How to Design an AI Marketing Strategy

Idea in Brief

The Challenge
At many companies, the advertising and marketing perform is quickly embracing synthetic intelligence. But so as to totally notice the expertise’s huge potential, chief advertising and marketing officers should perceive the assorted forms of functions—and the way they may evolve.
The Framework
Classifying AI by its intelligence stage (whether or not it’s easy job automation or makes use of superior machine studying) and construction (whether or not it’s a stand-alone utility or is built-in into bigger platforms) can assist companies plan which applied sciences to pursue and when.
Companies ought to take a stepped strategy, beginning with rule-based, stand-alone functions that assist workers make higher choices, and over time deploying more-sophisticated and built-in AI techniques in customer-facing conditions.

Of all an organization’s capabilities, advertising and marketing has maybe essentially the most to acquire from synthetic intelligence. Marketing’s core actions are understanding buyer wants, matching them to services, and persuading individuals to purchase—capabilities that AI can dramatically improve. No marvel a 2018 McKinsey evaluation of greater than 400 superior use instances confirmed that advertising and marketing was the area the place AI would contribute the best worth.
Chief advertising and marketing officers are more and more embracing the expertise: An August 2019 survey by the American Marketing Association revealed that implementation of AI had jumped 27% within the earlier 12 months and a half. And a 2020 Deloitte international survey of early AI adopters confirmed that three of the highest 5 AI aims have been marketing-oriented: enhancing current services, creating new services, and enhancing relationships with prospects.

While AI has made inroads in advertising and marketing, we count on it to tackle bigger and bigger roles throughout the perform within the coming years. Given the expertise’s huge potential, it’s essential for CMOs to perceive the forms of advertising and marketing AI functions accessible at the moment and the way they might evolve. Drawing on greater than a decade of expertise finding out information analytics, AI, and advertising and marketing and advising corporations throughout industries about them, we’ve developed a framework that may assist CMOs classify current AI initiatives and plan the rollout of future ones. But earlier than we describe the framework, let’s have a look at the present state of play.
Today’s AI
Many companies now use AI to deal with slim duties, reminiscent of digital advert placement (often known as “programmatic shopping for”); help with broad duties, like enhancing the accuracy of predictions (suppose gross sales forecasts); and increase human efforts in structured duties, reminiscent of customer support. (See the sidebar “Well-Established AI Applications in Marketing” for an inventory of some frequent actions AI can help.)

Firms additionally make use of AI at each stage of the shopper journey. When potential prospects are within the “consideration” section and researching a product, AI will goal advertisements at them and can assist information their search. We see this taking place on the on-line furnishings retailer Wayfair, which makes use of AI to decide which prospects are almost certainly to be persuadable and, on the premise of their looking histories, select merchandise to present them. And AI-enabled bots from corporations reminiscent of Vee24 can assist entrepreneurs perceive prospects’ wants, improve their engagement in a search, nudge them in a desired path (say, to a selected net web page), and if wanted, join them to a human gross sales agent by chat, telephone, video, and even “cobrowsing”—permitting an agent to assist the shopper navigate a shared display screen.
AI can streamline the gross sales course of through the use of extraordinarily detailed information on people, together with real-time geolocation information, to create extremely customized services or products provides. Later within the journey, AI assists in upselling and cross-selling and might cut back the probability that prospects will abandon their digital buying carts. For instance, after a buyer fills a cart, AI bots can present a motivating testimonial to assist shut the sale—reminiscent of “Great buy! James from Vermont purchased the identical mattress.” Such initiatives can improve conversion charges fivefold or extra.
After the sale, AI-enabled service brokers from companies like Amelia (previously IPsoft) and Interactions can be found 24/7 to triage prospects’ requests—and are in a position to cope with fluctuating volumes of service requests higher than human brokers are. They can deal with easy queries about, say, supply time or scheduling an appointment and might escalate more-complex points to a human agent. In some instances AI assists human reps by analyzing prospects’ tone and suggesting differential responses, teaching brokers about how greatest to fulfill prospects’ wants, or suggesting intervention by a supervisor.
The Framework
Marketing AI might be categorized in accordance to two dimensions: intelligence stage and whether or not it’s stand-alone or a part of a broader platform. Some applied sciences, reminiscent of chatbots or advice engines, can fall into any of the classes; it’s how they’re applied inside a selected utility that determines their classification.
Let’s have a look at the 2 forms of intelligence first.
Task automation.
These functions carry out repetitive, structured duties that require comparatively low ranges of intelligence. They’re designed to observe a algorithm or execute a predetermined sequence of operations primarily based on a given enter, however they will’t deal with advanced issues reminiscent of nuanced buyer requests. An instance can be a system that robotically sends a welcome e mail to every new buyer. Simpler chatbots, reminiscent of these accessible by Facebook Messenger and different social media suppliers, additionally fall into this class. They can present some assist to prospects throughout fundamental interactions, taking prospects down an outlined choice tree, however they will’t discern prospects’ intent, provide custom-made responses, or study from interactions over time.
Machine studying.
These algorithms are educated utilizing giant portions of information to make comparatively advanced predictions and choices. Such fashions can acknowledge photos, decipher textual content, section prospects, and anticipate how prospects will reply to numerous initiatives, reminiscent of promotions. Machine studying already drives programmatic shopping for in internet advertising, e-commerce advice engines, and gross sales propensity fashions in buyer relationship administration (CRM) techniques. It and its extra refined variant, deep studying, are the most well liked applied sciences in AI and are quickly turning into highly effective instruments in advertising and marketing. That mentioned, it’s vital to make clear that current machine-learning functions nonetheless simply carry out slim duties and want to be educated utilizing voluminous quantities of information.
Now let’s think about stand-alone versus built-in AI.
Stand-alone functions.
These are greatest understood as clearly demarcated, or remoted, AI applications. They’re separate from the first channels by which prospects find out about, purchase, or get help for utilizing an organization’s choices, or the channels workers use to market, promote, or service these choices. Put merely, prospects or workers have to make a particular journey past these channels to use the AI.
Consider the color-discovery app created by Behr, the paint firm. Using IBM Watson’s pure language processing and Tone Analyzer capabilities (which detect feelings in textual content), the appliance delivers a number of customized Behr paint-color suggestions which can be primarily based on the temper customers want for his or her house. Customers use the app to short-list two or three colours for the room they intend to paint. The precise sale of paint is then executed exterior the app, though it does permit a connection to order from Home Depot.
Integrated functions.
Embedded inside current techniques, these AI functions are sometimes much less seen than stand-alone ones to the purchasers, entrepreneurs, and salespeople who use them. For instance, machine studying that makes split-second choices about which digital advertisements to provide customers is constructed into platforms that deal with the complete course of of shopping for and putting advertisements. Netflix’s built-in machine studying has supplied prospects video suggestions for greater than a decade; its picks merely seem within the menu of choices viewers see once they go to the positioning. If the advice engine have been stand-alone, they would want to go to a devoted app and request strategies.
Makers of CRM techniques more and more construct machine-learning capabilities into their merchandise. At Salesforce, the Sales Cloud Einstein suite has a number of capabilities, together with an AI-based lead-scoring system that robotically ranks B2B buyer leads by the probability of buy. Vendors like Cogito, which sells AI that coaches name heart salespeople, additionally combine their functions with Salesforce’s CRM system.

Combining the 2 forms of intelligence and two forms of construction yields the 4 quadrants of our framework: stand-alone machine-learning apps, built-in machine-learning apps, stand-alone task-automation apps, and built-in task-automation apps.

Understanding which quadrant functions fall into can assist entrepreneurs plan and sequence the introduction of recent makes use of.
A Stepped Approach
We consider that entrepreneurs will in the end see the best worth by pursuing built-in machine-learning functions, although easy rule-based and task-automation techniques can improve extremely structured processes and provide affordable potential for business returns. Note, nonetheless, that these days job automation is more and more mixed with machine studying—to extract key information from messages, make more-complex choices, and personalize communications—a hybrid that straddles quadrants.
Stand-alone functions proceed to have their place the place integration is tough or not possible, although there are limits to their advantages. Therefore, we advise entrepreneurs to transfer over time towards integrating AI inside present advertising and marketing techniques relatively than proceed with stand-alone functions. And certainly, many corporations are heading in that general path; within the 2020 Deloitte survey, 74% of worldwide AI executives agreed that “AI might be built-in into all enterprise functions inside three years.”
Getting Started
For companies with restricted AI expertise, a great way to start is by constructing or shopping for easy rule-based functions. Many companies pursue a “crawl-walk-run” strategy, beginning with a stand-alone non-customer-facing task-automation app, reminiscent of one which guides human service brokers who interact with prospects.
Once corporations purchase fundamental AI abilities and an abundance of buyer and market information, they will begin transferring from job automation to machine studying. A superb instance of the latter is Stitch Fix’s clothing-selection AI, which helps its stylists curate provides for purchasers and relies on their self-reported model preferences, the objects they maintain and return, and their suggestions. These fashions grew to become much more efficient when the corporate started to ask prospects to select amongst Style Shuffle pictures, making a worthwhile supply of recent information.
New sources of information—reminiscent of inside transactions, exterior suppliers, and even potential acquisitions—are one thing entrepreneurs ought to search for always, since most AI functions, notably machine studying, require huge quantities of high-quality information. Consider the machine-learning-based pricing mannequin that the constitution jet agency XO used to improve its EBITDA by 5%: The key was to faucet exterior sources for information on the provision of personal jets and on components that have an effect on demand, reminiscent of main occasions, the macroeconomy, seasonal exercise, and the climate. The information XO makes use of is publicly accessible, however it’s a good suggestion to additionally search proprietary sources every time attainable, as a result of fashions utilizing public information might be copied by opponents.
As corporations develop into extra refined of their use of promoting AI, many totally automate sure forms of choices, taking people out of the loop solely. With repetitive, high-speed choices, reminiscent of these required for programmatic advert shopping for (the place digital advertisements are served up nearly instantaneously to customers), this strategy is crucial. In different domains AI could solely current suggestions to an individual confronted with a alternative—for instance, suggesting a film to a shopper or a technique to a advertising and marketing govt. Human decision-making is often reserved for essentially the most consequential questions, reminiscent of whether or not to proceed a marketing campaign or to approve an costly TV advert.
Firms ought to transfer to more-automated choices every time attainable. We consider that is the place the best returns from advertising and marketing AI might be discovered.
Challenges and Risks
Implementing even the best AI functions can current difficulties. Stand-alone task-automation AI, regardless of its decrease technical sophistication, can nonetheless be laborious to configure for particular workflows and requires corporations to purchase appropriate AI abilities. Bringing any sort of AI right into a workflow calls for cautious integration of human and machine duties in order that the AI augments individuals’s abilities and isn’t deployed in ways in which create issues. For occasion, whereas many organizations use rule-based chatbots to automate customer support, less-capable bots can irritate prospects. It could also be higher to have such bots help human brokers or advisers relatively than work together with prospects.
As corporations undertake more-sophisticated and built-in functions, different issues come up. Incorporating AI into third-party platforms, specifically, might be difficult. A living proof is obtainable by Procter & Gamble’s Olay Skin Advisor, which makes use of deep studying to analyze selfies that prospects have taken, assess their age and pores and skin sort, and suggest applicable merchandise. It is built-in into an e-commerce and loyalty platform,, and has improved conversion charges, bounce charges, and common basket sizes in some geographies. However, it has been more durable to combine it with retail shops and Amazon, third events that account for a excessive proportion of Olay’s gross sales. The Skin Advisor shouldn’t be accessible on Olay’s intensive retailer web site on Amazon, hampering the model’s capacity to ship a seamless, AI-assisted buyer expertise there.
Finally, corporations should maintain prospects’ pursuits prime of thoughts. The smarter and extra built-in AI functions are, the extra worries prospects could have about privateness, safety, and information possession. Customers could also be skittish about apps that seize and share location information with out their information or about good audio system that could be eavesdropping on them. In common, customers have proven a willingness (even eagerness) to swap some private information and privateness in change for the worth that progressive apps can present. Concerns about AI functions like Alexa appear to be dwarfed by appreciation of their advantages. Thus the important thing for entrepreneurs as they broaden the intelligence and attain of their AI is to be sure that its privateness and safety controls are clear, that prospects have some say over how their information is collected and used, and that they get honest worth from the agency in change. To assure these protections and preserve prospects’ belief, CMOs ought to set up ethics and privateness overview boards—with each advertising and marketing and authorized consultants—to vet AI initiatives, notably those who contain buyer information or algorithms that could be susceptible to bias, reminiscent of credit score scoring.
. . .
While advertising and marketing AI holds huge promise, we urge CMOs to be life like about its present capabilities. Despite the hype, AI can nonetheless accomplish solely slim duties, not run an whole advertising and marketing perform or course of. Nevertheless, it’s already providing substantial advantages to entrepreneurs—and actually is crucial in some advertising and marketing actions—and its capabilities are quickly rising. We consider that AI will in the end rework advertising and marketing, however it’s a journey that can take a long time. The advertising and marketing perform and the organizations that help it, IT specifically, will want to pay long-term consideration to constructing AI capabilities and addressing any potential dangers. We urge entrepreneurs to begin growing a technique at the moment to benefit from AI’s present performance and its probably future.

A model of this text appeared within the July–August 2021 concern of Harvard Business Review.

Recommended For You