Artificial intelligence (AI) – a subset of machine learning – is moving into the mainstream with the availability of increased computing power as well as the scalability and efficiency offered by the cloud.
AI has the potential to revolutionise the way we work and communicate, not least of all in the marketing function.
CMOs and CDOs are evaluating how and where to apply AI as part of their customer engagement capability in order to take advantage of the efficiencies, interactions and competitive edge it can offer. Since widescale AI deployment is still maturing and because the current capability of the tech is progressing quickly, wise marketers will take a progressive and iterative approach to adoption.
Machine learning is the foundation of AI
Machine learning involves setting a computer a problem and letting it process vast amounts of data (guided by some algorithms) until it creates its own rules – a model. There are different approaches to machine learning – perhaps the most common of these models are neural networks based on animal brains. AI uses underlying models to analyse input (data) and solve problems in a way that emulates human thinking.
Why the hype?
AI has evolved slowly for several decades with no dramatic breakthroughs. So why the recent hype? Because two major developments have now elevated the power and usefulness of AI to the next level: an explosion in the amount of data available and the scalable computing power available to crunch massive data sets.
Technologists have applied the latest advances in AI to enhance and augment every aspect of our lives from healthcare and lifestyle to transport and business.
How does AI help marketing?
Technologists have applied the latest advances in AI to enhance and augment every aspect of our lives from healthcare and lifestyle to transport and business. They are also developing advanced models advanced models that help with marketing applications and can be grouped as follows:
Vision – the ability to identify, label and categorise images. In marketing, the technology could be used to recognise and tag branded digital assets or to analyse customer-generated content.
Speech – the ability to convert text to voice and voice to text. Allowing users to speak to a machine rather than needing to type promotes more authentic and complete search requests. This improves accessibility, increases the richness of the search data collected and offers a more natural and human interaction. Always-on voice assistants such as Alexa and Google Home remove the screen from interactions between people and technology, creating a new relationship between the marketer and the household.
Natural language processing – the ability to extract meaning from human speech and sentences. This technology creates far more powerful searches, recognising the quirks and indirect links of natural language. For instance, the question ‘what is AI’ is similar to ‘tell me about AI’, and a search for holidays in Italy can bring back an article about hotels in Rome even if the words ‘holiday’ and ‘Italy’ don’t appear in the article. NLP also supports translation. Google’s Duplex is a great current demonstration of this AI capability.
Reasoning – a broad category that encompasses the ability to identify patterns to create predictive models and anticipate behaviour. For example, the technology can anticipate which customers will churn and determine how the churn can be prevented, as well as create recommendations tailored to the individual customer. This modelling can be performed in real time, making it powerful for applications such as identifying high-value customers as they browse a website or a beacon enabled store – marketers can respond in real-time rather than analysing their behaviour after the fact.
Deploying AI for Marketing
The business of productising AI is nascent with new vendors and solutions launching every day.
- Platform vendors such as Google, Microsoft, IBM and Amazon have used their access to big data and computational power to develop machine platforms and learning frameworks such as Tensorflow. Data scientists can customise these frameworks and tools such as H2O.ai and Data Iku to serve as the foundation of their production AI.
- Black box services such as Google Cloud platforms Vision, Speech and NLP APIs can be configured by data engineers to manage their companies’ inputs and outputs. Even more simply, companies can purchase marketing products which already incorporate machine learning components – e.g. Salesforce Einstein. These products do not directly require data engineering skills and can be operated by marketers.
The Ethics of using Machine Learning in Marketing
The vast power of AI means corporations and individual employees shoulder a heavy ethical burden in the consumption and processing of data. Just as they expect companies to show environmental and social responsibility, consumers are beginning to demand companies use their data ethically.
The new European directive covering the use of consumer data (GDPR) recognises this and prohibits machine learning modelling of customer data without consent .
The challenge for marketers is to look past the hype and assess the real value of AI products to the business. CMOs and CDOs should:
- Agree how the company can and should use AI in the short term to support business objectives and enhance the customer experience.
- Set a marketing technology roadmap to identify where AI can be applied or where foundation work is needed.
- Determine a deployment model – is the business is ready to recruit an inhouse team of data scientists and make use of bespoke models, or should it start its AI journey with software which already incorporates machine learning models?
- Prepare for the long-term – build a foundation now to enable the company to move quickly and competitively later as AI capabilities continue to evolve.
- Decide how to use AI to deliver the benefits, without compromising customers’ privacy and trust.
- AI today may be rudimentary and offer narrow applications, yet it can be powerful when augmenting existing capabilities. Marketers should prepare to adopt AI, take advantage of the ever increasing capabilities vendors are able to support and build new engagement methods as the technology matures.
 There are also provisions for modelling where contractual obligations make it necessary or the work is covered by separate laws, however for most companies consent is the most likely option.