The Drum Awards Festival - Extended Deadline

-d -h -min -sec

Digital Transformation Artificial Intelligence Creative Works

How to make the perfect AI prompt: Add plenty of data

By Lindsay Hong, CEO & Co-founder



The Drum Network article

This content is produced by The Drum Network, a paid-for membership club for CEOs and their agencies who want to share their expertise and grow their business.

Find out more

July 8, 2024 | 7 min read

What makes a good prompt engineer? The quick wits of a stand-up, the savviness of a data scientist – and lots of good data, says Lindsay Hong of SmartAssets.

A glass flask containing a green liquid surrounded by a white vapour

A good prompt engineer should know just the right question / Jordyn St. John via Unsplash

By now, all marketers should have grasped that impactful content starts with loaded data and a deep consumer understanding.

We have oodles of creative data available at our fingertips. Now, with just a splash of AI, marketers can make sense of previously unfathomable data sets and gauge true creative impact beyond performance metrics to understand why some ads perform better than others.

But if we know this to be true, then the million-dollar question is: Why aren’t all brands taking advantage of these creative insights?

We’re now about a year and a half out from the explosion of a rather extraordinary AI boom. And, outside the hype of just talking about AI, around 65% of companies still haven’t taken a stride forward to being the adopters of change.

Powered by AI

Explore frequently asked questions

Calling the prompts

An overlooked opportunity to ramp up content with AI is bringing prompters on board. Or ‘prompt engineers,‘ if we’re being formal about it. Today, roughly 1,879 prompt engineer-related job opportunities in the UK alone are advertised on LinkedIn, and jobs requiring AI skills are increasing 3.6x faster than other job postings.

Yet, when we talk about AI upskilling and prompting, what exactly do we mean? In my opinion, prompt engineers need to combine the linguistic creativity and think-on-your-feet chops of a stand-up comedian; the gift of gab of a content marketer; and the savviness of a data scientist capable of having full-length conversations with bots and manipulating them to produce the precise outcome that they want.

A bit more obviously, a prompt engineer should also be like any other good briefer. Prompters should have a great depth of knowledge in their subject area and know how to get exactly what they want from DALL·E, GPT4, or whatever the latest software may be. Yet, if they could back it all up with data, wouldn’t that be even better?

Let’s say an abridged prompt reads something like: “Create a 30-second video clip. It should include our signature washing powder and prominent branding. The clip should convey strong feelings of joy. Its objective is to boost engagement and improve click-through rates,” and so on.

The prompter could feed an engine with all the information necessary to produce a clip that they think is going to be great. But, what’s that based on? And how do they know this will work if there’s no objective data to inform their prompts?

Devil in the data

If creative decisions aren‘t backed up by data, they are usually instead based on subjective opinion and amount to shooting-in-the-dark guesses. Thankfully, data crunching and consumer insights are the advertising 101. But, what’s important to ask is: are these insights being used to spur creativity, and are they even touching the production process?

For example: How do you know what your audience wants to see? What products, colors, scenery, emotions, and calls-to-action will resonate? How can you improve memorability and attention rates to make your creative truly effective for your target demographic?

In the past, we achieved this through surveys, focus groups, or other research-based means. But a new suite of creative data is bringing deeper insight than ever before, getting into what makes great assets tick. By leveraging ad tech – such as AI-generated creative tagging – prompters can now break into the deepest recesses of creative data to see beyond the asset, and into what the asset consists of, and finally answer why it worked.

These insights are why the latest ad tech – infused with AI – has essentially become prompting fuel. We should now be looking at this AI-data combo as an informant – and in many cases a catalyst – to support creative ideas and better creative effectiveness. And this means letting go of the idea that AI will eventually stifle creativity.

Prompting and creativity

It’s not all about ballsy campaigns, striking colors, and quirky calls to action. These can be effective, and will always have a place in advertising no matter what. But where bold content screams, effective content often whispers.

The role of a prompter shouldn’t be to make a half-baked extrapolation on what might work, but to abseil down into creative data recesses and understand what we should be feeding into prompts.

The good news is that even if brands haven’t been harvesting creative data for the past few decades (with countless iterations and A/B tests), a creative repository and campaign data are more than enough to get started for a creative health check.

The tech space is evolving rapidly, and some new roles can evolve almost as soon as they’re introduced. But creativity is, for the most part, a distinct human quality, which can be both measured and stimulated by AI.

Digital Transformation Artificial Intelligence Creative Works

Content by The Drum Network member:


Better ad creatives with AI

Find out more

More from Digital Transformation

View all


Industry insights

View all
Add your own content +