The Drum 2018 Review: AI to bring better efficiencies, experiences and perhaps programmatic bias

A few big things will happen in the next 12 months. The first is that bias will become part of the conversation around AI. If organisations aren’t thinking about the data they’re using for training the AI systems they use, they run the risk of those systems exhibiting bias. That creates two problems. One is that the bias affects everything that’s informed by the AI, so if you use it programmatically, the bias will appear in the results. The other problem is that people tend to believe the outputs of AI - “the machine says it so it must be true” - even if what they’re seeing is the result of human bias in the training data.

The second big thing that will happen is the use of reinforcement learning will become common practice. So instead of having a few messages that you test by sending them out to everyone, the AI picks the best message for each audience segment based on their previous behaviour.

We’re also seeing higher and higher levels of automation of customer service across all channels. Everything is moving towards verbal or free-text-based communication now, because that’s how we humans like to communicate, so brands need to think about this; what their brand tone-of-voice is, what their brand’s voice actually sounds like, what the best way is of dealing with their customers as part of a conversation. We’re seeing huge engagement with conversational interaction - both text and voice - and click-through rates ten times higher than we’ve seen for any other channel.

Then in five years time we’ll be seeing augmented intelligence supporting everyone in their jobs. It’ll help us to do better what we as humans are good at. The AI will provide the information we need, in whatever way works best, so that we can focus on having ideas, fixing things that are broken - all those weird connections that humans are so good at making.

Andrew Cocker, senior marketing director, Expedia

AI is very much embedded within Expedia’s operations today, but we’re still only scratching the surface, moving from machine learning to machine intelligence.

For example, when customers first arrive on our app, site or through our voice integrations, one of our many machine-learning based algorithms may be able to adjust the hotels they see based on their preferences. Often, consumers want to book multiple travel items on one visit, so we’ll put other relevant options in front of them, often chosen by an AI-based algorithm. After they’ve stayed, an AI-based algorithm may even check and summarise the review they wrote about their hotel.

The scale of the opportunity is huge. Travel is an immensely complicated business, due to the size of data sets we need to manipulate in real time. Let’s say you’re looking at the options to fly from Seattle to Atlanta - this produced 19.7 quadrillion (19 with 15 zeros after it) potential round trip itineraries! Hence ML and AI are integral to helping customers find the right trip for them.

The direction we are heading is into much deeper personalisation and more automated, but personalised, customer service. A great travel agent would get to know their customers preferences, then personalise solutions for them. A great travel agent would help predict the pain points before they happen - such as price changes, delays due to weather - and automatically change pieces of the itinerary to ensure a smooth end-to-end experience. AI will be the beating heart of this relentless personalisation drive.

For marketing in general over the next five years, we’re likely to see huge increases in the efficiency and incrementality of our performance. Machine-led approaches are already significantly outperforming humans at expedia and this trend is only going to increase.

The area of emotionally-driven, longer-term strategic communications will greatly improve, as AI will have proven that it’s ultimately the most efficient way of driving sustained long-term shareholder value. Brands will be expert at tugging on the right emotions at the right time, to elicit a specific response as the lines blur between entertainment and advertising.

Alastair Cole, chief innovation officer, Engine Group

Two areas in which Engine is using machine learning are: helping brands predict customer satisfaction; and improving experiences.

Using data collected along the customer journey we're able to identify customers we believe are likely to spread negativity about their experiences. We’re then able to intervene using direct channels of communication, and proactively address these challenges. Our ultimate aim is to make more customers feel satisfied so they’ll buy again, and spread a positive message about the brand - leading to greater profitability.

We’re also employing machine learning techniques to improve customer experiences, increase basket sizes and grow share-of-wallet. Our clients are benefitting from new voice services that create frictionless interactions, and from tools using image recognition to identify existing products and behaviours, and make real-time recommendations.

Twelve months from now AI will be helping businesses deliver hyper-personalisation at scale. Business leaders appreciate they must create high-quality experiences that differentiate, and personalisation is one way to achieve that. As the availability of GDPR-compliant data increases, businesses will be motivated to experiment with more advanced hyper-personalisation strategies. These “new nudges” will affect behaviour change first by enhancing digital communications, then through catalysing the next wave of hyper-personal recommendation engines.

Five years from now AI will be improving every aspect of operating a business. It will help eliminate some tasks entirely, streamline the ones that remain, and facilitate complex task automation. This is underlined by Gartner’s prediction that voice interfaces will be ready for wide-scale adoption in the digital workplace, within the next two years. We’ve already seen a few markers for this - not least the introduction of voice-based search within Google Analytics - and it’s not hard to imagine a smart speaker in the corner of every meeting room, ready to answer tough questions, and initiate complex procedures.

As long as these enhancements are seen by the workforce as part of a graceful and natural progression, they will be adopted and behaviours will adapt accordingly. The upside is potentially huge, with Forrester predicting that by 2020 businesses using AI to uncover new insights, will take $1.2 trillion yearly from competitors that don't.

Richie Barter, CEO, AltViz

We’ve seen a lot of AI being used in marketing automation, trying to find patterns in companies’ data to link things up across the customer journey and to allow them to do attribution, but that’s probably marketing automation version 1.0.

Where marketing teams have an untapped opportunity is in thinking how the adoption of automation across the whole business can create opportunities for them. As supply chains are automated, for example, what does that mean for marketing campaigns? AI will take marketers out of their bubble.

One of the things we’re looking at is quality assurance around customer service calls. We’re trying to ensure they go well by using AI to monitor them. What opportunities could the data arising out of that create for marketing?

Then looking ahead, there’s the question of infrastructure. The adoption of the public cloud is progressing. People are cleaning up their data and bringing it together while rationalising their IT stack. So, the big opportunity will be co-ordinating and scaling automation across the entire business. There are currently pockets of automation projects in businesses, but as things mature, you’ll need a cloud-based automation platform. I see the operational and IT teams involved in these discussions, but generally not marketing. Marketing isn’t looking into the supply chain, for example, but what happens there will have an enormous impact on their campaigns and brand.

Darren Savage, chief strategy officer, Tribal

We’re increasingly working with clients to use the AI capability that already exists in their tech stack to analyse the huge amounts of unstructured data that they’re collecting. Then we’re starting to do predictive modelling, for example with fashion companies. They’re using purchasing behaviour across their stores to try and predict where fashion is going, where in the past they’ve tended to work off the catwalk shows. Now they can use AI driven insight systems to blend the two things.

The industry is currently in a state of informed bewilderment - there’s lots of data from lots of different sources, but we don’t necessarily know what it all means. Machine learning is very good at making sense of that sort of data, and in the next couple of years we’ll start to see AI insight systems being used in areas like product design and customer experience.

Then the flipside of that will be consumers using AI to manage their purchasing. Jeans, for example, are one of the trickiest purchases people make. Sizing varies from manufacturer to manufacturer, and even from batch to batch. For instance, someone we work with has built an AI system that brings together data including manufacturer’s sizing, user reviews, materials information, and knowledge of what’s in your wardrobe to find jeans that will fit you perfectly. That way you can buy online and not worry about the hassle of returning them if they don’t fit.

Then we’ll start to see AI help people create bespoke items. NikeiD has been very successful, because people see the value of being able to create your own pair of running shoes. But they could start to embed sensors in the shoes so that Nike can make recommendations based on your exercise behaviour. They could even predict when injuries are about to occur by sensing small changes in how you run. AI will be helping companies predict what customers will want to buy, but also helping customers navigate difficult purchasing decisions.

Norm Johnston, global CEO, Unruly

It’s still early days with AI and our understanding of its full implication for marketers. We've made good strides using machine learning to improve media targeting and optimisation by recognising repeatable patterns and building models around them. For example, we've enriched our UnrulyEQ emotional testing tools via IBM Watson.

Arguably, the biggest future impact for marketers is that advertising to AI will become just as important as advertising to humans. Convincing Alexa or Siri to recommend your product or your content will become critical in a world where AI becomes your personal gatekeeper, even making decisions on your behalf. As an advertiser it makes your job even harder. You need to either emotionally move people to go out of their way to select your product, or optimise your way into the algorithm to ensure you rise to the top. Entire teams will be focused on advertising to different AI ecosystems, just as search teams currently optimise for Google. Search Engine Optimisation will evolve into Siri Ecosystem Optimisation.