Speaking at Advertising Week Europe, Google’s EMEA head of media, Emily Henderson, explained that AI could be deployed in advertising beyond optimising ad-buying and into creative executions.
Working with creative agency Essence, the search giant needed to go beyond a dynamic banner in order to showcase its Google Home smart speaker. (For a full treatment, read WARC’s report: Beyond the magic banner: Google taps AI to sell AI.)
The Guardian, the UK-based media organisation, as a long-time partner and a premium publisher, opened up its content API so that Google’s new banner idea could adapt to the content on the page.
Working across teams in a ‘hackathon’ session, all three organisations – client, agency and publisher – brought engineers, creatives and planners to the table to think up an ad that could work with the Guardian content while also depicting the Google Home’s capability.
“We thought, very quickly, what we could do is to show the Assistant recommending dishes that would accompany the recipe that you were looking at on the [Guardian] site,” said Andrew Shebbeare, co-founder and chairman of Essence.
Initially, using a simple keyword search to work out if the recipe in question was sweet or savoury, the team thought they could recommend additional ingredients or accompaniments that would work with the dish. However, words like ‘tart’ which can describe both savoury and sweet dishes caused problems.
The solution was to focus the understanding on images as well as text using Google’s own AutoML program, which would create a model that the content could be served against.
“So we’ve analysed our image, we reckon that’s definitely a sweet dish, we cite the copy and recognise that the recipe that’s being built here is for a lemon and ginger friand (whatever that is). And we’ve also analysed the text and extracted the fact that the key ingredient in this recipe is ginger”, Shebbeare added.
With that information, the ad could illustrate a potential command to the Google Home device, with the copy reading: “Hey Google, add lemons and ginger to my shopping list.”
Ultimately, the lesson here is that AI, or more specifically machine learning, can help to create contextually relevant copy at scale by understanding both words and pictures.
You don’t need a computer science degree and you don’t have to use only Google’s platform to access services like this. Amazon and Microsoft have comparable products that make machine learning accessible.
Sourced from WARC