it. That means having the best tools and the best operations around them.

The biggest challenge for marketers in mastering AI will be choosing the right tools for the right jobs.

Not all AI is created equal. Different bots use different LLMs (Language Learning Models). And since they learn differently, they respond differently. I’m sure many of the curious out there have gone through the exercise of typing the same prompt into multiple Generative AI platforms.

For instance, when I asked several of them, “In 15 words or less, what is the best way to get a competitive advantage using Gen AI?”, here’s what I got back.

ChatGPT4: Leverage AI for unique data insights, automation, personalized services, and improved decision-making efficiency.

Claude: Improve decision-making with personalized, data-driven insights from assisted AI.

Bard: Use Gen AI to create unique value for customers.

Pi.ai: Synthesize and analyze data, drive innovation, and optimize processes.

Perplexity gave me 122 words…. and 6 sources.

Ah. Now, we’re getting somewhere. Many Gen AI bots hallucinate. When they don’t have information in their data set, they make something up. Or if they have conflicting information, they guess what is correct.

So, Perplexity can’t count to 15. But they quote actual sources so I can verify the answers it gives me. What matters more?

Jasper.ai gave me 506 words. They are worse at counting than even Perplexity.

But, they also prompted me to see if I wanted it to be a blank document, a new blog post, new from a template, or new art. So, if my goal is counting, Jasper is an abject failure. But if my goal is to create an SOW or a Project Brief or anything else that fits in a template, it may be the best solution.

The use case completely changes the perspective on what “best” means. Jasper also lets me choose the tone of voice to use in its response (I chose “smartass”) and even read all the content on Traction’s website to create our own branded tone of voice.

Of course, it was simply too tempting not to use the same prompt to generate “New Art”.

While much of the world is focused on changing their job title on LinkedIn to Prompt Engineer, a far better investment of your time might be identifying the use cases you need to solve for, and then researching the best tools to achieve them.

This problem will only accelerate. In a recent meeting of The Futureproof Project, an AI marketing think tank Traction curates with over 100 senior brand leaders as members, former Canva and Uber growth marketing executive (and current Traction AI Adoption Consultant), Sunil Subhedar, predicted the growth of smaller, more niche LLMs that deliver less costly and more impactful nuanced industry applications.

That means identifying the perfect tool for your specific problem may become harder and harder just as the value unlocked by AI becomes greater and greater.

Sunil also predicts that “Multimodal LLM” will be a game-changer for more complex multitask completion. Multimodal AI will allow analysis of multiple kinds of inputs—voice, text, images, numerical data—and use multiple models to process it. We’re also seeing AIs that are taking multiple steps to solve problems. For instance, ask AgentGPT to perform a task and it will first develop a checklist of all the steps required to accomplish that task, and then develops sequential prompts for each item on the task list to provide a comprehensive plan.

All of this exists today. Who knows what will exist in three weeks, let alone three years?

Finding the right tool on the diner menu of potential solutions can cause analysis paralysis. But don’t let it stop you. The best step forward is to order something that looks good, and give it a try.