• Leveraging Midjourney to generate images for storyboards

• Combining Midjourney and Photoshop to create mock-ups of packaging

• Developing a video treatment with Stable Diffusion

What have we learned about AI?

1. Large language models are a powerful tool for more than just writing copy.

Large language models lend a hand for rapid concepting and content outlining during the ideation phase. They allow users to generate ideas quickly so they can hit the ground running. LLMs can identify sentiment, translate languages and more, assessing the content and copy generated and relaying that information to the user. Used wisely, plug-ins, such as those that augment SEO activities, help create a more streamlined, integrated approach.

However, while speedy, LLMs often generate content that lacks a unique voice, fails to speak to the target audience or just plain misses the mark.

As one example, we asked ChatGPT to come up with a title for our webinar on AI using this prompt:

“Generate 3 webinar title options for a webinar hosted by a marketing agency for their clients that will cover these topics: Our approach to researching and testing new AI tools, Five promising uses of AI in marketing, What we’ve learned so far during our AI exploration. The tone should be conversational and light.”

Here are the suggestions ChatGPT provided:

• Navigating the AI Wonderland: Tales From Our Marketing Expedition

• AI Unleashed: Crafting Smiles, Clicks and Conversions in Marketing!

• Chit-Chatting with AI: Exploring Marketing Magic and Musings

• Brewing Success: A Sip of AI in Marketing and Our Flavorful Journey

In case you’re wondering, we did NOT select one of these titles, all of which were a little too dramatic and off-brand for our taste. Instead, we went with “A Conversation About AI and Marketing” (written by, you guessed it, a human).

Furthermore, most pre-trained data is current only as of September 2021. Worse, LLMs can “hallucinate” or generate false or incorrect information, generating work with the wrong calculations and facts.

These downfalls leave it up to the user to bring a critical eye and diligence to the editing process.

2. Advertising and marketing platforms are evolving their AI integrations.

The machine-learning subset of AI isn’t new — it’s been powering ad platforms for years.

Today, AI integrations in media-buying platforms such as The Trade Desk are making ad-buying and spend optimizations more efficient. Meanwhile, HubSpot’s Content Assistant generates copy for marketing emails, landing pages, social media and more.

However, the internal workings of these integrations are black boxed, meaning the user can’t view or influence them. As with LLMs, content generated by these AI integrations are not magic bullets and should be reviewed and revised with a critical eye.

3. Working with data can be powerful, but it comes with risks.

Manual processes make it difficult to identify patterns in large datasets. That makes AI tools designed to work with data especially attractive. Tools such as ChatAible enable rapid data analyses with built-in guardrails. LLMs equipped with data analysis features can also help explain key findings in an easy-to-understand way.

But be careful with sensitive information. Feeding data to an AI tool carries clear risks: The data could become compromised through unauthorized access or data leakages, posing security threats to you or your business. Exposure to biased data may also result in skewed analyses that reflect or amplify existing biases.

When testing these platforms, Wray Ward is intentional about safeguarding sensitive data by avoiding client names and identifying information. Instead, we include only general category identifiers and anonymize all data before submitting it.

4. Image and video creation are mind-blowing but risk a race to the middle.

AI platforms such as Midjourney, DALL-E 2, Stable Diffusion and Runway slash the amount of time and human effort required to create dynamic, eye-catching visual content. They’re superpowered concepting tools for testing image and video ideas. For example, at Wray Ward, we used Stable Diffusion to develop a video treatment, feeding the AI engine static images as a guide in the development of a dreamlike video execution. It was an effective, efficient way to bring an idea to life without having to invest time and money to capture original photography.

On the other hand, overuse of these tools may lead to generic content. Visuals generated by AI also create risks related to content originality and rights. This kind of content requires vigilant oversight to maintain quality and avoid problems down the road.

It’s also worth noting that, as with any platform, AI technology for image and video generation is only as effective as the person using it. Think about it: Anyone can get in a sports car and drive it around a track, but only a trained driver can push that car to its limits — without crashing — to win the race.

Mastering AI tools also requires a creative mind and a creative eye.

5. Coding for digital platforms is giving a boost to productivity.

GitHub Copilot is one of many AI tools developed to assist with code creation and error detection. These pre-trained, deep-learning networks reduce the need for extensive human labor. Automated optimization can lead to more efficient code outputs.

But these quicker-to-completion processes may also overlook nuanced issues or introduce bugs. In addition, too much reliance on AI for coding may limit innovative thinking. Like everything else I covered above, the outputs of these tools should not be taken as gospel or released into the world without careful consideration.

(Not) the Last Word

Our clients trust us to help differentiate and elevate their brand. And while many of the AI technologies we’ve tested show promise and, yes, aid our work, they can’t replicate humans’ ability to establish and maintain an authentic brand voice or develop a deep understanding of what makes the brand unique.

If anything, it’s clear that the craftspeople still matter, and these new tools are just that — tools.

The AI landscape is evolving at a breakneck pace. But in many ways, it’s still in the “crawl” stage, and it will be a while before it’s ready to walk, much less run. The most important thing is that we’re moving in the right direction, embracing the change while remaining wary of the risks. We can’t predict exactly what will happen next in this story, but we can promise you that at Wray Ward, we’ll never stop seeking ways to deliver better-performing work for our clients.

Want to read more about how we’re approaching AI or what these tools could do for your business? Check out the related content below.