What Exactly is Artificial Intelligence?
Put simply by Demis Hassabis, founder and CEO of Google-owned AI company DeepMind, artificial intelligence is the “science of making machines smart” (Ahmed, 2015). As broad of a definition as this may be, it is well fitting because AI is an umbrella term for a wide variety of manifestations.
Within the umbrella of AI includes subcategories, such as machine learning and deep learning, that produce real-world applications of AI, such as voice recognition, image recognition, virtual assistants, search suggestions. These are all forms of weak AI. While equally fascinating, general artificial intelligence, or even super-artificial intelligence (SAI), are far from being practical to implement in business today. For this reason, people typically mean "Weak AI" when they discuss AI, rather than General or Strong AI.
Below are some of the need-to-know definitions related to AI. You can find a more detailed list of AI terminology here.
Algorithms that analyze massive amounts of data to identify patterns and trends then take action on those insights automatically, improving itself in the process.
A subset of AI that uses programs to learn and improve upon itself and process large amounts of data. Machine learning is the aspect of AI that allows for it to learn without being explicitly coded to do so.
1) A type of machine learning in which human input and supervision are an integral part of the machine learning process on an ongoing basis. In supervised learning, there is a clear outcome to the machine’s data mining and its target function is to achieve this outcome, nothing more.
2) A class of machine learning algorithms that learn patterns from outcome data. Supervised learning algorithms make predictions based on a set of examples.
A category of algorithms that are trained using a dataset that has not been labeled. Unsupervised learning algorithms look for patterns, underlying structures, and hidden relationships within a training dataset. The algorithm then creates a function to model these relationships between inputs and outputs in an effort to achieve accurate predictions on a previously unseen input.
The general term for to machine learning using layered (or deep) algorithms to learn patterns in data. It is most often used for supervised learning problems.
Natural Language Processing (NLP)
A branch of artificial intelligence that helps computers understand, interpret, and manipulate human language. This field of study focuses on helping machines to better understand human language in order to improve human-computer interfaces with use cases like moderation, information extraction, summarization, and more.
The ability for computers to “see” imagery through mathematical representations of three-dimensional shape and appearance. Computer vision lets a computer comprehend the meaning and context of an image in a similar way as human vision, allowing for sentiment analysis, facial recognition, and much more.
Adoption of AI
AI has a long history of being sold as a panacea for businesses. Ever since its inception around the late 1950's, AI has gone through periods of being hyped and oversold, while failing to meet high expectations due to a lack of computational power, technological advancements, data availability, and other factors. When the technology failed to meet expectations, an inevitable "AI Winter" would ensue, characterized by a general lack of trust and funding going towards AI projects.
This pattern can still be seen today. Many of companies today are sold AI solutions that don't even use AI technology. Some companies get caught in the hype of AI and rush to use it throughout all aspects of their business, even if it's not the best technology for a given task. Other companies may start slow when implementing AI, only for their initiatives to fade away due to lack of proper resources.
However, one thing is clear about modern artificial intelligence: it is radically transforming entire industries from tech and retail to healthcare and manufacturing. The companies that properly implementing AI are already seeing massive gains.
A myriad of factors has contributed to the rise in interest and feasibility in recent years. Some of these include:
- Increased computing capabilities to process AI algorithms at scale cheaper than ever before,
- Big Data and the data management advancements that came with it,
- Heightened interest and funding,
- Increasingly large pool of highly talented professionals eager to advance the industry.
All of the factors listed above also contribute to the growing divide between AI leaders and laggards. Noteworthy AI leaders such as Amazon and Google tend to attract a large portion of the scare pool of AI talent. This makes it increasingly difficult for small and medium size companies to implement robust AI systems.
While large-scale AI systems may be out of reach for most companies, a growing number of small-scale AI solutions are becoming available that require much fewer resources. For example, no-code solutions such as Google's AutoML provide "a suite of machine learning products that enables developers with limited machine learning expertise to train high-quality models specific to their business needs" using "Google’s state-of-the-art transfer learning and neural architecture search technology".
No-code and low-code AI solutions give companies of all sizes the ability to implement powerful AI into their websites, products, and services.
Marketing implementation of AI varies widely. AI can be implemented within an organization at a basic level for retargeting ad campaigns, and it can be implemented at a complex level with robust customer relationship management systems (CRMs) and omnichannel integration.
As such, understanding the current state of artificial intelligence in marketing can be tricky. Part of my undergrad honors thesis research led to me creating a scale to understand marketing implementation of AI better.
A key factor for differentiating the various degrees of marketing implementation of AI is what I termed “involvement.” The level of involvement that an application of AI exhibits reveals key details of how it is integrated into the company.
The amount of resources required to implement the AI. Low involvement entails low resources required, whereas high involvement entails a high amount of financial requirements. Similarly, low involvement solutions require less data or make use of collective pools of data, but high involvement solutions typically require large amounts of proprietary company data.
The labor requirements for implementing the AI. Low involvement AI solutions can be used by entry-level employees, but high involvement AI solutions require experts to implement and manage the systems. Hiring top-quality AI talent is both difficult and expensive at the moment, creating a barrier to entry for smaller companies.
The degree to which the AI plays a critical role in the day-to-day and high-level business operations. Low involvement AI offers a limited competitive advantage and is not a determinant factor in the brand’s value proposition. Contrastingly, high involvement AI is deeply rooted in the brand’s core product or service offerings.
Low Involvement AI
Low involvement AI consists of third-party solutions that do not require large amounts of resources (money, time, expertise, etc.) to implement and manage. Low involvement AI provides the benefit of lower barriers to entry at the cost of less powerful technology.
They can be subscription software available for a small monthly fee, free applications, or other readily accessible solutions. An essential trait of low involvement AI is that it does not act as a core value proposition or competitive advantage for the company. These are small-scale, single-use applications that often subtly use AI in their backend programming.
Examples of Low Involvement AI Implementation Today:
- Search functionalities from the user’s perspective. This refers to an individual using Google, or some form of internal website searches as a part of day-to-day tasks, as opposed to a brand developing custom search functionalities with recommendation systems, personalization, etc.
- Creative applications used by individuals. Again, referring to Adobe Creative Cloud programs like Photoshop and After Effects, mobile apps for editing pictures, etc. as opposed to the development of such applications (which have complex AI in their backend)
- Built-in functionalities of other products or services (Photo tagging on Facebook, Siri in iPhones)
Low-Medium Involvement AI
The low-medium involvement range of AI are automated systems made to handle lower-level, task-oriented tasks. Use cases of low-medium involvement AI reduce human participation in labor and time intensive tasks but require manual oversight and input from the user to operate.
Examples of Low-Medium Involvement AI Implementation Today:
- Entry-level usage of programmatic advertising
- Basic retargeting ads
- Newsletter automation and simple personalization (Mailchimp)
- Low-level marketing automation (Kit for Shopify)
Medium Involvement AI
Medium involvement AI applications make use of moderately robust technological capabilities in specific areas of the business. Rather than playing an integral role in the company’s high-level strategy, these forms of AI offer more narrow-focused applications characterized by interactivity from the customer with the brand.
Examples of Medium Involvement AI Implementation Today:
- Customized chatbots (for customer service, eCommerce, or personal assistant use cases)
- Custom voice assistant applications (through Amazon Alexa, Google Home, etc.
- AI-generated music for localizing marketing campaigns
- Experiential marketing (one-off marketing campaigns or tactics that stage interactive experiences using AI, such as Walt Disney Co. using language processing to trigger an audio soundtrack as a parent reads a story to their child (Chow, 2017))
- Real-time personalized UI/UX (such as using Wordsmith to contextualize the content and layout of a brand’s homepage to match the local market characteristics of a specific user)
- Creative augmentation, creative intelligence (Adobe Sensei)
Medium-High Involvement AI
Medium-high involvement AI branches into systems that play an integral role in the firm’s marketing efforts. This category of AI implementation is characterized by usage across multiple areas of marketing efforts as opposed to single-use applications (e.g., systems to handle the entire customer relationship management process versus only handling email personalization). For large firms, this stage may be characterized by regional/location specific usage of AI, rather than company-wide usage.
Examples of Medium-High Involvement AI Implementation Today:
- Customer relationship management (Salesforce Einstein that utilizes predictive analysis, intelligent recommendations, image recognition, sentiment analysis and task automation)
- Advanced programmatic advertising and digital marketing (Albert AI that uses in-depth behavioral customer segmentation, highly targeted automated media buying, high-level customer personalization, profound insights into consumer behavior, and cross-channel execution of marketing efforts)
- Omnichannel integrated systems to distribute creative assets (Adobe Experience Cloud)
High Involvement AI
High involvement AI implementation encompasses the most advanced, integral technology available to businesses. These may be highly customized, in-house AI systems explicitly built for the company used throughout crucial decisions and operations.
Examples of High Involvement AI Implementation Today:
- High involvement AI implies the technology is part of the core value proposition of the company and has a direct impact on its performance. These are robust systems that are at the cutting edge of the technology’s capabilities and highly intertwined with how the company operates.
- Custom AI systems built by the company (such as the creators of AI applications themselves like Google or Facebook)
- Company-wide, cross-department usage of third party AI systems for critical operations (full-scale implementation of applications such as IBM Watson as a core feature of the company’s service offerings or operating procedures)
- Pervasive AI (cognitive environments that act as integrated brand experiences using smart technology in different settings (IBM, n.d.), omnichannel CRM’s that can track and segment customers using sentiment analysis online, facial recognition in a retail environment, etc.)
The 5 Core Benefits of AI in Marketing
AI allows for brands to connect with and delight their customers on an individualized level at scale. By being able to track and analyze new sources of data entirely, marketers can achieve comprehensive views of their customers to fuel their marketing efforts.
This allows brands to deliver highly personalized, omnichannel marketing efforts that are much more likely to resonate. Hyper-Personalization using AI facilitates a level of intimacy between brands and consumers that has never before been possible, leading the way for enjoyable and individually-relevant experiences that align closely with Generation Z and Millennial consumers (Vision Critical, 2016).
The combination of automating time-intensive tasks, comprehensive understandings of target audiences, and deep insights into marketing performance means marketers can see lowered costs in many of their activities.
The ability to get their message in front of the right customers in the right way at scale means marketers no longer need to spend money marketing to customers that don’t contribute to their bottom line.
Less employee time spent on repetitive or menial tasks allows for greater focus on value-adding activities as well. This means that customers may be able to enjoy higher value from the brands they interact and shop with at no extra cost to them, as well as the potential to be reached by new brands that may be a good fit for them.
AI offers brands improved visibility of their customers. This means that customer segmentation can be made more effective through deep insights into consumer behavior at macro and micro levels, and campaigns can be tracked and analyzed faster and more intelligently to inform marketers’ strategic decisions.
Companies can tap into data from sources that were previously unavailable to them to achieve a highly comprehensive understanding of their customers. For consumers, this entails increased relevancy of the marketing messages they are exposed to and highly contextualized and personalized experiences with brands.
Profound insights can set the foundation for brands facilitating long-lasting, meaningful relationships with their customers by reducing their frustrations and allowing for 1:1 experiences to take place.
AI enables brands to automate many customer-facing activities that previously required human labor to execute, such as customer service representatives or sales associates.
Brands can see lowered costs in hiring, training, and managing employees, while simultaneously enabling customers to access companies in a myriad of self-service channels (such as chatbots, voice-powered applications, personal assistants, and more).
Customers having the ability to access a brand at their convenience means interacting with brands will increasingly become a hassle-free experience. As AI can fill customer-facing roles with incredible efficiency and efficacy, customers and employees can have access to companies at their convenience.
Marketers of the future will have an arsenal of tools available to create experiences, communicate meaningful stories, generate value for their customers at scale. Being able to reach customers in highly personalized ways means customers can be entertained, enlightened, and empowered through 1:1 marketing.
Highly interactive experiences using AI (such as biometrics, facial recognition, emotion analysis, etc.) allow for more meaningful interactions between brands and consumers. With brands being able to individualize marketing efforts at scale, customers can enjoy content tailored to their interests, aspirations, and needs.