AI is hitting the headlines frequently right now and there is much debate about how it will shape the future. However, the truth is that we are already surrounded by AI, with many of us enjoying its benefits (knowingly or otherwise) across a plethora of AI-enabled devices. It’s here to stay, its applications will grow and there will undoubtedly be disruption as a result.
With seismic change coming, it’s vital that loyalty leaders understand how AI is impacting CRM and loyalty today together with its potential in this sphere.
In this short paper, we look at what AI is and how its technologies can be applied before considering the changing customer expectations and barriers that might prevent companies from accelerating their loyalty offers. Finally we examine the practical advantages AI will bring to loyalty management solutions and how these can help to overcome the barriers that businesses and brands encounter.
A basic understanding of Artificial Intelligence (AI)
Wave 4 -
Artificial General Intelligence
Future - Refers to a hypothetical form of artificial intelligence that possesses the ability to understand, learn and apply knowledge across a wide range of tasks and domains in a manner similar to human intelligence. AGI would be capable of performing any intellectual task that a human being can, including reasoning, problem-solving, creativity and social interaction.
Demystifying AI and leveraging it in loyalty
The AI digital age is here and is permeating our daily lives at speed! Kids are learning how to use Chat GPT to better understand subjects at school, from asking how Shakespeare addresses themes such as poverty or race to understanding how electrons work and even using it to generate computer program code.
We can see that customers are also looking at this technology and starting to seek ways to capitalise on the opportunity in front of us. In fact 56% of them have already started, which also means that just under half are already losing ground.
The four waves of AI
Before considering how AI can be applied to the future of loyalty, we take a look at the AI patterns developed
in four key waves:
A more detailed understanding of the first three waves and the technology patterns/models and how they apply are detailed in Appendices 1, 2 and 3 of this document. These models can be applied to your CRM/loyalty solution now to generate information and/or activity as appropriate.
Leveraging AI on your CRM / Loyalty Platform requires connection of data sources with the relevant AI model and configuration of defined inputs and outputs from the model.
For example:
• Pre-built items like “Next Best Action”
• Analytics Dashboards for Insights
• Automation flows for process automation / Job To Be Done (JTBD)
Wave 3 -
Autonomous & Agents
Example:
Acts independently, making decisions and taking actions without direct human intervention based on what is has learned from existing data.
Chatbots, autonomous vehicles
Wave 2 -
Generative
Example:
Assists - Creates news things based on what it has learned from existing data.
Chat GPT.
Wave 1 -
Predictive
Example:
Informs - Makes predictions based on what it has seen on existing data.
NBA (Next best action) engines.
Want to know more?
Get in touch.
Acquiring customer data and sustaining customer loyalty for long periods is a never-ending challenge that every business faces. While most consumers are members of more than 12 customer loyalty programmes, they are active in less than 50% of those programmes. What makes some programmes more successful than others and how do you make your customer loyalty programme one of those that people use, share and talk about?
At Collinson we work tirelessly with customers to unlock the magic within their business, to design build and deliver the best loyalty programmes that set them apart from their competition and drive desired change – more mindshare, wallet share, advocacy, and Loyalty.
To learn what we’ve done for businesses like yours and how we can leverage Salesforce loyalty management technology across Customer 360 to help you achieve your customer vision and bring your loyalty strategy to life, please get in touch.
Collinson
About Us
Stephen Gilbert
Vice President Salesforce Loyalty
Contact me
With data quality being one of the most important factors to successfully implement AI technology, loyalty management solutions are uniquely placed to help CRM solutions’ data quality through several mechanisms:
How loyalty management benefits the AI story
AI programs can improve loyalty software in several ways, enhancing effectiveness and operational performance at scale providing additional value to both businesses and customers:
How AI technologies benefit loyalty management
Conclusion
In summary, we can conclude that:
1. AI technologies are not difficult to understand and you do not need to get deeply into the science to understand how they work. As a result the application of AI to loyalty management should not be seen as technically difficult task:
a) A range of AI models for CRM can be targeted to analyse CRM and Loyalty data.
b) A relevant and simple user interface like Co-Pilot is required to push inputs to and generate outputs and actions from the AI models.
2. AI Predictive, Generative and Autonomous models work best when applied to a data store that is grounded on CRM data. This store should be up to date, consistent, complete and trusted by the business. Not having such a source is a significant barrier to potential AI adoption. If your CRM data is not grounded then a loyalty programme is a good way of starting this development as the loyalty data being either Zero or First party will be the most trusted.
3. Leveraging AI and loyalty can deliver significant business benefits including, greater customer satisfaction, higher degree of personalisation, improved programme ROI and enhanced user productivity in Marketing and Service departments.
Loyalty & AI Mutually Beneficial Technologies
Customer expectations continue to rise as businesses face increasing costs, the need to drive ever greater efficiencies and navigate privacy hurdles.
Customers demand fast, consistent, and personalised interactions every time they interact with a company. They’re seeking brands that offer exceptional products and services and are loyal to those that deliver. With customer loyalty at stake, this is a pivotal moment for businesses to thoughtfully use generative AI to supercharge the seamless experiences their customers have come to expect.
Growing customer expectations
Compounding factors like inflation and technological advancement are leading people to re-think what’s important, including where and how they spend their money.
Nevertheless, customers are not single-minded in their quest for a good deal, leaving room for brands to compete on more than price. Product quality remains essential, as does customer experience.
80% of customers say the experience a company provides is as important as its products and services.
Customers are re-visiting their priorities
The changing face of CRM and Loyalty
The top concern for business readiness for AI is one of low data readiness, this concern is even greater than privacy and regulatory concerns:
A key contributor to the data barrier is the consistency of data across channels, devices and departments. However, consistency is not the norm, and many customers find themselves repeating information to different representatives — a sign of siloed information or worse, having a customer data mis-match.
Ultimately, poorly integrated technology and processes leave 55% of customers feeling like they generally engage with separate departments rather than holistically with one unified company.
Companies struggle to provide a connected customer experience
Barriers to AI Adoption
Our research shows that growing customer expectations are what brands - and specifically loyalty operators - see as having the greatest impact on programmes over the next three years - both positively and negatively.
Brands are somewhat fearful of growing customer expectations
Organisations’ abilities to rapidly understand and offer solutions that are competitive and engaging are going to be influenced, in no small part, by their willingness and capacity to adopt AI technologies.
Personalisation gives favourite brands a leg up
Underpinning growing customer expectation is the expectation that companies will adapt experiences to match their changing needs and preferences. This means personalisation is a fundamental tenet of modern customer engagement, with almost two-thirds of customers saying they expect personalised experiences.
Today though, most customers feel companies treat them as a number rather than as an individual. This is particularly true of consumers whose relatively small transactions don’t prompt the tailored attention of large and slow-moving business deals.
Customers know that companies depend on data to meet their standards for personalised engagement but expect something in return when they hand it over. The more data customers provide, the better the experience they expect.
Data and technology up the ante for customer engagement
Customers’ favourite brands have an edge against the competition when it comes to matching changing needs, but even they have room for improvement.
Technological breakthroughs like generative AI can help businesses scale support and personalisation and are further raising customers’ standards: 81% of customers expect faster service as technology advances, and 73% expect better personalisation.
This issue is also equally true for AI systems, where the consistency and accuracy of the data against a unified customer or member profile will drive better AI predictive recommendations or generative actions.
So, in our attempt to demystify where we are today with loyalty and AI, we need to focus on how we get a consolidated and timely customer data solution that can provide a 360 degree view of the customer or member in a timely fashion. Ultimately this is what is critical to the first stages of implementation of effective and trustworthy AI.
Zero party / first party data: Loyalty solutions via the inbuilt Value Exchange mechanism by default gather the most accurate data, whether this is via the member association with transactional data (1st party) or explicit requests for information (0 party). Additionally this data is generally given across customer service lines, allowing companies to quickly “stitch” disparate customer profiles across their technology estate.
Increased customer engagement: Loyalty programmes incentivise customers to engage more frequently via zero and first party models with a business or brand for example:
• Completion of member profile including likes and affinities
• Commentary on social platforms and referrals
• Feedback on offers and products
• Engagement with loyalty programme partners
• Behavioural actions such as choosing to opt into promotions
This leads to a higher volume of data being generated from customer interactions. More data means more opportunities for AI algorithms to learn and improve their accuracy and performance.
Richer data profiles: Loyalty programmes often collect detailed information about customers, including their purchase history, preferences, demographics and behaviour. This is again extended and enriched by loyalty partner relationships. By leveraging this rich data, AI algorithms can build more comprehensive customer profiles, enabling more personalised and targeted recommendations, promotions and experiences.
Feedback mechanisms: Loyalty programmes provide channels for customers to provide feedback, reviews and ratings on products and services. This feedback data can be valuable for training AI models to understand customer preferences, sentiment and satisfaction levels, leading to better-informed decision-making and product improvements.
Predictive analytics: Loyalty programmes generate data that can be used for predictive analytics, such as forecasting customer churn, predicting future purchases and identifying trends and patterns in customer behaviour. AI algorithms can analyse this data to anticipate customer needs and preferences, allowing businesses to proactively address customer concerns and improve retention rates.
Segmentation and targeting: Loyalty programme data can be used to segment customers into distinct groups based on their purchasing behaviour, preferences and other characteristics. AI algorithms can analyse these segments to identify patterns and trends within each group and tailor marketing strategies and promotions accordingly, leading to more effective targeting and higher conversion rates
Personalised rewards and offers: AI algorithms can analyse customer data, including purchase history, preferences and behaviour, to personalise rewards and offers. By understanding individual customer needs and preferences, loyalty software powered by AI can tailor rewards and promotions to each customer, increasing their relevance and effectiveness.
Predictive analytics: AI algorithms can analyse historical data from loyalty programmes to identify patterns and trends in customer behaviour. By leveraging predictive analytics, loyalty software can anticipate future customer needs, predict churn, identify opportunities and perform Partner Whitespace analysis for targeted engagement and retention strategies.
Dynamic point valuation: AI algorithms can optimise the valuation of loyalty points based on factors such as customer behaviour, market conditions and inventory levels. By dynamically adjusting point values, loyalty software can incentivise desired behaviours, maximise customer engagement and optimise programme profitability.
Segmentation and targeting: AI algorithms can segment customers into distinct groups based on their preferences, demographics and behaviour. Loyalty software can then use these segments to target customers with personalised communications, promotions and rewards that are tailored to their specific needs and interests.
Fraud detection and prevention: AI-powered fraud detection algorithms can analyse transaction data from loyalty programmes to identify suspicious patterns and anomalies indicative of fraudulent activity, such as point theft or account takeover. By detecting and preventing fraud, loyalty software can protect the integrity of the programme and maintain customer trust.
Customer service and Loyalty operations optimisation: AI-powered chatbots and virtual assistants can provide personalised support and assistance to loyalty programme members, helping them with account enquiries, reward redemptions and other issues in real-time. By automating routine customer service tasks, loyalty software can improve efficiency and enhance the overall customer experience. Co-pilots can also help marketeers and service agents with their daily tasks.
Feedback analysis: AI algorithms can analyse feedback from loyalty programme members, including reviews, ratings, and survey responses, to identify trends and insights that can inform programme improvements and enhancements. By listening to customer feedback and acting on it, loyalty software can continuously evolve and adapt to meet the changing needs of customers.
Predictive AI models are a class of artificial intelligence algorithms designed to make predictions or forecasts based on input data. These models learn patterns and relationships within the data to predict future outcomes or infer missing information. Predictive AI models are widely used in various fields for tasks such as forecasting, classification, regression and anomaly detection. Here are some common types of predictive AI models:
Appendices
Appendix 1: Predictive AI models
Regression models: Regression models are used to predict continuous numerical values based on input variables. Examples include linear regression, polynomial regression and logistic regression (for binary classification). Regression models learn the relationship between input features and the target variable and use this knowledge to make predictions on new data.
Classification models: Classification models are used to predict categorical outcomes or assign labels to input data. Examples include logistic regression (for binary classification), decision trees, random forests, support vector machines (SVM) and neural networks. Classification models learn decision boundaries between different classes in the input data and use them to classify new instances.
Time series forecasting models: Time series forecasting models are used to predict future values of a time-dependent variable based on its past values. Examples include autoregressive integrated moving average (ARIMA) models, exponential smoothing methods, recurrent neural networks (RNNs) and long short-term memory (LSTM) networks. These models analyse the temporal patterns and trends in the time series data to make predictions about future values.
Anomaly detection models: Anomaly detection models are used to identify outliers or abnormal instances in a dataset. Examples include statistical methods such as z-score, isolation forests, one-class SVM and autoencoder neural networks. Anomaly detection models learn the normal behaviour of the data and flag instances that deviate significantly from this norm as anomalies.
Recommendation systems: Recommendation systems are used to predict user preferences and recommend items or content that are likely to be of interest to users. Examples include collaborative filtering, content-based filtering and hybrid recommendation systems. Recommendation systems analyse user behaviour and item characteristics to make personalised recommendations.
Ensemble models: Ensemble models combine multiple base models to improve prediction accuracy and robustness. Examples include bagging (e.g. random forests), boosting (e.g. AdaBoost, gradient boosting machines) and stacking. Ensemble models leverage the diversity of multiple models to reduce overfitting and improve generalisation performance.
Predictive AI models play a crucial role in decision-making processes across various industries, including finance, healthcare, e-commerce, marketing and manufacturing. By leveraging historical data and learning patterns from past observations, these models enable organisations to make informed predictions, optimise processes and drive better outcomes.
Generative AI models are a class of artificial intelligence algorithms designed to generate new data that resembles a given dataset. These models learn the underlying patterns and structures of the data and use that knowledge to generate new samples that are similar to the original data. There are several types of generative AI models, each with its own approach to generating data:
Generative adversarial networks (GANs): GANs consist of two neural networks, a generator and a discriminator, that are trained simultaneously in a competitive manner. The generator generates new data samples, while the discriminator distinguishes between real and fake samples. Through this adversarial training process, GANs learn to generate highly realistic and diverse data samples, such as images, text and audio.
Variational autoencoders (VAEs): VAEs are probabilistic models that learn to encode and decode data in a lower-dimensional latent space. The encoder network maps input data to a latent space, and the decoder network reconstructs the input data from samples drawn from the latent space. By sampling from the latent space, VAEs can generate new data samples that resemble the original data distribution while allowing for continuous interpolation between samples.
Auto-regressive models: Auto-regressive models and transformer-based models like GPT (Generative Pre-trained Transformer) generate data sequentially by predicting the next element in a sequence given the previous elements. These models are commonly used for generating text, where each word or token is generated based on the context provided by the preceding words.
PixelCNN and PixelRNN: PixelCNN and PixelRNN are generative models that generate images pixel by pixel. These models model the conditional probability distribution of each pixel given the previous pixels in the image. By sampling from this distribution, PixelCNN and PixelRNN can generate highly detailed and realistic images.
Flow-based models: Flow-based models learn a bijective mapping between input and output spaces, allowing for efficient and reversible generation of data samples. These models are particularly well-suited for generating high-dimensional data, such as images and audio and have been used for tasks such as image generation, style transfer and image inpainting.
Generative AI models have a wide range of applications across various domains, including image generation, text generation, music composition, drug discovery and more. By learning the underlying patterns and structures of complex data distributions, these models can generate new and diverse samples that exhibit characteristics similar to the original data, unlocking new opportunities for creativity, innovation and discovery.
Appendix 2: Generative AI models
Autonomous and agent-based AI models are two approaches to developing intelligent systems that can interact with their environment and make decisions without human intervention. While they share similarities, they also have distinct characteristics:
Autonomous AI models:
While autonomous AI models focus on individual systems operating independently, agent-based AI models emphasise the interactions and dynamics between multiple autonomous entities. Both approaches have applications in various domains, including robotics, simulation, optimisation, and decision support systems.
In summary, autonomous AI models operate independently to achieve specific goals, while agent-based AI models consist of autonomous entities (agents) interacting with each other and their environment to accomplish tasks or solve problems. Both approaches contribute to the development of intelligent systems capable of autonomous decision-making and behaviour.
Appendix 3: Autonomous and agent based models
Autonomy: Autonomous AI models operate independently, making decisions and taking actions based on their internal programming and environmental input.
Goal-directed Behaviour: Autonomous AI models are typically designed to achieve specific goals or objectives. They continuously assess their environment and adjust their behaviour to achieve these goals.
Examples: Autonomous vehicles, autonomous drones, and autonomous robots are examples of autonomous AI models. These systems navigate and interact with their surroundings, making decisions in real-time to accomplish tasks such as driving, flying or performing physical tasks.
Agent-based AI models:
Agents: Agent-based AI models consist of autonomous entities called agents, each of which perceives its environment through sensors, processes information and takes actions to achieve its goals.
Interactions: Agent-based AI models often involve interactions between multiple agents, which may cooperate, compete or communicate with each other to accomplish tasks or solve problems.
Emergent behaviour: Agent-based AI models can exhibit emergent behaviour, where complex patterns and phenomena emerge from the interactions between individual agents.
Examples: Multi-agent systems used in simulations, game environments, and economic models are examples of agent-based AI models. These systems simulate the behaviour of multiple agents with different goals and decision-making capabilities interacting within a shared environment.
Sources:
References
(1) Salesforce Customer Research & Insights 2023
(2a+b) Salesforce State of the Connected Customer Report 2023
(3a+b) Collinson research 2022, UAE, KSA, UK
(4a+b+c) Salesforce State of the Connected Customer Report 2023
(5a+b) Salesforce State of the Connected Customer Report 2023
(6) Salesforce Research & Insights 2024
(7a+b) Salesforce State of the Connected Customer Report 2023
- Salesforce “State of the connected Customer” 6th Edition
- ChatGPT
- Personalization, Data Security, and Speed Drive Customer Loyalty Amid Uncertainty — Salesforce Research
- Salesforce Research & Insights 2024
- Collinson research 2022, UAE, KSA, UK
Gary Chandler,
Product Director
Fiona Perks,
Business Solutions Director
BY
Learn more at our exclusive AI Marketplace of Ideas event this Septemeber at our London Bridge Office.
REGISTER
(1)
(2a)
(2b)
(3a)
(3b)
(4a)
(4b)
(4c)
(5a)
(5b)
(6)
(7b)
(7a)