Table 4 Empirical grounding of the AEL... | Emerald Publishing

Table 4

Empirical grounding of the AEL model: stage definitions, indicators and cross-case evidence

AEL stage	Stage definition	Empirical indicators observed across cases	Cross-case examples	Quotes from the interviews
Stage 1: Data-enabled reach and salience	Organizational practices that use AI to allocate attention and optimize exposure sequencing across platforms – not creating awareness from zero, but dynamically managing reach, frequency and contextual relevance	Real-time audience modeling; platform-driven reach optimization; rapid testing; automated creative versioning for micro-contexts; localization of global assets	Case A: rapid creative variant production and localization; reduced time-to-market and improved relevance through scalable adaptation. Case B: dynamic email/app interfaces adapting to preferences. Case C: DTC/app ecosystem enabling continuous behavioral sensing and personalized nudges	“We used AI to localize the ad content by finding alternatives of our iconic advertising context and stayed relevant in every market we serve without starting the ad production from scratch” (CASE A) “With the help of our platform's AI, we identified beauty audiences at scale that we would never find manually, thereby expanding reach beyond existing consumer segments” (CASE B) “The AI integrated in our systems helps us determine which segments are getting popular and which less, while helping us communicate to athletes and build a relationship with them at the exact time when they are searching for items and comparing it” (CASE C)
Stage 2: Dynamic engagement	AI-mediated interaction design that turns touchpoints into adaptive dialogues and service-like experiences; engagement emerges from responsiveness, tone control, and context-sensitive interaction flows	Conversational flows; adaptive tone; service–communication integration; sentiment detection and routing; escalation to humans; multimodal assistance (e.g. AR try-on guidance)	Case A: interactive UGC/campaign participation enabled by GenAI. Case B: GPT-4–powered chatbot interactions; tone adaptation via sentiment analysis; AR/vision-enabled guidance; reported engagement lift (e.g. “11% higher” interactions vs static pages)	“Our creative ads used to be static and we treated it as a movie creation, while now with the AI we treat creative as a living system and learn with the audience when they engage with our ads in real time” (CASE A) “AI is actually changing the total experience for our consumers, as it adapts the communication tone and choices based on the AI-assessed mood of the consumer, making the experience sound much more like they were talking to a real beauty advisor” (CASE B) “We continuously refresh the ad creative based on what AI tells us that our potential consumers actually watched” (CASE C)
Stage 3: Predictive conversion	Machine-supported prioritization of “next best actions” that determines when, how, and whether to intervene with offers, content or pushed data – integrating propensity, timing and sequencing	Next-best-action models; offer timing optimization; churn propensity triggers; automated variant allocation; dynamic CTA; automated budget shifts; conversion path tailoring	Case B: conversion gains within AI-mediated journeys. Case C: predictive AI embedded into app-based push; continuous segmentation; concerns about over-personalization/cognitive fatigue and partner tensions	“The AI is used to predict, based on viewing behavior data, which story version of the ad would resonate best with our target audiences” (CASE A) “For the consumer journey to be felt like a consultation, the predictive ability of AI to shift the communication tone, for instance, from inspiration to problem solving, is critical” (CASE B) “The AI model helps us to figure out when is the best time to engage with the athletes, which is critical, as in case we act too soon, we may lose trust” (CASE C)
Stage 4: Co-creation-enabled relationship deepening	AI-supported mechanisms that lower the cost of identity-relevant participation and community contribution, expanding loyalty beyond repeat purchase to include advocacy, co-creation, and engagement behaviors	Co-created assets; remixing and sharing; community participation; identity-relevant customization; UGC participation; engagement beyond purchase; sustained interaction loops	Case A: GenAI-enabled consumer co-creation/remixed packaging and UGC uptake; concerns about authenticity and brand voice drift. Case C: lifestyle ecosystem and daily-life integration via app pushes and personalization	“We invited each single one of our consumers to co-create ads with the help of AI and share them, which gave everybody a personal feel, while still keeping our brand's iconography intact” (CASE A) “We want the community to help shape the brand's beauty conversation by letting consumers build their looks and share them, thereby influencing what we feature in our marketing communication” (CASE B) “We want athletes to tell and shape the story of our products and brand itself – our ultimate goal is to turn the personalization into participation” (CASE C)
Stage 5: Ethical optimization	Embedded oversight that defines acceptable AI behavior and ensures accountability (i.e. creative guardrails, disclosure, audits, and escalation protocols) is treated as a constitutive stage rather than an add-on	Guardrails and style constraints; human-in-the-loop review; disclosure policies; bias/quality audits; privacy/consent controls; compliance monitoring; incident response	Case A: protocols to prevent dilution and generic outputs; internal ethical guidelines. Case B: bias audits for AR try-on; performance improvements after mitigation (e.g. darker skin tone issues; “27% uplift”). Case C: algorithmic diversity strategies and partner data constraints	“Establishing AI governance is critical for mitigating brand-safety risk, as otherwise speed could quickly become a risk to our brand's reputation” (CASE A) “Responsible use of AI in beauty means ensuring privacy and conducting constant performance checks. For instance, when we identified performance gaps between skin tones, we stopped the rollout until it was fixed” (CASE B) “AI should never compromise an athlete's trust, thus we are very strict when it comes to transparency and what data we use and share” (CASE C)

AEL stage	Stage definition	Empirical indicators observed across cases	Cross-case examples	Quotes from the interviews
Stage 1: Data-enabled reach and salience	Organizational practices that use AI to allocate attention and optimize exposure sequencing across platforms – not creating awareness from zero, but dynamically managing reach, frequency and contextual relevance	Real-time audience modeling; platform-driven reach optimization; rapid testing; automated creative versioning for micro-contexts; localization of global assets	Case A: rapid creative variant production and localization; reduced time-to-market and improved relevance through scalable adaptation. Case B: dynamic email/app interfaces adapting to preferences. Case C: DTC/app ecosystem enabling continuous behavioral sensing and personalized nudges	“We used AI to localize the ad content by finding alternatives of our iconic advertising context and stayed relevant in every market we serve without starting the ad production from scratch” (CASE A)“With the help of our platform's AI, we identified beauty audiences at scale that we would never find manually, thereby expanding reach beyond existing consumer segments” (CASE B)“The AI integrated in our systems helps us determine which segments are getting popular and which less, while helping us communicate to athletes and build a relationship with them at the exact time when they are searching for items and comparing it” (CASE C)
Stage 2: Dynamic engagement	AI-mediated interaction design that turns touchpoints into adaptive dialogues and service-like experiences; engagement emerges from responsiveness, tone control, and context-sensitive interaction flows	Conversational flows; adaptive tone; service–communication integration; sentiment detection and routing; escalation to humans; multimodal assistance (e.g. AR try-on guidance)	Case A: interactive UGC/campaign participation enabled by GenAI. Case B: GPT-4–powered chatbot interactions; tone adaptation via sentiment analysis; AR/vision-enabled guidance; reported engagement lift (e.g. “11% higher” interactions vs static pages)	“Our creative ads used to be static and we treated it as a movie creation, while now with the AI we treat creative as a living system and learn with the audience when they engage with our ads in real time” (CASE A)“AI is actually changing the total experience for our consumers, as it adapts the communication tone and choices based on the AI-assessed mood of the consumer, making the experience sound much more like they were talking to a real beauty advisor” (CASE B)“We continuously refresh the ad creative based on what AI tells us that our potential consumers actually watched” (CASE C)
Stage 3: Predictive conversion	Machine-supported prioritization of “next best actions” that determines when, how, and whether to intervene with offers, content or pushed data – integrating propensity, timing and sequencing	Next-best-action models; offer timing optimization; churn propensity triggers; automated variant allocation; dynamic CTA; automated budget shifts; conversion path tailoring	Case B: conversion gains within AI-mediated journeys. Case C: predictive AI embedded into app-based push; continuous segmentation; concerns about over-personalization/cognitive fatigue and partner tensions	“The AI is used to predict, based on viewing behavior data, which story version of the ad would resonate best with our target audiences” (CASE A)“For the consumer journey to be felt like a consultation, the predictive ability of AI to shift the communication tone, for instance, from inspiration to problem solving, is critical” (CASE B)“The AI model helps us to figure out when is the best time to engage with the athletes, which is critical, as in case we act too soon, we may lose trust” (CASE C)
Stage 4: Co-creation-enabled relationship deepening	AI-supported mechanisms that lower the cost of identity-relevant participation and community contribution, expanding loyalty beyond repeat purchase to include advocacy, co-creation, and engagement behaviors	Co-created assets; remixing and sharing; community participation; identity-relevant customization; UGC participation; engagement beyond purchase; sustained interaction loops	Case A: GenAI-enabled consumer co-creation/remixed packaging and UGC uptake; concerns about authenticity and brand voice drift. Case C: lifestyle ecosystem and daily-life integration via app pushes and personalization	“We invited each single one of our consumers to co-create ads with the help of AI and share them, which gave everybody a personal feel, while still keeping our brand's iconography intact” (CASE A)“We want the community to help shape the brand's beauty conversation by letting consumers build their looks and share them, thereby influencing what we feature in our marketing communication” (CASE B)“We want athletes to tell and shape the story of our products and brand itself – our ultimate goal is to turn the personalization into participation” (CASE C)
Stage 5: Ethical optimization	Embedded oversight that defines acceptable AI behavior and ensures accountability (i.e. creative guardrails, disclosure, audits, and escalation protocols) is treated as a constitutive stage rather than an add-on	Guardrails and style constraints; human-in-the-loop review; disclosure policies; bias/quality audits; privacy/consent controls; compliance monitoring; incident response	Case A: protocols to prevent dilution and generic outputs; internal ethical guidelines. Case B: bias audits for AR try-on; performance improvements after mitigation (e.g. darker skin tone issues; “27% uplift”). Case C: algorithmic diversity strategies and partner data constraints	“Establishing AI governance is critical for mitigating brand-safety risk, as otherwise speed could quickly become a risk to our brand's reputation” (CASE A)“Responsible use of AI in beauty means ensuring privacy and conducting constant performance checks. For instance, when we identified performance gaps between skin tones, we stopped the rollout until it was fixed” (CASE B)“AI should never compromise an athlete's trust, thus we are very strict when it comes to transparency and what data we use and share” (CASE C)