MASTER'S THESIS
on the topic
Agentic Artificial Intelligence in Digital Marketing: Transition to Autonomy
and the Concept of the “Dual Addressee”
, SUMMARY OF THE MASTER'S THESIS
This master's thesis explores the transition from rule-based automation to the systematic
implementation of autonomous marketing agents (Agentic AI) in digital marketing. The main
objective of research is to analyze how this technological evolution transforms communication
models and imposes the need to restructure digital content for the needs of algorithmic
intermediaries.
Theoretically, the study defines and argues for the concept of the “dual addressee” (the Human-
AI-Human model). In this new communication framework, the classic B2C (business-to-
customer) approach is upgraded by integrating a mandatory intermediate level of interaction,
B2A (business-to-agent). To effectively reach the end user, the modern marketing message must
be structured and optimized initially for autonomous AI assistants through Answer Engine
Optimization (AEO) methods.
The empirical part of the thesis applies a structured contextual analysis of multinational
companies from three different sectors: FMCG (Unilever, P&G), e-commerce (Amazon), and
financial services (ING Bank, JPMorgan Chase). The data shows that the implementation of
agent autonomy, contextual memory, and API orchestration leads to statistically significant
improvements in key metrics, including growth in return on advertising spend (ROAS) and
increase in long-term customer value (LTV).
Alongside the benefits identified, the study identifies the risks associated with delayed
technological adaptation and outlines the trend towards increasing competitive polarization.
Organizations that do not adapt their information architectures to the requirements of AI
intermediaries face the risk of significant loss of market share and digital presence.
In conclusion, the thesis highlights the need to redefine the role of the marketer from an
operational campaign executor to a system architect managing algorithmic trust. The study
provides theoretically ground guidelines for adapting marketing strategies to an environment in
which artificial intelligence plays the role of an active mediator in consumer decision-making.
, Contents
List of tables and figures 6
INTRODUCTION 7
CHAPTER ONE 10
THEORETICAL FRAMEWORK OF DIGITAL MARKETING AND ARTIFICIAL
INTELLIGENCE 10
1.1. Evolution of digital marketing and transition to automation 10
1.2. Defining Artificial Intelligence (AI) in a Marketing Context 12
1.3. Strategic Planning through AI: Adapting Classic Models 13
1.4. The New Paradigm: Agent AI as an Object of Marketing and the Concept of the "Dual
Addressee" (Human-AI-Human Marketing) 15
CHAPTER TWO 18
TYPOLOGY AND TOOLKIT OF AGENT ARTIFICIAL INTELLIGENCE 18
2.1. Architecture of agent-based AI systems: the transition from passive algorithms to
autonomous agents 18
2.2. Autonomous agents for content orchestration and hyper-personalized campaigns 20
2.3. AI agents in search engines: The transition from SEO to AEO (Answer Engine
Optimization) and conversational search 22
2.4. Conversational Agents and the Transition to Fully Autonomous Customer Service 23
2.5. Proactive agent systems for big data analysis and attribution modeling 25
CHAPTER THREE. 28
RESEARCH METHODOLOGY AND ROLE OF AI IN SELECTED SECTORS 28
3.1. Methodological approach and research plan 28
3.1.1. Research paradigm and essential position 28
3.1.2. Research design: qualitative research with quantitative components 29
3.1.3. Working hypotheses and criteria for their operational verification 30
3.2. Systematic analysis of secondary data: protocol and hierarchy of sources 31
3.2.1. Justification of the method 31
3.2.2. Criteria for inclusion and exclusion of sources 32
3.2.3. Hierarchy and assessment of the reliability of secondary sources 33
3.3. Structured contextual analysis of multiple objects (according to the methodology of Yin,
2018) 34
3.3.1. Theoretical justification of the research method 34
3.3.2. Criteria for site selection and logic of sector grouping 36
3.3.3. Unit of analysis, data collection protocol and chain of evidence 37
3.4. Validity, study limitations and ethical considerations 38
3.4.1. Construct validity and triangulation 38
3.4.2. Limitations of the method and management of research biases 39
3.4.3. Research ethics and academic transparency 39
CHAPTER FOUR 41
EMPIRICAL RESEARCH AND ANALYSIS OF SELECTED OBJECTS 41
4.1. Macro analysis of the global AI transformation in digital marketing (2019-2026) 41
4.1.1. Investment dynamics and evolution of marketing budgets 42
4.1.2. Impact of agent autonomy on key operational metrics 43
4.1.3. Critical Reading: The Programmatic Ecosystem and the Limits of Automation 46
, 4.2. Object 1: Fast Moving Consumer Goods (FMCG) Sector – Analysis of Unilever and
Procter & Gamble 47
4.2.1. AI Transformation at Unilever: Autonomy in Content Production and Orchestration 47
4.2.2. Procter & Gamble: Proactive analytics and working with external tools 48
4.2.3. Comparative analysis and synthesis of results 49
4.2.4. Critical assessment: Strategic deficits and the problem of the “dual addressee” 51
4.3. Object 2: E-commerce Sector – AI Attribution Modeling and Closed Data Loop at
Amazon 52
4.3.1. Amazon Marketing Cloud (AMC): Autonomous Attribution and Memory Building 52
4.3.2. The Rufus Assistant: Practical Implementation of the “Dual Addressee” 55
4.3.3. Critical assessment: Monopolistic dynamics and conflict of interest 55
4.4. Object 3: Financial Sector – Agent AI, Personalized User Journeys and Trust
Management at ING Bank and JPMorgan Chase 56
4.4.1. ING Bank: Autonomy through Next Best Action and contextual memory 57
4.4.2. JPMorgan Chase: Proactive Customer Analytics and Language Model Orchestration57
4.4.3. Agent AI as a financial intermediary and the concept of the “dual addressee” 60
4.4.4. Critical Assessment: The Paradox of Explainability and Trust 60
4.5. Comparative analysis of digital objects and empirical verification of hypotheses 62
4.5.1. Synthetic matrix of industrial sites 62
4.5.2. Universal patterns and contextual modifiers 64
4.5.3. Empirical verification of working hypotheses 65
CONCLUSION 68
REFERENCES 71
on the topic
Agentic Artificial Intelligence in Digital Marketing: Transition to Autonomy
and the Concept of the “Dual Addressee”
, SUMMARY OF THE MASTER'S THESIS
This master's thesis explores the transition from rule-based automation to the systematic
implementation of autonomous marketing agents (Agentic AI) in digital marketing. The main
objective of research is to analyze how this technological evolution transforms communication
models and imposes the need to restructure digital content for the needs of algorithmic
intermediaries.
Theoretically, the study defines and argues for the concept of the “dual addressee” (the Human-
AI-Human model). In this new communication framework, the classic B2C (business-to-
customer) approach is upgraded by integrating a mandatory intermediate level of interaction,
B2A (business-to-agent). To effectively reach the end user, the modern marketing message must
be structured and optimized initially for autonomous AI assistants through Answer Engine
Optimization (AEO) methods.
The empirical part of the thesis applies a structured contextual analysis of multinational
companies from three different sectors: FMCG (Unilever, P&G), e-commerce (Amazon), and
financial services (ING Bank, JPMorgan Chase). The data shows that the implementation of
agent autonomy, contextual memory, and API orchestration leads to statistically significant
improvements in key metrics, including growth in return on advertising spend (ROAS) and
increase in long-term customer value (LTV).
Alongside the benefits identified, the study identifies the risks associated with delayed
technological adaptation and outlines the trend towards increasing competitive polarization.
Organizations that do not adapt their information architectures to the requirements of AI
intermediaries face the risk of significant loss of market share and digital presence.
In conclusion, the thesis highlights the need to redefine the role of the marketer from an
operational campaign executor to a system architect managing algorithmic trust. The study
provides theoretically ground guidelines for adapting marketing strategies to an environment in
which artificial intelligence plays the role of an active mediator in consumer decision-making.
, Contents
List of tables and figures 6
INTRODUCTION 7
CHAPTER ONE 10
THEORETICAL FRAMEWORK OF DIGITAL MARKETING AND ARTIFICIAL
INTELLIGENCE 10
1.1. Evolution of digital marketing and transition to automation 10
1.2. Defining Artificial Intelligence (AI) in a Marketing Context 12
1.3. Strategic Planning through AI: Adapting Classic Models 13
1.4. The New Paradigm: Agent AI as an Object of Marketing and the Concept of the "Dual
Addressee" (Human-AI-Human Marketing) 15
CHAPTER TWO 18
TYPOLOGY AND TOOLKIT OF AGENT ARTIFICIAL INTELLIGENCE 18
2.1. Architecture of agent-based AI systems: the transition from passive algorithms to
autonomous agents 18
2.2. Autonomous agents for content orchestration and hyper-personalized campaigns 20
2.3. AI agents in search engines: The transition from SEO to AEO (Answer Engine
Optimization) and conversational search 22
2.4. Conversational Agents and the Transition to Fully Autonomous Customer Service 23
2.5. Proactive agent systems for big data analysis and attribution modeling 25
CHAPTER THREE. 28
RESEARCH METHODOLOGY AND ROLE OF AI IN SELECTED SECTORS 28
3.1. Methodological approach and research plan 28
3.1.1. Research paradigm and essential position 28
3.1.2. Research design: qualitative research with quantitative components 29
3.1.3. Working hypotheses and criteria for their operational verification 30
3.2. Systematic analysis of secondary data: protocol and hierarchy of sources 31
3.2.1. Justification of the method 31
3.2.2. Criteria for inclusion and exclusion of sources 32
3.2.3. Hierarchy and assessment of the reliability of secondary sources 33
3.3. Structured contextual analysis of multiple objects (according to the methodology of Yin,
2018) 34
3.3.1. Theoretical justification of the research method 34
3.3.2. Criteria for site selection and logic of sector grouping 36
3.3.3. Unit of analysis, data collection protocol and chain of evidence 37
3.4. Validity, study limitations and ethical considerations 38
3.4.1. Construct validity and triangulation 38
3.4.2. Limitations of the method and management of research biases 39
3.4.3. Research ethics and academic transparency 39
CHAPTER FOUR 41
EMPIRICAL RESEARCH AND ANALYSIS OF SELECTED OBJECTS 41
4.1. Macro analysis of the global AI transformation in digital marketing (2019-2026) 41
4.1.1. Investment dynamics and evolution of marketing budgets 42
4.1.2. Impact of agent autonomy on key operational metrics 43
4.1.3. Critical Reading: The Programmatic Ecosystem and the Limits of Automation 46
, 4.2. Object 1: Fast Moving Consumer Goods (FMCG) Sector – Analysis of Unilever and
Procter & Gamble 47
4.2.1. AI Transformation at Unilever: Autonomy in Content Production and Orchestration 47
4.2.2. Procter & Gamble: Proactive analytics and working with external tools 48
4.2.3. Comparative analysis and synthesis of results 49
4.2.4. Critical assessment: Strategic deficits and the problem of the “dual addressee” 51
4.3. Object 2: E-commerce Sector – AI Attribution Modeling and Closed Data Loop at
Amazon 52
4.3.1. Amazon Marketing Cloud (AMC): Autonomous Attribution and Memory Building 52
4.3.2. The Rufus Assistant: Practical Implementation of the “Dual Addressee” 55
4.3.3. Critical assessment: Monopolistic dynamics and conflict of interest 55
4.4. Object 3: Financial Sector – Agent AI, Personalized User Journeys and Trust
Management at ING Bank and JPMorgan Chase 56
4.4.1. ING Bank: Autonomy through Next Best Action and contextual memory 57
4.4.2. JPMorgan Chase: Proactive Customer Analytics and Language Model Orchestration57
4.4.3. Agent AI as a financial intermediary and the concept of the “dual addressee” 60
4.4.4. Critical Assessment: The Paradox of Explainability and Trust 60
4.5. Comparative analysis of digital objects and empirical verification of hypotheses 62
4.5.1. Synthetic matrix of industrial sites 62
4.5.2. Universal patterns and contextual modifiers 64
4.5.3. Empirical verification of working hypotheses 65
CONCLUSION 68
REFERENCES 71
