8 things you need to know about AI in marketing right now
- 11 May, 2017 06:08
Artificial intelligence (AI) is THE buzzword of 2017, and is quickly reaching the dizzying heights of ‘cloud computing’ and ‘big data’.
And like the game-changing technologies that have preceded it, AI’s implications on marketing are extensive.
All the biggest martech vendors are rapidly adding AI into their arsenal: From IBM’s Watson cognitive computing platform, to Salesforce’s Einstein, Oracle’s Adaptive Intelligence, Adobe’s Sensei, and Marketo’s Adaptive Engagement platform. Equally, there are plenty of startups entering the space – Conversica, Ampsy, Sentient and Complexica, to name a few.
Over recent weeks, and during the recent Oracle and Marketo digital marketing summits in the US, CMO heard from and spoke to a raft of different industry players about what actually fits under the umbrella term ‘AI’, the ramifications on the way marketing operates, and importantly, how AI could change consumer buying and engagement behaviour.
Here, we provide a snapshot of some key things you need to know right now about AI and where it’s headed.
1. AI means lots of different things to different people
It’s a new technology, so it shouldn’t surprise many to know the term ‘artificial intelligence’ means lots of different things to different people. In a media roundtable at the Oracle Modern CX Summit in Las Vegas, Oracle’s VP of products and data science for its Adaptive Intelligence AI platform, Jack Berkowitz, agreed definitions were problematic and sought to provide some clarity.
“People look at it and are trying to sell something, or scare someone. It’s about having a machine understand the context of a situation and then give you some aid to be able to amplify your capabilities in that situation,” Berkowitz said. “Think about what Google Translator and image recognition is doing: It’s doing something that’s automated, but it’s actually about understanding the context of that situation. It does that by combining and analysing lots of data.
“There may be jobs that go away as a result of AI, but this also will create new opportunities, new capability for folk as it’s our assistant to the next step. Think about it in the same way you think about automation.”
Berkowitz said AI could help a brand adapt marketing to a prospect or customer in real time, gain access to data at scale in a less stressful way, or optimise activity to be more successful. But whatever the scenario at this point, applications of AI requires a partnership between people and systems, Berkowitz said.
“You choose the partnerships that make sense for you,” he said.
Many in fact argue AI right now is just advanced, or more intelligent automation. That’s certainly the view of Oracle Marketing Cloud's group VP of customer success APAC, Paul Cross, who caught up with CMO to discuss AI’s implications during the summit.
Thanks to the work done to date around building the martech execution stack, marketers now have an end-to-end execution engine, making it simple to apply AI to it, he said. AI for marketing couldn’t exist without two core things in place: Data and the execution framework, he claimed.
“The last two decades have been about building that ecosystem,” Cross said. “Not many other sectors have the platforms already in place so AI can be applied today. Once martech software companies realised AI could be applied, it was obvious. It’s an enhancement.”
What AI needs to work is data, content and a delivery framework, Cross said.
“The good thing is we’ve been chopping content up in a way that could be machine executed – that’s what we have been doing… in marketing through emails, websites,” he explained. “Over the last five years, we’ve brought in all sorts of data too, including behavioural, geolocation, transactional, and put that against a person so we’re able to better optimise content.
“Then we’ve followed up with recommendation engines, which could be seen as first-level automation. The hard work has been done, which is to have the delivery platform.
“So we have unshackled content into selectable components, and organised behaviour data to the nth degree. We’ve automated some of that. The next level is to utilise all of that capability and use AI for the purpose of optimisation.”
Oracle group VP of product, Steve Krause, agreed the use case for AI in marketing is already there: To match the right offer to the right person. Oracle’s approach to AI is to firstly take what the marketer does and make those processes behind-the-scenes smarter, such as things being done manually like creative formatting. The second objective is to launch systems like Adaptive Intelligent offers, that help the end user get better marketing, he said.
“In the world of AI offers, we’ve combined the AI with extra data to get the one-plus-one-equals three solution,” he said.
2. AI is going into everything
In addition, other marketing technology vendors, such as Marketo, Adobe, Salesforce and Emarsys, are rearchitecting their platforms in order to support AI processing in order to assist the trillions of data points being used to improve the way brands engage and respond to customers.
Marketo’s initial approach to AI for example, suggests AI can be flicked on much like a light switch to assist the physical marketer. The way the marketing automation vendor is building AI capability is to fuel what it calls its ‘Adaptive Engagement’ platform.
“With the flip of a switch, I can decide if I want to make the content, channel and the cadence adaptive, then the adaptive engine will select the best content to send at the right cadence and to the channel best to reach the audience,” Marketo’s group VP product management and UX, Cheryl Chavez, said at the vendor’s recent Marketing Nation summit.
Marketing isn’t the only profession gaining from AI, of course. By 2020, 50 per cent of companies will be using cognitive computing in their applications, according to IDC research. AI is going into search, consumer and business products and services, industrial applications and functional applications.
“All these quintessential AI capabilities - image processing, sensory input, pattern recognition, emotion recognition, sentiment analysis, predictive power, natural language processing and so on – is being offered as a service,” IDC principal analyst, Gerry Murray, said during Marketo’s summit. “There are interesting mash-ups in terms of drawing on these different services and putting AI into application environments.”
Emarsys’ global CMO, Allen Nance, said the big plus with AI for marketers is that it finally enables personalisation at scale beyond the human capacity.
“When I look at AI and its implications on marketing, it’s a simple concept people are making too hard. And that is human-driven personalisation at scale,” he said.
Nance was quick to point out machine learning is not new, nor is it necessarily the competitive advantage in and of itself. The whole idea of machine learning is to take an algorithm, apply it to data and learn something from it, he said. And to work well, AI must be vertically integrated into a platform, accessing the data foundation, intelligence layer, execution layer and deployment engines.
“If those are the layers of our marketing cake, the reality is AI can’t now be a module, it has to vertically integrated with all of that – it needs the data, the intelligence, the segmentation, then access to deployment engines in order to execute a campaign,” Nance said. “The best companies investing into AI are integrating it as a technical capability.”
3. AI will initially manifest as 'buttons'
Over the next 1-2 years, Murray claimed AI will be reflected through simple buttons that better automate tasks already being undertaken by marketers. As an example, he suggested an AI engine that will go and find look-a-like audience members, as well as tools to help with optimising segmentation models based on pre-determined criteria specified by the marketing function.
“In the end, it may be available as push button utility to many different types of marketers,” Murray said. “But that doesn’t mean the marketer understands all the underlying assumptions and data sets being used to define the model. But someone will need to understand that – you’re going to need to understand the back end to a certain degree to so you can trust it, and have a high level of confidence in the back-end model.”
4. AI requires lots of ingredients and importantly, good data
Like any marketing system, the technology can only go so far, however. Bad data in, and you’ll still end up with poor results out, Oracle’s Krause said.
“Part of AI is training the system and having good data and signal going in,” he said. “The marketer can benefit hugely from a system that’s not only smart algorithmically, but is also smart about its data sources.”
Comparing it to cooking in the kitchen, Murray cited widely diverse sources of “perishable ingredients” in an AI recipe. These include real-time data sources, to batch data, unstructured and structured data, all supplied through different dealers and providers, and each with different cycles, seasonality, and quality assurance processes.
“You’re bringing all these raw ingredients into a kitchen, where you have a relatively small set of people – data scientists - with specialised training and tools to create recipes,” he explained. “They’re trying to take all these ingredients then put these into a model that’s going to be consumed by the organisation’s ‘business manager’.
“But you don’t have chef at the table, or the data scientists in front of marketing directly, there is a vital role in the process: The waitress. They work with the customer, interact with them, they try to explain the food, tell them where the ingredients come from, preferences, and allergies. A critical role here is the role of business analyst.”
AI also requires content to exist first, Cross said. “AI is not useful without the communications and content piece,” he said. “There has to be an execution from it.”
According to Cross, access to audiences and their data is quickly becoming a level playing field thanks to players such as Facebook and the rise of digital connectivity. What will ultimately win the brand war still comes back to creative and content, he said.
“It comes back to the quality of the messaging – it has to have the quality pieces of content,” Cross said. “You still need ‘brand’ – brand building is no different for it’s what determines the way you distribute your messaging. There’s still the question too of what is value, not just transactional.It could be driving reach, desire, advocacy - what AI is then doing is improved placement of messaging.”
5. AI is going to change the marketing skills mix
Third-party services are emerging to help with emerging AI products and services, but given the scope for AI, and its reliance on disparate data sources and modelling, it’s clear skills inside the marketing function are going to have to change. Murray suggested teams will need to have some understanding internally in order to use AI effectively.
He outlined a new ‘cognitive marketing’ team that’s required, incorporating data scientist (for data representation, featurisation and model building); the data-driven marketer (outlining business objectives, data cultivation, governance); the business analyst (project management, business assumptions, communications); the marketing technologist (infrastructure management, system integration and IT liaison).
“You could outsource some of these roles,” Murray said, but the capability is going to need to be there.
6. AI will transform consumer behaviour too
What’s arguably more interesting to Murray is how AI will pervade the consumer side of the brand interaction equation. One way IDC expects this to materialise is through bots on the buy side.
“They’ll shop for us, do research for us, run various tasks, and be authorised to make purchases in certain conditions,” Murray said. “As marketers, we’ll have to go beyond the buyer persona to bot personas.Using AI, bots will be doing preliminary research for us over time, looking for offers, intelligence, specifications. This has huge implications for marketers and especially advertisers.
“Think about the content you’re putting out: It may no longer being consumed initially by a human, but a bot. It’s going to be in the way. If the bot likes what you have and it meets the criteria it’s looking for, you’ll pass the test and the bot will let you through. The bot will pay attention to offers, and it may turn them off. It means a very big change to buyer behaviour. It’s going to take a while, but just look at Alexa, Siri – it’s going to happen.”
Several revs down the road, Murray predicted bots and language-based systems such as Alexa will become the executive assistants consumers depend on to manage their lives.
“It will no longer be passive, it’ll be active, constantly looking for new and cools things for us based on our definition and needs,” he said. “If you combine this with IoT [Internet of Things], everything becomes a compute environment – our homes, vehicles, places of business, shopping and more.”
7. AI apps for marketing are already out there
As a use case for AI and machine learning already in play, chief strategy officer at DigitasLBi/Spindrift, Michael Fasosin, pointed to work done to use chatbots for customer service with Bank of America.
At HPE, data is being used to ascertain intent and demand signals in paid, owned and earned channels, analysed via machine learning, to improve media planning, Fasosin said. In each case, it’s realising data is capital that’s the first step; AI then helps execute off the back of that.
IDC’s Murray went a lot further, outlining 10 AI-powered apps for marketing already in use in live environments:
- Chat bots – led by the launch of Facebook’s Messenger chatbots in 2016. Other local examples include Disney’s chatbot to automate Muppet chats with fans, and Domain’s property messenger bot, allowing consumers to find properties by interacting with Facebook.
- Virtual sales rep/email avatar – such as Conversica’s sales enablement tool, acting as a virtual sales assistant to better qualify sales leads before they’re passed on to human agents.
- Real-time sentiment analysis – such as the work undertaken by IBM’s Watson platform.
- Recommendation engines – good examples being the product recommendations produced by Amazon and Netflix for subscribers and customers.
- Live event monetisation – such as Ampsy, which tracks all social streams around a live event, such as sporting events or rock concerts, to get a sense of who is most engaged, sentiment around the event and influencers, then intervenes in real time to improve event experiences. The platform rewards influencers, engages with people having bad time, and has been proven to push the needle on onsite merchandising, Murray said.
- Cognitive commerce - Sentient is one example in this space, providing a distributed AI platform based on deep learning, computation and big data scale being used for digital marketing, shopping and funnel optimisation and finance trading strategies.
- Cognitive content - Persado is one of many examples of a platform leveraging AI for content production. The platform uses sentiment analysis on keywords to produce content that appeals to different kinds of audience. It uses an emotional ‘wheel’ to segment audiences, then works on content using language that will appeal to those audiences.
- Media mix optimisation – an example here is startup, Marketing Evolution, which is putting together an AI-powered attribution an real-time optimisation offering for in-flight creative rotation, targeting and media mix changes to the individual level.
- Attribution analysis
8. You need an action plan for AI
With AI right on the marketing doorstep, Murray offered the following tips for how marketers can get an action plan together in order to take advantage of the emerging capabilities. The first is to make sure you have use case with narrow swim lane if you’re getting started.
“AI isn’t going to generate prophecies, it’s really good for incremental steps in narrowly defined boundaries,” Murray claimed.
The second piece of advice is to focus on use cases and data sources and make sure you trust the data, Murray said, adding it’s vital teams understand data and assumptions going into any model used in AI.
Marketers also need to leverage knowledge resources in order to take advantage of AI.Murray’s fourth piece of advice is to avoid “black box cognitive/AI tools that defy understanding of how they work or reach decisions”. “If anyone tells you they can’t provide transparency… stay away,” he said.
As a final warning, Murray said marketers shouldn’t jump into any AI deployment if the outcomes are both uncertain and potentially harmful. You need a clear metric behind any investment or experiment, he said.
“You want to affect known metrics everyone in the organisation trusts,” he added. “That’s how you generate the confidence and economic models behind AI acquisition.”