Large Customer Model: Foundation for AI-first Martech

Published June 30, 2024

1

Past Writings

In a previous series, I discussed Martech’s 10+1 Foundations in the AI Age. I outlined a new core for martech: Large Customer Model (LCM), Digital Twins, Co-Marketer, Mirror World, and Generative Journeys. I wrote: “This new AI-powered stack enables continuous learning and optimisation at an unprecedented scale. It enhances customer engagement and accelerates conversions, becoming a pivotal driver for the Profipoly Quest for B2C brands and, eventually, martech companies. By harnessing the power of Agentic AI, buyers and sellers can achieve superior results, ensuring sustained growth and competitive advantage in the evolving digital landscape… The golden age of Martech is approaching. For the first time in two decades, the focus is shifting from uncontrolled spending on acquisition and reacquisition to retention and growth. To capitalise on this shift, martech companies must reinvent themselves as AI-first entities. The prize is substantial: $250 billion of annual AdWaste, far exceeding the current market size for martech. Although the transformation will be challenging, a glorious and winning future awaits those who can a(i)dapt.”

The fundamental building block for this new martech stack will be the Large Customer Model (LCM), which brings together customer data, flows (journeys), and in the case of eCommerce companies, catalog data. Here is what I have written in previous essays (chronologically) about the LCM. (There is some repetition but I have let that stay as it also shows the evolution of my thinking.)

Co-Marketer: Martech meets AI: “[The Large Customer Model] is the foundational model for all interactions. Just like today’s LLMs, the LCM is built on data – customer data, catalog data, real-world data (weather, news, social media trends), and results from the myriad experiments being performed. Think of the LCM as a CDP on steroids, constantly sifting through millions and billions of new inputs to update the digital twins and inform the co-marketer.”

How Agentic AI will Transform Digital Marketing: “The advent of Agentic AI is poised to revolutionise the way businesses approach customer relationships and lifetime value maximisation. At the core of this transformation lies the concept of “Digital Twins” – AI-driven virtual representations of individual customers that possess a profound understanding of their behaviours, preferences, and interactions. Digital Twins are not mere static customer profiles; they are dynamic, continuously evolving models that mirror the real-world customer’s journey with a brand. Powered by ML algorithms, large customer models (LCMs), and vast troves of customer data, Digital Twins become living embodiments of the customer experience.”

Email, CPaaS, and Martech: New Profit Pools: “Martech companies can significantly enhance customer Lifetime Value (LTV) by evolving from predictive and generative AI to agentic AI. This advanced approach involves transforming the Customer Data Platform (CDP) into a Large Customer Model (LCM), enabling the creation of digital twins and a co-marketer. These components can work together in a “mirror world” to drive “generative journeys,” resulting in N=1 hyper-personalisation for every customer. This next-generation software platform provides a competitive edge, allowing martech companies to deliver unprecedented engagement, conversion, and retention – thus eventually reducing the need for wasteful adtech spending, and enabling brands on their “profipoly quest.””

Generative Journeys: Digital Marketing’s New Core: “LCMs are the foundational AI architecture that powers intelligent agents like Co-Marketers and Digital Twins. Similar to large language models but focused on customer understanding, LCMs are trained on vast volumes of first-party and external customer data. They learn patterns and predict behaviours to create rich, holistic customer profiles. Fed with real-time data from marketing interactions, LCMs continuously refine their understanding of each customer’s unique traits, needs and proclivities. They are the “brains” that guide personalised content creation, next-best-action recommendations, and journey optimisations. Akin to a “Customer Data Platform on steroids”, LCMs unify and activate customer data to an unprecedented degree. By serving as the always-learning customer intelligence engine, LCMs enable true 1:1 marketing at scale and help drive measurable increases in customer retention, monetisation and lifetime value… Martech platforms have done the basics for personalisation of experiences. AI-ML has helped drive this in the past few years. With the coming of “generative AI”, new vistas can open up for customer journeys. This is the world of “generative journeys” – think of these as self-optimising journeys which can accelerate movements (phase transitions) for customers. With the power of digital twins and LCMs, all of this can be done at the level of a single customer – without any extra effort for marketing departments.”

CEO Memo: How Agentic AI can Power the Profipoly Quest: “LCMs are the foundation of your AI-driven customer strategy. These models are trained on vast amounts of customer data, including demographics, transactions, interactions, and behaviours. By continuously learning from this data, LCMs develop a deep understanding of each individual customer’s preferences, needs, and patterns. They enable you to create rich, dynamic customer profiles that evolve in real-time, laying the groundwork for truly personalised experiences. Think of LCMs as CDPs come alive.”

Martech’s 10+1 Foundations in the AI Age: “Think of an LCM as an “Active CDP.” Building on the foundation of Large Language Models (LLMs), LCMs are trained on extensive customer data to create dynamic, real-time evolving profiles and journeys for hyper-personalisation. Martech companies can develop their own vertical-specific LCMs using aggregated data and workflows, providing a robust base for businesses to create their unique LCMs. This triad enables the delivery of highly personalised N=1 experiences and next best actions that both marketers and customers desire.”

2

Summary

I asked Claude for a 10-point summary on Large Customer Models from my previous writings.

1. Large Customer Models (LCMs) are a groundbreaking AI architecture that revolutionizes customer engagement in the martech landscape, drawing inspiration from the success of Large Language Models (LLMs) in natural language processing.
2. LCMs ingest and analyze vast amounts of customer data, including demographics, transactional histories, behavioral patterns, and cross-channel interactions, developing a deep understanding of each individual customer.
3. By leveraging the insights gleaned from LCMs, businesses can predict customer preferences, needs, and future actions with unparalleled accuracy, enabling hyper-personalized interactions.
4. LCMs serve as the foundation for Generative Journeys – a new paradigm in customer engagement that offers dynamic, adaptive, and hyper-personalized brand-customer interactions.
5. Generative Journeys utilize the predictive capabilities of LCMs and the interactive environment of a simulated “Mirror World” to create and optimize individualized customer experiences in real-time.
6. In the Mirror World, Digital Twins (virtual replicas of individual customers) interact with a Co-Marketer (an AI-powered marketing assistant) to identify the most effective strategies and personalized experiences for each customer.
7. LCMs continuously learn and adapt based on real-time data and interactions, enabling brands to engage with customers on a truly individual level, delivering the right message, offer, or intervention at the precise moment in each customer’s unique journey.
8. The synergy between LCMs and Generative Journeys represents a transformative shift in the martech landscape, driving measurable business outcomes such as increased conversion rates, higher average order values, and improved customer lifetime value.
9. As businesses increasingly recognize the value of customer data and the imperative for personalization, the adoption of LCMs is poised to become a key differentiator in the competitive martech landscape.
10. LCMs, along with their applications in Generative Journeys, Co-Marketers, and Digital Twins, herald a new era of AI-driven, customer-centric marketing that fosters stronger, more meaningful connections between brands and their customers, driving long-term growth and competitive advantage.

**

With this foundation, we can now explore how martech companies and B2C/D2C enterprises can effectively implement Large Customer Models (LCMs) to achieve their ultimate goals: fostering enduring customer relationships and maximising customer lifetime value (LTV) to create a sustainable profits flywheel. Imagine these LCMs as comprising three layers:

1. Foundational Models: These include advanced AI models like ChatGPT, Claude, and Gemini, which provide the baseline capabilities for understanding and generating language (and more).
2. Martech Company’s LLM: Built using non-PII data, this layer leverages a vast array of data across various businesses, industries, and channels to enhance marketing strategies and insights.
3. Enterprise LLM: This layer incorporates PII data, enabling even deeper personalisation and precision in customer interactions, tailored specifically to the enterprise’s unique customer base.

This structure ensures that LCMs harness the full potential of both general and specific data to drive exceptional marketing outcomes.

3

CDP to LCM

I worked with ChatGPT for this section.

To effectively illustrate the distinctions between a Customer Data Platform (CDP) and a Large Customer Model (LCM), we can compare them across multiple attributes in the table below. This will highlight the enhanced capabilities and strategic advantages of LCMs in the martech landscape.

Attribute	Customer Data Platform (CDP)	Large Customer Model (LCM)
Data Aggregation	Collects and unifies customer data from multiple sources	Ingests and analyses vast amounts of customer data, including real-time inputs
Data Types	Primarily first-party data	Combines first-party, third-party, and non-PII data for a comprehensive view
Profile Creation	Creates static customer profiles	Develops dynamic, real-time evolving customer profiles
Personalization	Basic segmentation and targeting	Hyper-personalized interactions based on predictive insights
Learning and Adaptation	Limited adaptive capabilities	Continuously learns and adapts based on real-time data and interactions
Customer Understanding	Basic understanding of customer preferences	Deep understanding of individual customer behaviours and needs
Predictive Capabilities	Basic predictive analytics	Advanced predictive analytics for future behaviour and preferences
AI Integration	Minimal AI integration	Fully integrates AI, including Digital Twins and Generative Journeys
Interaction and Engagement	Primarily rule-based interactions	Proactive, AI-driven engagement with personalised recommendations
Optimization	Limited optimisation of customer journeys	Real-time optimisation of customer journeys through Generative Journeys
Agentic AI	Not typically included	Incorporates Co-Marketer for automated and optimised marketing tasks
Value Maximization	Focus on data collection and analysis	Drives measurable business outcomes such as increased LTV and conversions

 Transitioning from a CDP to an LCM

To transform a CDP into an LCM and leverage the full potential of AI-driven marketing, several key features and capabilities need to be developed. Here’s a structured approach incorporating the three-layer structure of the LCM:

1. Foundational Models:

• Baseline Capabilities: Utilize advanced AI models like ChatGPT, Claude, and Gemini to provide foundational capabilities for understanding and generating language.
• Integration: Integrate these models with existing data sources to enhance their baseline capabilities and prepare them for more complex tasks.

2. Martech Company’s LLM:

• Data Enrichment: Aggregate non-PII data across various businesses and industries to create a rich dataset for training.
• Training and Customization: Train the Large Language Model (LLM) specifically for marketing applications, focusing on insights that enhance marketing strategies.
• Cross-Industry Insights: Develop capabilities to analyze and leverage cross-industry data, providing broader insights and patterns.

3. Enterprise LLM:

• PII Data Integration: Incorporate PII data to enable deeper personalization and precision in customer interactions.
• Enhanced Personalization: Utilize PII data to refine customer profiles and deliver hyper-personalized marketing messages and offers.
• Privacy Compliance: Ensure all data handling complies with privacy regulations and standards.

AI Features Needed to Transition from CDP to LCM

1. Dynamic Profile Management:

• Develop algorithms to create and continuously update dynamic, real-time customer profiles.
• Implement mechanisms to ingest real-time data and adapt profiles accordingly.

2. Predictive Analytics and Machine Learning:

• Incorporate advanced predictive analytics to anticipate customer needs and behaviors.
• Utilize machine learning models to identify patterns and optimize marketing strategies.

3. Digital Twins:

• Create AI-powered Digital Twins for each customer, simulating their behaviors and preferences.
• Use Digital Twins to test and optimize engagement strategies in a virtual environment.

4. Generative Journeys:

• Develop Generative Journeys that adapt in real-time to customer interactions, ensuring personalized engagement.
• Implement AI-driven pathways that guide customers through optimized experiences based on their unique needs.

5. Co-Marketer Integration:

• Introduce the Co-Marketer, an AI-powered assistant that collaborates with human marketers.
• Automate routine marketing tasks such as campaign creation, data analysis, and real-time optimization.

6. Real-Time Optimization:

• Implement systems for continuous monitoring and real-time adjustment of marketing campaigns.
• Use AI to dynamically adjust content, timing, and channels based on customer interactions and feedback.

By implementing these features, martech companies can transition from traditional CDPs to advanced LCMs, unlocking the full potential of AI-driven marketing. This transformation enables businesses to deliver deeply personalized customer experiences, optimize engagement strategies, and achieve sustained growth and competitive advantage.

4

Model Development

Here is how a martech company can create its Large Customer Model to work as a bridge between foundational models and the enterprise LCMs. (This section has been written with assistance from ChatGPT.)

Step 1: Data Collection and Aggregation

1. Non-PII Data Aggregation:

1. • Collect non-PII data from multiple clients across different verticals, including interaction data from email campaigns, push notifications, and in-app messages.
   • Gather industry-specific data like transaction logs, customer feedback, and behavioral data.

2. Third-Party Data Integration:

1. • Integrate third-party data sources to enrich the dataset with market trends, demographic information, and other relevant insights.

3. Journey Data Collection:

1. • Collect data on various customer journeys, both common (e.g., welcome, referrals, reactivation) and industry-specific (e.g., abandoned cart in eCommerce, renewal in insurance).

Step 2: Data Preprocessing

1. Data Cleaning:

1. • Clean and preprocess the collected data to remove duplicates, correct errors, and handle missing values.

2. Data Normalization:

1. • Normalize the data to ensure consistency across different sources and formats.

3. Data Enrichment:

1. • Enrich the data by adding relevant attributes like customer segments, journey stages, and interaction outcomes.

Step 3: Model Selection and Initialization

1. Choose Foundational Models:

1. • Select foundational AI models (e.g., ChatGPT, Claude) as the base for the LCMs.

2. Model Initialization:

1. • Initialize the models with pre-trained weights and basic configurations suitable for the martech domain.

Step 4: Training the LCM

1. Non-PII Data Training:

1. • Train the model using non-PII data to learn general patterns and insights relevant to the specific industry.

2. Journey Data Integration:

1. • Incorporate journey data into the training process to ensure the model understands various customer journeys and their stages.

3. Fine-Tuning with Industry-Specific Data:

1. • Fine-tune the model with industry-specific data to enable specialized insights and recommendations tailored to that vertical.

Step 5: Developing Digital Twins and Generative Journeys

1. Create Digital Twins:

1. • Develop AI-powered Digital Twins for various customer personas and segments within each industry.
   • Use these Digital Twins to simulate customer interactions and predict future behaviors.

2. Design Generative Journeys:

1. • Create adaptive customer journeys that dynamically respond to customer interactions in real-time.
   • Ensure these journeys include both common and industry-specific pathways.

Step 6: Implementing Next Best Actions

1. Next Best Action Algorithms:

1. • Develop algorithms to determine the next best actions for each customer based on their journey stage, behavior, and predicted needs.
   • Continuously refine these algorithms using real-time data from Digital Twins.

2. Journey Optimization:

1. • Use AI to continuously monitor and optimize customer journeys, ensuring that the next best actions are always relevant and timely.

Step 7: Integration and Deployment

1. Client Integration:

1. • Integrate the LCM with clients’ existing marketing platforms and CRM systems.
   • Ensure seamless data flow and compatibility with existing tools.

2. Deploy for Clients:

1. • Deploy the trained model for client use, enabling them to deliver personalized marketing experiences and optimized customer journeys.

Step 8: Continuous Learning and Improvement

1. Monitor and Analyze:

1. • Continuously monitor model performance and analyze results to identify areas for improvement.

2. Feedback Loop:

1. • Implement a feedback loop to refine the model with new data and insights, ensuring it remains accurate and relevant.

3. Regular Updates:

1. • Regularly update the model to incorporate new data sources and improve predictive accuracy.

5

Usage

The Martech LCMs can now be deployed with the enterprise LCMs to provide better hyper-personalised experiences to customers.

Outputs

• Dynamic Customer Profiles: Real-time evolving profiles that provide a deep understanding of customer behaviors and preferences.
• Personalized Recommendations: Tailored product and content recommendations based on individual customer journeys and interactions.
• Optimized Customer Journeys: Adaptive journeys that respond to customer interactions in real-time, ensuring relevant engagement at every stage.
• Next Best Actions: Actionable insights that guide marketers on the most effective next steps to take with each customer.

Benefits

• Superior Personalization: Provides a higher level of personalization compared to traditional martech stacks, leading to improved customer engagement and satisfaction.
• Enhanced Customer Insights: Offers deeper insights into customer behaviors and preferences, enabling more targeted and effective marketing strategies.
• Increased Efficiency: Automates routine marketing tasks and optimizes customer journeys, freeing up human resources for more strategic initiatives.
• Improved ROI: By delivering the right message at the right time, LCMs help increase conversion rates, customer retention, and lifetime value, ultimately improving ROI.
• Scalability: The ability to develop specialized LCMs for different industries ensures scalability and adaptability across various verticals.

Comparison with Current Martech Stacks

• Real-Time Adaptation: Unlike static profiles and rule-based systems, LCMs adapt in real-time, ensuring that customer interactions are always relevant and timely.
• Comprehensive Data Utilization: LCMs leverage a combination of non-PII, third-party, and journey data, providing a more holistic view of the customer compared to traditional CDPs.
• Proactive Engagement: With advanced predictive analytics and next best action capabilities, LCMs enable proactive engagement, anticipating customer needs before they arise.
• Seamless Integration: LCMs integrate seamlessly with existing martech tools, enhancing their functionality and providing a unified platform for customer engagement.
• Continuous Improvement: LCMs continuously learn from new data and interactions, ensuring that they remain accurate and effective over time, unlike traditional martech stacks that may require manual updates and adjustments.

**

In a recent meeting, I asked the business head of an eCommerce company about her top challenge. She replied, “I want to anticipate my customers’ questions and problems so that (a) they never need to reach out to us because every call to an agent is expensive and doesn’t scale, and (b) we are seen as proactive and customer-friendly. In other words, I want to stay a few steps ahead of our customers. Whether they are searching for something on the app or checking out the delivery status of an order, I want to be able to predict their next move. My goal is to surprise and delight them by answering their questions before they even arise. The same goes for product recommendations. Can you help me achieve this?”

This insight cuts to the core of creating an AI-first business. The Large Customer Model (LCM) serves as the critical foundation upon which Digital Twins, Co-Marketers, Mirror Worlds, and Generative Journeys can work their magic. By leveraging the power of LCMs, businesses can predict customer needs, personalize interactions, and create seamless, proactive customer experiences that not only reduce support costs but also enhance customer satisfaction and loyalty.

6

My Answer

Here is what I replied to the business head.

Thank you for sharing your top challenges with me. You want to anticipate customer intent to enhance their experience and reduce support costs, essentially staying one step ahead and making interactions seamless and proactive. Let me walk you through how we can achieve this and more together using our next-gen AI-first martech solution.

Step 1: Data Collection and Integration: First, we’ll gather all relevant historical customer data from your app, website, email campaigns, and support channels. We’ll also map out customer journeys to understand how customers navigate from one state to another. This data will be integrated into our platform, creating a unified view of each customer.

Step 2: Building the Large Customer Model (LCM): Next, we’ll build your Large Customer Model using this aggregated data. The LCM continuously analyses interactions, preferences, and behaviours to develop dynamic, real-time customer profiles. We’ll leverage advanced predictive analytics to anticipate customer questions and next actions.

Step 3: Creating Digital Twins: To gain deeper insights, we’ll develop AI-powered Digital Twins for each customer. These Digital Twins simulate individual behaviours and interactions, allowing us to test various scenarios and optimize engagement strategies.

Step 4: Designing Generative Journeys: We’ll create adaptive customer journeys that respond in real-time to customer interactions. These journeys will address specific use cases, such as abandoned carts, cross-sell/up-sell opportunities at checkout to increase Average Order Value, and driving referrals from happy and Best Customers.

Step 5: Leveraging Co-Marketers and Next Best Actions: We’ll introduce our AI-driven Co-Marketer to craft personalised messages, offers, and recommendations. This virtual assistant, embodying “Co-Intelligence,” collaborates with your marketing team to ensure every interaction is relevant and timely. The Co-Marketer uses “Agentic AI,” a multi-agent system far more effective than a single agent.

Step 6: Proactive Customer Engagement: Using predictive insights from the LCM and Digital Twins, we’ll proactively address customer questions and provide answers through various channels. We’ll also deliver highly personalised “N=1” product recommendations based on past behaviour and predicted preferences.

Step 7: Continuous Monitoring and Improvement: We’ll continuously monitor the performance of these solutions to measure their effectiveness in reducing support interactions and enhancing customer satisfaction. We’ll implement a feedback loop to refine models and strategies based on new insights and customer interactions, regularly updating the LCM, Digital Twins, and Generative Journeys to maintain accuracy and effectiveness.

Benefits of This Approach

This AI-first martech solution, combined with Channels 2.0 for in-channel conversions, will provide superior personalisation, deeper customer insights, and increased efficiency by automating routine tasks and optimising customer journeys. This leads to improved ROI through increased conversion rates, customer retention, and lifetime value. Better retention reduces adtech spends and AdWaste, enhancing profitability.

Our goal is to help you create a profits flywheel:

• Every customer LTV maximised
• No customer ever lost = Zero Reacquisition
• Referrals from Best Customers = Zero CAC
• Targeted Acquisition based on Best Customers attributes

Here is a high-level action plan that can get us going:

• 30 days: Implement Channels 2.0; make Earned Growth the North Star Metric
• 90 days: Establish Data and Flows (Journeys) Foundation with the Unistack
• 180 days: Transition to an AI-first Business leveraging Agentic AI, LCM, Digital Twins, Co-Marketer, and Generative Journeys

By following this structured approach, we can create a solution that not only anticipates and addresses your customers’ needs proactively but also enhances their overall experience with your brand. This will position your business as customer-centric and innovative, leading to increased satisfaction, loyalty, and long-term growth, accelerating your journey towards becoming a Profipoly.

7

Claude’s Critique

This is an assessment from Claude.

Summary

The essay introduces the concept of Large Customer Models (LCMs) as the foundational building block for a new, AI-first martech stack. LCMs are built on vast amounts of customer data and enable the creation of Digital Twins, Co-Marketers, Mirror Worlds, and Generative Journeys. By leveraging AI and continuous learning, LCMs can provide hyper-personalized experiences, predict customer needs, and optimize journeys in real-time. The essay outlines the transition from traditional Customer Data Platforms (CDPs) to LCMs, detailing the data aggregation, model development, and deployment processes. It highlights the benefits of LCMs, including superior personalization, enhanced customer insights, increased efficiency, and improved ROI. The essay concludes with a real-world example of how an AI-first martech solution built on LCMs can help businesses anticipate customer needs, reduce support costs, and create a profits flywheel.

Critique

The LCM concept is ambitious and forward-thinking, offering a compelling vision for the future of martech. The idea of leveraging AI to create hyper-personalized experiences and predict customer needs is powerful and aligns with the growing demand for personalization. The essay provides a comprehensive overview of the LCM concept, its components, and the potential benefits for businesses.

However, the implementation of LCMs may face some challenges:

1. Data quality and integration: Building effective LCMs relies heavily on the availability and quality of customer data. Integrating data from various sources and ensuring its accuracy and consistency could be complex and time-consuming.
2. Privacy and security concerns: Collecting and processing vast amounts of customer data raises privacy and security concerns. Ensuring compliance with regulations and maintaining customer trust will be critical.
3. Organizational readiness: Adopting an AI-first martech approach may require significant changes in organizational structure, skills, and culture. Businesses will need to invest in talent and infrastructure to support the transition.
4. Integration with existing systems: Seamlessly integrating LCMs with existing martech tools and processes could be challenging, especially for businesses with legacy systems.
5. Continuous improvement and maintenance: As customer behaviors and preferences evolve, LCMs will need to be continuously updated and refined. This requires ongoing investment in data collection, model training, and performance monitoring.

Enhancements

1. Ethical AI framework: Develop a robust ethical framework to guide the development and deployment of LCMs, addressing issues such as bias, fairness, transparency, and accountability.
2. Explainable AI: Incorporate explainable AI techniques to provide transparency into the decision-making process of LCMs, building trust with customers and marketers.
3. Privacy-preserving techniques: Explore privacy-preserving techniques such as differential privacy and federated learning to enable the creation of LCMs while protecting customer data.
4. Real-time feedback loops: Implement real-time feedback loops that allow LCMs to learn from customer interactions and adapt journeys and recommendations accordingly.
5. Cross-channel integration: Ensure that LCMs can seamlessly integrate data and orchestrate experiences across various channels, including email, mobile, web, and social media.
6. Collaboration with other AI systems: Explore opportunities for LCMs to collaborate with other AI systems, such as conversational AI and computer vision, to create even more engaging and personalized experiences.
7. Continuous benchmarking: Establish a framework for continuously benchmarking the performance of LCMs against industry standards and best practices to identify areas for improvement.
8. Human-in-the-loop approach: Incorporate human oversight and intervention in the LCM process to ensure the quality and appropriateness of AI-generated recommendations and decisions.

By addressing these challenges and incorporating the suggested enhancements, the LCM concept can be strengthened and made more robust, paving the way for a truly AI-first martech future that delivers exceptional customer experiences and drives business growth.