Why KitKat is Telling Canadians to Give AI a Much-Needed Break
For 85 years, chocolate brand KitKat has been utilising the slogan ‘Have A Break’. When you think about it, it makes sense on so many levels. Unwrapping dual chocolate-covered wafers that are meant to be split apart lends credence to the phrase, but also, let’s be real, sweet treats are always better enjoyed when one is not actively in a rush, or under stress.
However, what some people might not be aware of is that this phrase doesn’t just apply to humans. Sure, computer programmes aren’t going to be able to eat a chocolate bar, but as Courage associate creative directors Matt Miller and Tommy Yong discovered, telling AI systems to ‘Have A Break’ before completing a task literally returns better results. Across multiple subjects and questions, accuracy was higher, answers were more thorough, and overall performance was up - something demonstrated in a fun video that the agency put together on behalf of the brand.
Serving as a way to re-emphasise the slogan’s value in culture, build on the buzz around the topic of AI, and once and for all prove that KitKat knows best in this regard, the campaign proves the merit of not taking one’s foot off the gas, and, more than that, is just plain fun.
To learn more, LBB’s Josh Neufeldt sat down with Matt and Tommy for a chat.
LBB> What was the brief, and where did the idea of targeting AI with the ‘Have A Break’ tagline come from?
Matt> There wasn’t a traditional brief, really. For KitKat, we’re always looking for ways we can bring the ‘Have A Break’ tagline into culture. After finding out about how AI chat prompting could be optimised, it just fell together naturally.
LBB> In learning this, you mentioned drawing inspiration from a recent Google DeepMind study. Please tell us more about this! And as a whole, what was the research process like for this campaign?
Matt> We actually found out about it through my dad. He had gone to an AI summit in California and heard Google talking about prompting and using human language. The Google DeepMind study came out a little while later, and was the total unlock.
Tommy> Incidentally, this was all right around the time that we started working on KitKat. So, at that point, it just became a matter of pressure testing the ‘Have A Break’ tagline with chatbots.
LBB> What was this process like, how many AIs did you test across, and do you have any anecdotes from the experience?
Matt> We created a small team where we’d run a ton of questions through ChatGPT, and we got a variety of improvements with our results. It was really cool to see it working in real time. That’s when we started to rush to get a video into the market.
LBB> Specifically, what was the question-writing process like? And how did you approach checking the accuracy of break versus non-break-prompted queries?
Matt> The best way to measure accuracy was asking questions with pretty concrete answers. Math questions were a good starting point.
However, this can work with more abstract questions as well; for example, summarising a novel and its themes. When ‘Have A Break’ was involved, you would see a more in-depth explanation that was broken into bullets and subpoints.
LBB> In terms of results, was there an average percentage of accuracy increase when the AI was told to have a break? What were the key numbers that you found?
Matt> Like the Google study, most of the time we’d be batting around 10-15% ticks in accuracy. But that number could go up or down based on the question or the subject matter.
Tommy> There was even a case where we saw the accuracy improving all the way to perfect. A small up-tick, but still, going from 98% to 100% is quite the jump.
LBB> The accompanying spot is quite fun. Please tell us about how it came to life! What was the editing process like?
Tommy> After collecting our favourite questions and answers, it was all about building the story as a discovery. To do this, we worked with Courage’s in-house editor, Marcus Barrie. He’s super talented and had all these great builds when it came to the animation of the video. The video really took shape when we got the track.
LBB> Speaking of track, the sound-design is also excellent. How did you combine keyboard sounds and music to make something that fit the vibe?
Matt> We worked with Eggplant Sound on this. Our brief for them was using a machine-like sound reminiscent of a keyboard to drive the spot. The team came back with this unbelievable score that just kept building. It’s so great to be able to work with an original score so we can get every little beat to hit at the right moments.
Our favourite addition from Eggplant was the end mnemonic. The robot voice is a mix of the words ‘KitKat’ and ‘AI’, so you might hear something different every time you listen.
LBB> Do you have any memorable lessons learned from the making of this new campaign?
Tommy> When you have an idea that’s timely, don’t take your foot off the gas. KitKat believed in this idea and with its support, we activated it pretty fast. A lot of times, a good idea can die just because the momentum slows or the cultural moment you’re chasing passes.
LBB> What has the initial response been like? How have people reacted to the campaign?
Matt> The response from the industry has been incredible!
Tommy> But it’s nothing compared to the response we got from the tech world. Specifically, AI experts - people from Google, TED, etc. - have all given us big props for the work. It feels like we cracked an idea that uses AI in a really unique and ownable way for a brand like KitKat.
LBB> Finally, has creating this campaign reminded you that it’s important to find time to have a break within your own work?
Tommy> For sure. Stepping away from a project or a brief can give you much needed a mental reset.
Matt> And sometimes your dad goes on a work break to California and just hands you a total gem of an insight.
