Three top trends for 2024: generative AI, hyperpersonalisation and Augmented Intelligence

The ceaseless evolution of data science continues to unleash enormous capabilities for understanding customer behaviours and optimising their experience. In this blog, dunnhumby's Head of Science, Chris Brooks, picks his three top trends in this for 2024 and discusses the impact they will have over this year and beyond. 

1. Generative AI

While there have been major advances in generative AI over the last few years, 2023 was the year when it really entered the public consciousness.  ChatGPT was the focal point, with everyone able to try it out and appreciate its impressive language, coding and conversational skills. 

And while generative AI could easily have been seen as over-hyped or a bit of a gimmick, it is starting to find practical applications with clear value.  After all, who wants to summarise a long slide deck into an executive summary when you can just ask your computer to do it for you? Meanwhile, simple code can be produced automatically, freeing the programmer from getting bogged down in too much implementation detail. 

As we enter 2024, the use of generative AI is only going to grow, and this will necessitate an increased focus on AI ethics, responsibility and safety.  Large language models truly are large, with their billions of parameters, and it’s not fully understood (yet) exactly how they work.  Guardrails will help, as they can catch incorrect or offensive responses before they are returned to the user. 

The introduction of smaller, more efficient language models such as Llama 2 and Mistral show that bigger is not always better.  In 2024, there will be a focus on reducing their size further without losing power – while simultaneously reducing the need for computing power and graphical processing units (GPUs) to run the models and opening up the possibility of using them directly on handheld devices.

2. Hyperpersonalisation 

Hyperpersonalisation in marketing and customer engagement is a strategy that aims to provide highly individualised and relevant experiences to customers at every point of interaction. 

Traditional personalisation often involves dividing customers into segmented audiences and delivering tailored content or recommendations to each segment. Hyperpersonalisation goes beyond that and relies on knowing every customer at every interaction. 

It relies on collecting data about each individual’s past interactions, preferences, behaviour, demographics, and more.  It is not limited to one channel or platform – it spans all customer touchpoints, whether online or offline.  This data is used to gain insights into each customer's unique needs and interests. 

Hyperpersonalisation is not a new concept – at dunnhumby we pioneered this space and already deliver more than 60 billion personalised offers each year.  However, with continued increases in computing power and the explosion of data sources, we see hyperpersonalisation as a major trend for 2024 that will permeate more sectors. 

Advances in computing hardware are opening up more sophisticated algorithms, such as graph neural networks which could help in areas where we may have less information on customers, for example low frequency purchases in general merchandise or recommending categories where a customer hasn’t bought before. 

Equally, as consumer lifestyles, habits, and spending patterns evolve over time, so too will the approaches for achieving hyperpersonalisation. In essence, strategies and data assets that are successful today might not yield the same increments tomorrow - so the need to work on mining and refining the context in which customer interactions take place will continue to be important.   

With the great power of individual level data sources comes great responsibility; not just to comply with growing privacy legislation around the world, but also to do the right thing for the customer by ensuring their data is only used in a safe and responsible way, to deliver the most relevant marketing and engagement at every touchpoint in 2024, and beyond. 

3. Augmented intelligence

Last year we identified Augmented Intelligence as a key trend for 2023, and we only see this trend increasing in 2024. 

Augmented Intelligence builds on Artificial Intelligence (AI) by ensuring there is always a human “in the loop” – allowing the sophistication of AI to efficiently recommend the optimal way forward, but ultimately leaving it up to a human to take the responsibility for the final decision.   

It also allows humans to bring their own experience and knowledge, which may not be straightforward to codify or capture in an AI model, into the decision-making process. 

With rapid advances in AI, there is always the danger that algorithms may not be aware of gaps in their training data, or situations they have never seen before, leading to bias.  With training data now sometimes getting too large to manually curate, there is also the danger of AI learning undesirable traits, or even producing offensive responses.  Augmented Intelligence mitigates this risk, by ensuring a human decision ratifies any output before it is used. 

Augmented Intelligence helps to ensure that decisions are fair, that the decision-maker is accountable for the action, and that there is clear transparency.  It ensures we understand how data is being used and what the rationale is behind decisions, to avoid anyone being disadvantaged.   

Examples of Augmented Intelligence include sense-checking optimised price recommendations before they are implemented, or visualising how changes to the range in a category will look on the shelf before they are confirmed. 

When used correctly, Augmented Intelligence helps business leaders to maximise the benefits of AI, while avoiding some of the potential negative consequences, allowing them to make the most of rapid and recent advances while minimising risk.