Before breakfast, I check my Facebook and LinkedIn newsfeeds for a quick synopsis of the day. As I jump in the shower, I hit “download” on a recommended movie on Netflix, knowing I have a long flight this evening. While wolfing down my cereal, I click once to buy a gift for a friend’s birthday next week. My iPhone pings to tell me that I need to leave now if I want to make that early meeting 54 miles away. And as I get in my car, I use voice activation to play my favorite Spotify playlist, and Apple Maps informs me it will take five minutes to drive to the train station this morning.
With real examples of demonstrable value in the market, we can no longer sarcastically joke that AI means “almost implemented.”
We are all being conditioned to rely on technology in our daily lives, not just for communication, but also for decision-making. This ever-deepening interface with technology is rewiring our brains to process information differently, as Nicholas Carr writes in The Shallows. It is the same with our customers.The ever-deepening interface w/ technology is rewiring our brains to process info differently, says @roughtype. Click To Tweet
Popular consumer apps have led to the unconscious mass adoption of advanced, predictive technology. And yet … while we are increasingly outsourcing our cognitive processes to myriad consumer apps and tools, the enterprise is only now waking up to this new level of customer expectation. This lopsided adoption is most clear when we consider that we now trust a car’s built-in collision-avoidance system to protect our lives, yet still question whether a machine can recommend what to write next in a marketing program or which customer should receive a new product offer.
We trust artificial intelligence to drive our cars safely but not to recommend marketing strategy.We trust artificial intelligence to drive our cars but not to recommend marketing strategy, says @andjdavies. Click To Tweet
HANDPICKED RELATED CONTENT: Cognitive Content Marketing: The Path to a More (Artificially) Intelligent Future
Inconvenient truth
Over the past 10 years, marketing automation has grown into a billion-dollar industry by promising to bring personalization and efficiency to marketing programs. The siren call of automated lead nurturing, lead scoring, and triggered responses to critical prospect activities has proved irresistible to B2B organizations: There were nearly 11 times more companies with marketing automation in 2014 than there were in early 2011 (SiriusDecisions), and 60% of companies turning over at least $500 million adopted marketing automation by 2014 (Raab Associates).
However, the inconvenient truth about first-generation marketing automation is that it is not really automated. It is a fantastic central workflow tool that can achieve scale, but it requires resource to set up, integrate, manage, and optimize. Indeed, in many B2B organizations, the phrase “feed the beast” has been accepted into marketing parlance as a way of describing the resource demands of marketing automation. Most fundamentally, there is the issue of rule creep. As you set up campaigns, you define business rules: “If A happens, then do B” or “If the individual has this characteristic, then put them in segment 4.” These can be simple to start with, but are always an inadequate reduction of complex and varied buyer journeys. So, you add more rules to make the campaign more targeted. And every time you measure results, the outcome is that more rules need to be written. Some of our enterprise clients estimate that they spend $500,000 per year on these manual elements of marketing automation – and that is disregarding the vital and significant investment in ongoing content creation.
While marketing automation promises the world, what it actually does is automate the execution of content marketing, while decision-making remains an impractically manual effort. It offers marketers a strong workflow and even insights, but fails to provide an automated way to act on those insights at scale. Fundamentally, the content in those systems is dumb; the system doesn’t understand what the content is about and who should read it. To track those looking at how to address this, Forrester recently started a new research theme it calls “content intelligence,” which it defines as “the use of artificial intelligence technologies to understand and capture the qualities inherent in any content.” As the marketing technology analyst David Raab says, “Something has to give: Either marketers stop trying to make the best decisions or they stop relying on rules.”First-generation marketing automation automates execution; decision-making remains manual effort. @andjdavies Click To Tweet
Expectation gap
In the face of relentlessly rising customer expectations, leading marketers are investing in AI-based tools – a category that encompasses everything from personalization tools that “learn” from individuals’ online behavior to recommend content more effectively, to tools that can detect minute patterns across massive consumer data sets and predict future behavior. These are some of the most interesting on the increasing list of potential applications for AI in marketing:
Content strategy – recommending what content to create next
Campaign strategy – recommending what sequence of communications to deliver
Personalization – recommending the right content for each customer based on behavior
Segmentation – clustering customers based on behavior or intent
Copy automation – automatically generating subject lines and descriptions
Lead or account prioritization – ranking leads or accounts by their likelihood to close
Sales strategy – recommending the right product/service offering and content to use in sales
Sales intent – predicting the right product offering, deal size, and close date
Retargeting – recommending the right content within retargeted ad units
Since the major marketing suites have yet to fully deploy or productize their AI offerings, adopting AI usually requires a blend of point solutions and data sets.
Indeed, marketers are increasingly piecing together their own technology stacks from best-in-class point solutions, allowing the technology to be built around customer need rather than vendor features. Especially in complex customer environments – for example, high-touch relationship sales with long purchase cycles – the application of AI promises to start bridging the gap between customer expectation and actual experience. This is most pertinent in global businesses, as AI solves for (and relies on) scale.
For Byron O’Dell, senior director of marketing at IHS Markit, employing predictive machine learning rather than marketing automation has been about overcoming the challenges of scale. He explains, “enabling marketing relevance at scale is challenging, but predictive machine learning is giving us a path to achieve this.”Predictive machine learning is giving us a path to achieve marketing relevance at scale, says @byronodell. Click To Tweet
Initially, most marketers are considering two key use cases: personalization and predictive lead scoring. Personalization entails matching content to the evolving customer need, particularly when content is produced at scale and often poorly classified. Predictive lead scoring is driven by the insatiable desire for new sales conversations, where the signals that identify an interested account are difficult to identify or uncover.
HANDPICKED RELATED CONTENT: Want to Scale Up Your Content Operations? 4 Things to Think Big About [Infographic]
Insights-driven business
These new approaches address a fundamental challenge: The buying process has changed, with the buyer increasingly empowered, informed, and connected, but enterprises are largely selling in the same way they always have. Using content to attract, engage, and convert is part of the solution, but leading marketers are also using content to understand the customer.Leading marketers use #content to understand the customer, says @andjdavies. Click To Tweet
In an increasingly competitive world, any business that does not understand its buyers will rapidly lose market share as new digital-first competitors grow. Disruptors obsess about their customer; they focus on delivering a superb and seamless customer experience; they are unencumbered by obsolete technology and rigid processes. They appreciate that gaining and acting on deeper customer understanding build competitive advantage.
Forrester Research is building a body of evidence around what it calls “insights-driven businesses.” One definition of these businesses is that they have no friction between the point of understanding the customer and the point of delivering the next response. There is a feedback loop that is completely automated. The cohort of businesses Forrester defines in this category – fast-growing companies innovating based on customer understanding and experience – should be truly terrifying to incumbents.
Marketing AI promises unstructured, real-time customer interactions that deliver value. Current rules-based systems simply cannot scale nor can marketing teams complete a manual process in the time required to deliver relevance.
Success factors
As an increasing number of businesses are investing in AI-based approaches, the commonalities among successful projects are becoming clearer.
Executive sponsorship – Time and again, clear executive sponsorship for the overall concept rises to the top of the list. While mid-level marketers may successfully buy point-solutions, larger organizations will find that to open the right data sets and drive overall business value, they eventually need an executive sponsor to champion a more automated approach.
Defined outcomes – Early innovators had to make leaps of faith without a known objective. But as the vendor landscape matures and client examples are documented, every project can and should have objectives linked to valued and measurable business outcomes.
Available data set – Most experts would agree a mediocre algorithm with a large data set always trumps a great algorithm with a small data set. Dig into the options available, clean up what you can, integrate new data sources, and run tests to see results.
Team composition – Although the aim of AI systems is to reduce manual tasks, the technology still needs to fit into a team and business process that understands its value. Increasingly, non-technical business users are being served, but for the meantime, it is important to ensure that the team understands data and are technical enough to grasp the strengths and shortcomings of an algorithmic approach. Perhaps more importantly, they must be humble and eager to learn, and data-driven (i.e., willing to link activity to results).
Vendor selection – Although there is a case for building in-house or using an agency for a bespoke application, the menu of options on the market from vendors is increasingly robust. To choose the right vendor, ask about the data set, try multiple competitive demos or trials, and push to understand whether the system is pre-trained or requires you to do it.
HANDPICKED RELATED CONTENT: New Tech Friends on the Marketing Block
Predictive enterprise
A shift toward the predictive enterprise requires an ideological and practical rededication to understanding the customer. The competitive advantage afforded by artificial intelligence is not based on the algorithm or the eventual application, but rather on understanding the customer in more depth – and acting on that insight in the moment.
The obvious obstacles are exclusively organization-centric: politics, technical roadblocks, resource constraints, and not-invented-here syndrome. Yet in a flat world, with disruptive new entrants focusing on a quality and seamless customer experience, the only sustainable option is to invest ahead of the competition.
To twist the overused Wayne Gretsky quote, it’s time to skate to where the market is going, not where it has been. The irony is that in this case, you don’t need to guess or rely on instinct. The customer already moved. As a customer, I expect a Facebook-inspired content feed, with the resultant privacy trade-off. I expect Amazon-like recommendations to be useful. And, a la Google, I expect you to anticipate my needs and offer help before I ask. Bring on the intelligent and predictive enterprise.
Thoughts on beginning
Initial forays into predictive marketing have hooked into the first-party profile data in large customer management and CRM systems. It’s not always clean data, but it is a good start. The deeper and more defendable approaches tackle a fundamentally harder problem: turning unstructured customer data into actionable insight.
Unstructured data, often called dark data, is largely unused within the enterprise, yet comprises 88% of all data gathered (IBM Research). At Idio, we summarize our approach to dark data with the thesis, “You are what you read.” What we mean is that the content you consume is highly indicative of your interests and highly predictive of your intent. AI-enabled tools analyze this dark data – essentially how your customers engage and behave with your content – to predict their interests and intent, and personalize their experience.
Consider using this project checklist to help your venture into predictive marketing:
Do I have executive sponsorship for an AI-based approach?
Have I defined several business outcomes?
Is there an urgency and clear time frame to achieve those outcomes?
Is there a data set to model?
Has my team bought in to the project?
Have I assessed the build-vs.-buy decision?
Have I created a short list of vendors?
Are their systems pre-trained or is there a lengthy training process?
Definitions of key terms
As you’re beginning to truly use the benefits of AI and predictive marketing, it’s important for everybody to be grounded with the same definitions. Here’s a brief primer:
Artificial intelligence (AI) is the science of building machines that do things that would be considered intelligent if done by a human.
Machine learning is the subset of AI that allows computers to learn without being explicitly programmed. Common machine-learning use cases are optimization (over time choosing the best option to achieve a set goal), identification (extracting meaning from images or text), anomaly detection (isolating an event that occurs outside of the norm), and segmentation (clustering based on inferred or known characteristics).
Content intelligence is the application of AI to content management, most notably the understanding and classification of content to improve targeting and measure performance.
Predictive marketing is the application of AI to marketing, usually to identify prospects, predict what they might be interested in, and recommend the next best piece of content or product information.
Conclusion
With this understanding of AI and some tips on how to get started, it’s your time to turn “almost implemented” into an AI reality to improve your enterprise marketing and truly understand and connect with your customers.
A version of this article originally appeared in the June issue of Chief Content Officer. Sign up to receive your free subscription to our bimonthly, print magazine.
Cover image by Joseph Kalinowski/Content Marketing Institute
Kommentare