This story appears in the April 2018 issue of strategy.

Marketers have kept themselves well-informed about the futuristic possibilities artificial intelligence, blockchain and voice-powered platforms can bring to their business. But before this technology can live up to what’s been promised, marketing departments will need to address a few nagging issues.

At the beginning of March, digital agency FCB/Six spun out its own consulting division in response to clients’ technology stacks not being set up to handle the agency’s creative or strategic ideas.

“We quickly realized they weren’t ready for [the ideas] we were recommending,” says Jacob Ciesielski, SVP of data and technology at the shop. “So we started going [down] a different path, trying to understand some of the challenges and work directly with their tech team to make recommendations.”

One issue, Ciesielski says, is that clients aren’t using their existing tools and systems to their fullest potential.

A common way marketers can access AI is through CRM platforms and marketing automation technology – either directly from companies like Salesforce or by integrating with AI platforms like IBM’s Watson. The link between AI and CRM is a clear one: artificial intelligence can be used to automate and predict the most effective personalized, one-to-one communications. But Ciesielski says, in his experience, most clients aren’t ready for AI. In fact, many of them are still struggling to structure their data to provide an accurate picture of consumer behaviour.

“Everybody is talking about delivering the right message on the right channel at the right time, but there’s a lot of systems still working in silos,” he says. “That isn’t giving them the full view of the customer across different platforms. Trying to deliver an experience based on a fragmented view is going to impact the quality.”

Adobe’s Digital Trends report found fully integrating their technology stack is a problem for even the most successful companies. Among the nearly 13,000 marketing professionals surveyed globally, only 25% of the top-performing businesses described their stack as being fully integrated, with 9% of the remaining businesses saying the same. In another survey by Accenture, 82% of Canadian executives expect their staff to be working with AI systems within the next two years. However, 84% of those same executives don’t believe they are doing enough to verify the accuracy of the data sets that power their AI systems.

An artificial intelligence platform will only work as well as the data it is “trained” on. Machine learning algorithms are fed data, helping them “learn” how to classify pieces of information and associate them with an action through constant reinforcement. The system is further refined after it’s put out into the world and  interacts with people, giving it more information to base its processes and decisions on.

But if a system is given an inaccurate or incomplete picture of consumer behaviour at the outset, any learning built on that is going to be flawed.

“If you put garbage in, you get garbage out,” says Andrew Lo, CEO and president of online insurance platform Kanetix. “We have data going back to 1999, but we still spent weeks cleaning up our data and turning it into something that made sense before we began our pilot with AI.”

“That isn’t giving them the full view of the customer across different platforms. Trying to deliver an experience based on a fragmented view is going to impact the quality.”

In September, Kanetix launched a project with Integrate.ai, a company building an AI-powered platform to optimize consumer engagement. Targeting auto insurance customers in Ontario, Integrate.ai used Kanetix’s consumer behaviour data to assign site visitors a score based on how likely they were to pursue a quote. The data included anonymized sales information, on-site behaviour, the amount of savings offered and user information (like their car model and driving history). That provided Kanetix with suggestions for who would be most likely to convert after being given a gift card, allowing the company to be more targeted with its offers.

“The surprise was we had to really roll up our sleeves to develop the actionable insights,” Lo says. “There was still a lot of back-and-forth to make sure we had the right operating metrics and that the data was responsive and actually delivering business value.”

And data is about to become even more complex with voice-powered platforms like Amazon’s Alexa and Google’s Assistant increasing their penetration, both through sales of devices like Echo and Home and by integrating those assistants into connected products made by other companies. In the conference hall at CES this year, what was once a prestige feature is now being made compatible with hundreds of new products from companies of all sizes.

Damien Lemaître, SVP of media at digital agency Isobar, says the ubiquity of voice platforms will give brands access to data they might not have had before and change the way they speak with consumers. “But only if marketers make sure they understand what that data actually says about their consumer; structure it so it’s accurate and doesn’t drive people away with a device they think is always listening.”

Responsibility, accountability and eliminating the “creepy factor”

A survey by Sklar Wilton and Associates found 75% of Canadians are concerned with the impact AI will have on their privacy. When asked to describe how they feel about the adoption of AI, 40% of those in the same survey picked words that can be best described as negative: “concern,” “fear,” and “anxiety.”

The issue isn’t that companies are collecting more data than before, says Jodie Wallis, managing director for artificial intelligence at Accenture Canada. It’s that good AI uses data more effectively, and to a degree that makes it more apparent to consumers how much of their information has been collected.

“There is a regulatory framework around responsible use of data, but what’s more important for marketers is the ‘creepy’ factor,” she says. “If a marketer is using data I didn’t expect them to have, whether it’s compliant or not, it may make me feel uncomfortable.”

Beyond ensuring your data comes from a trusted source and is collected with consent, Wallis suggests marketers fully understand what they are collecting and how it fits into the consumer’s life. She also recommends a simple “human test” – if you do something with AI that creeps you out, chances are it will do the same to your customer.

Bias is another ethical issue. AI could be used to analyze resumes to fill an open job, but if it’s trained on data related to who previously filled similar roles, the system could be biased against candidates that are women or people of colour. In a marketing context, that bias could impact who a targeted message is delivered to – which works if you’re trying to reach people you’ve historically performed well with, but less so if you don’t want other groups to feel alienated from your brand.

“That is the difficulty, because historical data is all we have,” Wallis says. “We don’t have future data. But at least we can understand the problem, and be mindful of the types of biases we need to look for.”

That’s partly why transparency and accountability are big factors for those looking to use AI responsibly. This means knowing exactly how AI comes to a decision, and who is on the hook if something goes wrong.

“What’s more important for marketers is the ‘creepy’ factor. If a marketer is using data I didn’t expect them to have, whether it’s compliant or not, it may make me feel uncomfortable.”

Wallis says understanding decisions in a low-stakes situation like automated product recommendations in online shopping may be less essential than when a customer demands to know why they were denied an insurance claim.

As an insurance company, Kanetix is subject to regulations that prevent an AI system from automatically performing an action based on its analysis, requiring a human to actually “push the button” (so to speak) and send an offer. But Lo says he’d recommend that kind of stop gap to any company, even if they aren’t mandated to.

“Automakers are not thinking about getting autonomous cars on the highway right now,” he says. “They’re thinking about integrating them into a ‘smart city’ where pedestrians won’t be walking in places autonomous cars are driving.
“You can’t just let AI face the customer and have it tackle moral issues or potentially do harm. There should always be a layer of abstraction. We do the same thing with chatbots, because we don’t just let it advise customers on decisions. We can see what the responses are and intercept a conversation at the right moment.”

The coming blockchain wave

Think of it as a scoreboard at a baseball game. Everyone in the stadium can see when a run is added, and once it’s on the board, it cannot be takedown. Now imagine using a special code to see which players scored each run.

Transparency is becoming an issue on marketers’ minds as the hype around blockchain continues to swell. First developed as the backbone of cryptocurrencies like Bitcoin, blockchain is a digital ledger of information and transactions. The ledger is encrypted, but also decentralized, with the same information stored across multiple systems and tracking every change made to it. This makes information stored on blockchain secure, but also accessible.

Businesses in sectors beyond cryptocurrency are finding useful ways to incorporate blockchain. Financial and insurance companies, for example, can give customers faster approval for new products and services because blockchain makes it easier to securely access, share and verify private information with third parties.

In marketing, a fully transparent ledger of info is seen by some as a solution for ad fraud – a driving force behind Unilever’s partnership with IBM to develop its own blockchain solution. It also has implications for CSR activities, giving industry watching groups and private citizens the ability to verify corporate donations and supply chain practices to ensure a company’s activities match their socially-conscious messages.

ATB Financial has been experimenting with applications for blockchain since 2016. It was the first to use the technology to make real-time payments over international borders; has experimented with Alberta’s oil companies to make transactions faster; and partnered with non-profit Sovrin Foundation to store people’s personal information and combat identity theft.

While ATB is ahead of some of its competition in exploring blockchain, Wellington Holbrook, chief transformation officer at the company, says it’s playing a “wait and see” game.
“We can experiment with it, but until our customer or other businesses start to embrace it, it’s kind of like being the only person with an email address,” Holbrook says. “But once the tide turns, it will go fast. It’s just a question of when, and that’s what everyone’s trying to be prepared for.”

He adds that once blockchain reaches mass adoption, its capabilities could result in a shift in consumer expectations. People might want more transparency and easier access to secured private data from companies, whether they are using blockchain or not. Holbrook likens it to the Beta versus VHS format war, consumers are demanding security and convenience around their personal data, but it might not be blockchain that provides it.

Taking those first steps

Regardless of the tech, it’s important to ensure it addresses your needs before investing. Ciesielski says most data and CRM platforms are “about 80%” the same.

“The challenge our clients are facing is they’re being sold products based on features and functionality,” he says. “What they’re not realizing is they are just buying a shiny new box, and it’s not always one that will solve their problems, because their infrastructure is not set up in the right way. Sometimes a tune up or reconfiguration of something they already have might solve the problem.”

Wallis say forethought is especially vital for a company to keep in mind before using AI. Having a sense of the problems AI is being used to solve can help marketers create a data infrastructure that helps the system work towards the right goal. And to do that, be ready to roll up your sleeves.

“You can’t just let AI face the customer and have it tackle moral issues or potentially do harm. There should always be a layer of abstraction.”

“What people stumble on, especially in larger organizations, is wrangling all of the historical data you have about your customer into your new data infrastructure,” she says. “You need all of the data to be clean and readable, and sometimes you’ve got different versions of old data and have to fill in gaps. That’s hard and expensive and there’s no way around it.”

Blockchain and AI require a great deal of computing power, and both Wallis and Holbrook agree using a cloud-based solution from a trusted partner is the best way to avoid spending on your own servers. However, that solution needs to be integrated with other systems. As fundamental as having an integrated data set is for setting up AI, it’s just as important to be able to feed the insights it creates back into an organization.

“Even if you get to the point where you generate really interesting insights using machine learning algorithms, unless you can embed action from those insights into your processes, you haven’t gained anything,” Wallis says. “A lot of traditional businesses still struggle with this.”

Ciesielski says another issue that can contribute to not having integrated tech is a lack of communication between the marketing and IT departments. Holbrook says ATB has had success addressing this by having members of the marketing team embedded within his tech and transformation division.

“What we’re talking about now are solutions where data and business processes are inextricably linked,” Wallis says. “It’s important that both teams recognize that working together is the way to gain a competitive advantage. We’re starting to see organizations that are fantastic at using technology in their marketing functions have been the ones with very tight relationships with marketing and IT.”

Harley-Davidson, meet Albert

To reduce seasonal overstock when it’s chilly and hitting the open road isn’t a priority, motorcycle dealership Harley-Davidson NYC engaged AI-powered marketing platform Albert.

Using the dealership’s KPIs, past campaign data and intelligence about existing customers, Albert identified effective combinations of creative and messaging; found new target audiences; and changed the marketing mix on the fly.

The implementation resulted in a 2,930% increase in leads per month crediting 40% of sales over six months to the project. Prior to this, the most bikes Harley-Davidson NYC sold over a weekend was eight; during the first two-days of working with Albert, it sold 15.

Unilever and IBM fight ad fraud with blockchain

When Unilever CMO Keith Weed announced his company would work with IBM to adopt blockchain, it gave credence to an idea for combating online ad fraud. Using a secured, transparent ledger provides the ability to see where impressions come from and verify that ads are actually being seen by the intended audience. The fact that data cannot be tampered with or added to the ledger without all parties – brand and agencies – agreeing to it means it also resolves discrepancies and brings trust to the relationship.

That might seem arduous in an online world where speed is paramount, but IBM has said it’s committing to a principle called “smart contracts” in its work. Smart contracts can instantly initiate a digital media buy after views are confirmed as coming from a legitimate source weeks or months after the buy already happened.