Getting better results with AI

Not getting the results you expected (or hoped for) with AI? Here are a few troubleshooting questions to ask yourself.

Author

Mariah Parker
July 27, 2023

Good afternoon, Collaborators!

Have you tried using an AI chatbot or content generator, gotten a (bad) first draft, and given up?

Did you implement an AI initiative, work on it for a few months, and realize that you have nothing to show for it?

This week, we’re looking at the most common reasons why you aren’t getting the results you want with AI, and how to improve your performance. Let’s dive in!

THE DEEP DIVE

Why you aren’t getting better results with AI

Not getting the results you expected (or hoped for) with AI? Here are a few troubleshooting questions to ask yourself:

Are you asking the right questions?

AI is only as good as the direction you give it. If you ask the wrong questions, it will give the wrong answers.

If you are using AI to solve a problem, make sure the questions you are asking it are tied to your real goals.

For example, if your goal is to increase revenue:

  • Try asking ad platforms to optimize for purchases or conversions instead of traffic. As Zillow surfing has taught us, website visitors aren’t necessarily people who actually buy.

  • Ask an AI model to identify which customers are most likely to stay instead of which customers are most likely to leave.

Is AI working on solutions to the actual challenges you’re facing? If not, how would you change the questions you are asking your AI tools?

Are you using the right data?

The success or failure of AI depends on the data used to train it. As George Fuechsel, an IBM instructor and programmer, likely said, “garbage in, garbage out.”

If you are using your own data to train AI models, take a close look at your data quality.

  • Are you pulling data from multiple sources or a single source of truth?

  • What gaps exist in your datasets?

  • Are there duplicates, outliers, or irrelevant data in your dataset?

You may need to address data quality issues to get better results from AI.

If you are using an AI content generator, you may not control the data it was trained on. However, you do control the quality of the final materials you use. Here are a few ways to reduce inaccuracies in AI-generated content.

What are you including in your generative AI prompts?

If you are having trouble creating high-quality, AI-generated content, your prompts may be playing a role.

The more detail you provide to AI content generators, the better the output. Make sure your prompts include:

  • Purpose - Make sure the AI knows why it is writing content. Should it educate or entertain customers? Is the content conversion-focused or awareness-focused?

  • Action - What action do you want users to take after reading the content?

  • Format - What form should the output take? Do you need a social media post, a table, or an internal email?

  • Audience - Who is the target audience for this content? 

  • Tone/style - What style or brand voice should the content generator use?

Are you giving feedback?

The law of (bad) first drafts also applies to AI.

Like humans, AI can’t improve without feedback. Ask chat-based AI to edit the output to make it shorter, funnier, more empathetic, or less sarcastic. You can ask it to make changes just like you would ask a junior marketer on your team.

You may be surprised by how much better the second draft is than the first.

Are you thinking about how AI can replace humans, or how it can augment them?

Generative AI tools are incredibly helpful.

However, as G/O media learned when they published unedited AI-generated work, AI tools also make mistakes.

AI is not ready to work without human supervision.

In these cash-strapped times, many marketing leaders are thinking about how to use AI to replace staff. However, you will get better results if you think about how to use AI to help your staff be more productive and free their time for higher-order tasks.

How can your team work together with AI to achieve its goals?

PROMPT OF THE WEEK

Creating video ad scripts with AI

Tone, audience, format, and purpose work together in this prompt to create highly customized video ads.

Fill in the prompt with your brand details and try it out!

Write a 15-second [channel] ad script for a company that [summary of company products/purpose]. The video script should be [empathetic/funny/uplifting]. The ad should appeal to [audience] and encourage them to [complete call to action].

AI NEWS FOR BUSY MARKETERS

A(I) World of Opportunity