FEATURE25 February 2019

How business can harness AI

x Sponsored content on Research Live and in Impact magazine is editorially independent.
Find out more about advertising and sponsorship.

AI Data analytics Features Impact Technology

Unilever’s Stan Sthanunathan has co-authored a new book exploring how marketers can embrace artificial intelligence. He talks to Jane Bainbridge about using these tools to reach customers more effectively.

Can harness AI

To most people, AI stands for artificial intelligence. But for Unilever’s executive vice-president, consumer and markets insight (CMI) Stan Sthanunathan, the better descriptor – and the one used internally at the FMCG giant – is augmented intelligence. “That means using artificial intelligence to enhance human intelligence,” he explains.

This subject is currently front of mind for Sthanunathan as his latest book – AI for Marketing and Product Innovation, written with Dr A K Pradeep and Andrew Appels – is published. He sees AI as an essential part of any marketer’s toolkit. “The amount of data flowing through is increasing exponentially and it’s impossible for anyone to mine it all – so using AI is inevitable, or you’re at risk of getting left behind,” he says.

Like all businesses, Unilever has had to make important decisions about how to incorporate AI into its frameworks – the first being how many methods to adopt.

Machine learning

“We concluded there’s no one solution for all the problems,” says Sthanunathan. “The space is rapidly evolving, so we will look at a whole range of tech companies, pilot them and experiment with them. We have scores of techniques – AI approaches – we’re using internally. There are some we’ve tried and censored; that doesn’t mean we’ve narrowed down the list of partners we’re working with – it’s actually increased over time.”

Like many organisations, Unilever is using AI and machine learning (ML) to mine structured quantitative data. But it has also been used to look at subliminal cultural influences – from the TV programmes that are watched, to magazines and blogs that are read.

“We applied machine learning to understand what big themes are emerging and to see if it can predict what’s coming in the near future.”
Initially, this process produced so many trends, Sthanunathan says, that it was “almost humanly impossible to make sense” of it. So, Unilever added two other dimensions – brand architecture and target markets.

“We’ve done enough qualitative research to know what those people think. So we bring brand architecture and mind drivers of consumers and put them through a blender – what comes out is a bunch of trends that are relevant for that brand and target market,” he explains.

It is these trends that the company’s brand marketers can use as they see fit. “We’re not saying this is prescriptive and thou shalt do this. We distil all the things into possible areas and marketers can look at – and respond to – them. We have given them the guard rails,” Sthanunathan adds.

A criticism often levelled at marketing today is that it’s obsessed with short-term gains, based on easily measurable attributes, to the detriment of long-term brand building, which may be much more difficult to quantify. Doesn’t AI exacerbate this problem?

“Everyone has to manage the next quarter and that’s a big challenge for businesses. What do you do to create a long-term sustainable idea? The previous example is all about long term,” Sthanunathan says. “I believe it must be a blend between winning in the short term and winning in the long term. Winning in the short term requires quick response to the market conditions to protect your brand.”

Citing one recent example of how AI helped Unilever address an issue quickly, he refers to a negative post about Lipton green tea with lemon, shared by a consumer on social media. The post said that if you pour the teabag contents on a black background you can see worms squiggling. “People saw the video and were outraged – it spread like wildfire,” Sthanunathan says.

Unilever’s algorithms spotted it early on and its R&D people quickly identified that the ‘worms’ were dehydrated lemon strands, which – because they are so light – appear to move, even in the slightest breeze. Unilever quickly made a video explaining all. “From start to finish, that issue was resolved in about four days. If you don’t have the capability of identifying issues at an early stage, you’re in trouble. It’s make or break for a brand.”

Good enough

In this brave new world of AI and ML, marketers must get to grips with new possibilities quickly. Does Sthanunathan think marketers are daunted or excited by the prospect of what this tech can do?

“A bit of both. Marketers must work in the old world and the new world – and that’s not easy. To that extent, they feel overwhelmed and marketing would be damned if people started believing that it was 100% science. It must be a blend of art and science. I see AI as releasing the time, so [marketers] can do what they’re capable of. Those that get it, really love the way it enables them.”

One of the biggest challenges for those working in marketing is managing multiple data sources and bringing together unstructured and structured data. Sthanunathan says there are two schools of thought: one says everything must be harmonised so it can be put into a single dataset and analysed; the other prefers to blend different datasets to create an integrated analysis. He tends more towards the latter.

“People have spent an enormous amount of time trying to create harmonised databases over the past 20 to 30 years, spending billions of dollars, and it is never perfect. So now people are saying 80% is good enough, speed is more important to me. More people are realising that perfection is the enemy of greatness,” Sthanunathan says, adding that people should get used to “perpetual beta”.

He believes it’s time for marketers to stop carrying on the myth that AI will kill creativity. “If you allow it to kill creativity, it will. That’s where the whole notion of augmented intelligence is critical – it’s a mindset.”

5 reasons why marketers should use ML in their strategies

  1. Real-time capability – consumers see ads and offers that change by the second, based on what they are searching for
  2. Reduction of marketing waste – by using behavioural data, marketing is more targeted
  3. Predictive analytics – ML can analyse big data, notice patterns and predict future occurrences
  4. Structured content – sentiment analysis that aims to determine the attitude of a speaker or writer, then recommends to marketers what to say, how to say it and the best time to say it, as well as how the audience is likely to react
  5. Cost reduction – overall, the money spent on marketing automation software is meant to be less than the money that would otherwise have been spent on sifting through collected data.


AI for Marketing and Product Innovation, by Dr A K Pradeep, Andrew Appels and Stan Sthanunathan, is published by Wiley