Read our new manifesto

Explore our new sections and topics

14 - 18 June

Are influencers the new source of creativity?

Jack Ibbetson

pr manager, EMEA

Qaiser Bachani

global digital marketing and Europe consumer experience lead

Lynn Lester

managing director of events

Katie Hunter

social and influencer lead

Lisa Targett

global head of sales

Expectations vs reality: questions to ask to get the true value of AI

The promise of artificial intelligence (AI) in marketing is huge. Once a product of the future, AI is now commonplace and is only becoming more relevant to our marketing needs. With its ability to personalise the online user experience, the $1.2 trillion industry is heralded as the silver bullet to the many challenges that marketers currently face. Yet, when it comes to implementing AI into the marketing mix, brands and agencies must look beyond the promises and assess what their AI provider is able to bring to the table.

Although marketers don’t need to know the technical details behind what drives AI, it is important to have a good understanding of how it works and what it can do. Not all AI systems are created equal, so having a grasp of where your AI platform derives its “intelligence” from is necessary so that you know its limitations and capabilities. From there, you’ll be able to discover if your AI provider is delivering the value it promises and how to maximise its potential.

Understanding AI systems and models

Essentially, there are two types of AI systems – rule-based and model-based. The rule-based system is a simpler version and has a static level of intelligence since it follows a fixed set of instructions and doesn’t have the capability to learn on its own. A model-based system, on the other hand, is much more sophisticated and allows for machine learning. It picks up patterns from new data that is accumulated, constantly fine-tuning to achieve the best outcome. Since today’s marketing is all about creating a personalised experience, it makes sense to follow a model-based AI system that can dive into the granular, individual needs of each customer.

Regardless of which system is used, AI models need to be updated frequently. Although AI has the ability to get smarter over time, it needs to be constantly improved and updated. Otherwise, it can fall into a concept drift, where it fails to recognise and analyse the correct data, creating less accurate predictions and learnings over time. Anything less than constant monitoring and optimisation will lead to a reduction of performance.

Questions to consider:

Do you know which AI system you are currently using?

What are you doing to optimise your current AI system?

How often do you implement new developments in AI?

Supporting AI with the right data

Since data is the crux of AI, having quality data is paramount to a well-performing AI platform. Make sure you have a class-leading data management platform (DMP) that provides clean data to support your AI capabilities. This is where the role of a data scientist becomes critical because while computers are great at matching patterns, the human element will always be necessary to set clear benchmarks and parameters for the platform to operate within.

It’s also very important to find out what data your AI model was built upon. Did the providers use full customer history or a subset of customer data? Subset data tends to be a more limited approach designed to work for a niche market. While this may work for your business, a full customer history is generally preferable as it allows for more flexibility.

Questions to consider:

Where do you source your data from?

How is your data and AI connected? How often are they updated?

Segmentation vs personalisation

A key benefit of AI is that it allows marketers to target the right customers through segmentation. However, as brands move towards more accurate targeting, it’s worth noting that segmentation isn’t the same as personalisation. Segmentation clusters a group of people with similar characteristics, but not everyone in that same segment may want or need the same things.

Personalisation, on the other hand, tailors to each individual’s needs based on their online footprint. For example, in personalised AI targeting, a customer’s purchase history can be used to determine their items of interest and purchase behaviour. From there, marketers can introduce similar products to them based on their past purchases.

Questions to consider:

As a marketer, do you know exactly who your target audience is?

How can you use personalised AI targeting to better reach your customers?

Human intelligence

Lastly, while technology is important, we mustn’t forget that the humans in the driving seat are really what defines effectiveness. The dawning of the AI era is already behind us, and so your AI provider’s data team ought to be bringing significant experience, ideally from a data science discipline, to ensure they are looking holistically at solving your business challenge rather than focussing solely on marketing metrics.

Questions to consider:

Do you have a budget set aside to hire good data scientists?

How many of your product teamwork in data science?

If the idea of improving the entire AI system in your organisation seems overwhelming, try breaking down the steps to make this more achievable. To start, ask the relevant head of departments what their needs are, then lay them all down for your AI provider to see how AI can be leveraged to meet these goals.

In order to maximise the potential of AI, marketers need to define success from the get-go. A great AI tool can only deliver its true value if you know what you’re looking for. With a sound understanding of AI and a clear goal in mind, you have the tools to truly understand and measure the value that your AI provider is delivering.

Pete O'Mara-Kane is general manager of International at LoopMe.