Assessing the Irreducibly Human in Marketing Education: A Voice,Judgement and Taste Framework for the Generative AI Era

Phase 1: Inductive Coding and Pattern Identification

Analysis followed Braun and Clarke’s (2019) reflexive thematic analysis framework, combined with inductive theory-building principles from qualitative organisational research (Gioia et al., 2013; Miles et al., 2014). Coding commenced alongside data collection and involved repeated transcript reading to identify practitioner descriptions of capability transformation within AI-mediated marketing work.

Initial coding focused on capturing practitioner language describing human contribution, including references to authorship, identity signalling, decision authority, ethical oversight, evaluation of AI outputs, and responsibility for marketing quality. Codes were intentionally descriptive during early analytic stages to remain close to practitioner meaning-making rather than prematurely abstracting conceptual categories.

Phase 2: Cross-Case Comparison and Capability Clustering

Iterative comparison across transcripts revealed recurring practitioner distinctions between tasks considered automatable and those requiring human oversight. Through cross-case pattern matching (Miles et al., 2014), these practitioner distinctions progressively clustered into three recurrent capability logics:

These clusters recurred across cases and were iteratively stabilised through cross-case comparison, suggesting patterned practitioner sensemaking rather than isolated individual perspectives.

Phase 3: Conceptual Abstraction and Framework Development

Following the identification of capability clusters, analysis progressed to conceptual abstraction consistent with interpretive theory-building approaches (Gioia et al., 2013). Through iterative memo writing, reflexive dialogue between authors, and repeated engagement with transcript data, these clusters were progressively refined into three higher-order constructs: voice, judgement, and taste.

Importantly, these constructs were not predefined analytic categories. Rather, they emerged as higher-order abstractions from recurring practitioner logics concerning expressive authorship, deployment authority over AI use, and evaluative selection among AI-generated alternatives. Voice was conceptualised as expressive authorship through which marketers signal identity, narrative coherence, and relational positioning. Judgement captured practitioner accounts of accountable decision authority, including contextual evaluation, risk assessment, and ethical oversight in AI-mediated workflows. Taste reflected patterned emphasis on evaluative discernment through which marketers select, refine, and differentiate outputs to achieve resonance and cultural relevance.

Consistent with reflexive thematic analysis (Braun & Clarke, 2019), these domains stabilised through iterative movement between data familiarisation, coding, memo writing, and abstraction to ensure interpretive grounding and theoretical coherence. Existing marketing competency and educational capability literatures were engaged only after these constructs had analytically stabilised, serving to position and refine the framework rather than to generate it deductively. Interpretive transparency was further supported through reflexive documentation of analytic decisions and collaborative interrogation of emerging interpretations. Selected illustrative quotations are presented to demonstrate how practitioner discourse informed the development of the final conceptual constructs and to enhance analytic traceability beyond the immediate dataset (see Table 2).

OPEN IN VIEWER

Table 2.

Illustrative Analytic Progression From Practitioner Insights to VJT Constructs.

Illustrative empirical evidence	Clustered capability domain	Conceptual construct
“I sort of feel like we’ve got a duty as an organization to have our own voice rather than be voiced by a computer. . . we want to have a strong brand identity and culture that we are putting out in our newsletters and in our copy.”	Human authorship and identity expression	Voice
“People know if they’re speaking to a person or a machine. . . people want to work with people. . . they don’t want to work with machinery.”
“If I go straight to an LLM. . . all of the natural ways in which the piece of writing could have flowed have been compromised. . . it stops me from embarking on the actual thinking process.”
“AI-produced text is frequently very formal and stiff. . . it needs a lot of editing to reflect how the organisation actually speaks.”
“AI-generated submissions lack a touch of human element. . . there is no thought behind it.”
“I very, very rarely copy and paste. . . my junior team member does and I have to sanity check it. . . it pulled the wool over his eyes but it wouldn’t pull it over mine.”	Decision authority and governance over AI use	Judgement
“Expand your knowledge of AI. . . but reject ideas if they’re not in line with your values as a marketeer.”
“It still needs to be guarded by humans. . . it takes a highly skilled human with marketing knowledge to really get the most out of it.”
AI can suggest propositions but “cannot decide which aligns with the brand’s goals.”
“Marketing has to do a lot with [human] taste. . . even if AI becomes indistinguishable, we’re still going to be biased toward wanting something created by humans.”	Discernment of quality, resonance, and distinctiveness	Taste
“Taste is knowing what’s good, what is worth keeping, and being able to explain why.”
“If you do not have taste, the tool drives you instead of the other way around.”
“The work will move toward understanding what is actually going to appeal to people. . . the conceptualization. . . the actual idea behind the campaign.”
“AI can make things quicker. . . but clever campaigns that appeal to people in a particular way still rely on human creativity.”

Reflexivity and Researcher Positionality

Consistent with reflexive thematic analysis, the researchers recognise that thematic interpretation reflects an interaction between practitioner accounts and researcher expertise (Braun & Clarke, 2019). The first and third authors’ prior research on marketing education and AI adoption informed sensitivity to practitioner discussions of capability transformation while necessitating ongoing reflexive attention to interpretive assumptions. This positioning supported theoretically informed sensemaking while maintaining grounding in participant meaning.

Voice: Expressive Authorship and Relational Positioning

Practitioners consistently described voice as the most vulnerable dimension of AI-assisted marketing work. While generative tools were widely valued for drafting and structural support, participants emphasised that model outputs frequently lacked intentionality, identity signalling, and relational authenticity.

Liam described an AI rewrite of his self-promotional personal statement as “very formal” and “stiff,” noting that although technically accurate, “it was not me” because it followed “a very particular template.” Maya similarly observed that many AI-generated outputs lack “a touch of human element,” producing multiple texts where “there is no thought behind it.” Warren reinforced this concern, suggesting that AI outputs often feel “a bit too polished and a bit hygienic,” generating patterns that “set off our bullshit detectors” when tone or cultural fit is misaligned.

Across accounts, voice was positioned as extending beyond stylistic editing to encompass narrative coherence, audience sensitivity, and embedded brand knowledge. Elliott described remaining “protective” of organisational voice even when working with internal models trained on brand datasets. Practitioners frequently framed AI as providing structural scaffolding, while communicative personality and relational positioning required deliberate human authorship. Voice therefore functioned as expressive authorship through which marketers shape tone, perspective, and audience relationships, signalling identity and relational intent. Participants viewed this capability as particularly fragile under AI-mediated production because it depends on tacit cultural and organisational knowledge that practitioners considered difficult to encode within model training.

Importantly, brand identity was one recurring site where voice became visible, but it did not exhaust how practitioners described the construct. While some participants explicitly linked voice to brand identity, practitioner accounts also revealed broader, non-brand-bound dimensions of voice as human authorship, cognitive engagement, and relational expression. Laura framed voice as organisational authorship, emphasising a “duty. . . to have our own voice rather than be voiced by a computer”. However, other participants positioned voice more generally as human presence and authenticity in communication, noting that “people want to work with people. . . they don’t want to work with. . . computers” and that audiences remain “biased to wanting something that’s created by humans. . . especially. . . as personal as marketing”. Voice was also described as emerging through cognitive and creative processes rather than surface-level stylistic features. Peter explained that relying on AI too early can prevent immersion, as “you never really immerse yourself in the idea” and instead lose “the unique combination. . . and the way my brain works”. Relatedly, participants distinguished human-authored voice from AI outputs characterised as “following formulas”, suggesting that voice involves non-formulaic expression and idea development rather than technical correctness alone. Finally, voice was implicitly linked to audience connection, with marketing described as requiring sensitivity to “how we. . . appeal to human emotions”. Across these accounts, voice extends beyond brand consistency to encompass human-authored expression, idea formation, and audience-oriented communication.

Judgement: Ethical, Contextual, and Strategic Decision Authority

Judgement emerged as the capability practitioners considered most consequential for maintaining professional accountability in AI-mediated workflows. Participants consistently described generative systems as producing outputs that require continuous human interrogation due to hallucinations, bias, and contextual misalignment. As Noah summarised, practitioners must “not take anything literal.”

Practitioner accounts revealed three interconnected dimensions of judgement.

Taste: Evaluative Discernment and Distinctiveness

Taste emerged as a subtle yet widely recognised capability through which practitioners differentiated high-quality marketing work from AI-generated volume. Participants frequently described generative outputs as technically competent but aesthetically or emotionally generic.

Jack characterised AI-generated consultancy language as “very generic,” requiring extensive refinement. Peter similarly argued that models “follow formulas” and “do not really speak to us the same way,” maintaining that “a good writer can write something way, way better” than an LLM. Warren emphasised that practitioners must still determine “what is good” and justify why particular outputs warrant development. Chloe captured the directional nature of this capability, observing that “if you do not have taste, the tool drives you instead of the other way around.”

Practitioners operationalised taste through three recurring practices: curation (filtering large volumes of generated options), originality sensing (recognising when outputs feel derivative or emotionally thin), and coherence construction (maintaining consistency across multimodal brand touchpoints). As Warren explained, AI produces scale, but “taste is knowing what’s good, what is worth keeping”—and being able to explain why. Oliver noted that AI outputs often “feel familiar,” whereas taste “pushes past familiar” to identify distinctive possibilities. Grace similarly described taste as what “holds [campaign elements] together, so it feels intentional.” Across interviews, taste was framed as a developmental capability grounded in exposure, critique, and iterative refinement, and as essential for sustaining distinctiveness under AI acceleration. Recent computational work similarly finds substantial divergence in human aesthetic and stylistic preferences when evaluating generated texts, reinforcing sociological accounts of taste as socially organised rather than universally shared (Chung et al., 2025).

Boundary Delineation and Theme Selection

During cross-case comparison, practitioner descriptions repeatedly converged on a deeper organising concern: how responsibility is retained when generative systems can produce plausible marketing outputs at scale. Although interviews surfaced numerous valued professional skills—including creativity, empathy, collaboration, data literacy, and adaptive learning—these capabilities consistently appeared as enabling mechanisms through which practitioners enacted voice, judgement, and taste, rather than as independent domains of human contribution. Creativity, for example, was associated with ideation and exploration, but practitioner discourse more consistently emphasised human involvement in selecting, refining, and validating ideas once generated. Empathy-informed voice through audience-sensitive communication, while complex problem-solving was commonly expressed through judgement as integrating ethical, contextual, and strategic considerations.

Practitioner accounts also indicated that voice and taste were enacted at different points in the workflow. Voice was invoked during the production of communication itself, where practitioners emphasised authorship, expression, and identity. Laura resisted delegating copy generation to AI because she believed “we’ve got a duty as an organization to have our own voice rather than be voiced by a computer,” framing authorship as integral to organisational identity. Peter similarly noted that beginning with AI can disrupt “all of the natural ways in which the piece of writing could have flowed,” locating voice in the shaping of expression rather than post hoc editing. By contrast, taste was more often articulated once outputs already existed and a comparative decision was required. Practitioners described evaluating multiple viable options and advancing those that demonstrated stronger resonance or distinctiveness, particularly in contexts involving audience engagement and emotional appeal. Taken together, these accounts distinguish voice as the shaping of communicative expression and taste as the comparative evaluation of generated alternatives.

Treating voice, judgement, and taste as analytically separable clarifies how these forms of expertise vary in practice. For example, a marketer may demonstrate a distinctive communicative voice by producing engaging, conversational copy aligned with audience expectations yet exercise poor judgement by publishing an AI-generated campaign claim (e.g., product performance or pricing information) without verification, exposing the organisation to factual error or regulatory risk. Conversely, a practitioner may demonstrate sound judgement by carefully fact-checking outputs and rejecting unreliable content, yet produce work that remains generic—for instance, approving safe but templated email copy that lacks differentiation and fails to engage the intended audience. Similarly, a marketer may generate multiple technically acceptable campaign variants but lack the evaluative sensitivity required to identify which option is most distinctive or strategically aligned—for example, selecting a headline that is grammatically correct but indistinguishable from competitors, rather than one that reflects clearer audience insight or brand positioning. These distinctions clarify how human expertise contributes under conditions where generative systems expand the scale and speed of content production. In AI-assisted workflows, the ease of generating plausible outputs can obscure weak judgement (e.g., unverified or non-compliant claims), weak voice (e.g., impersonal or formulaic tone), or weak taste (e.g., failure to differentiate among viable alternatives).

Alternative framings were considered during the analytical process. Practitioner accounts could plausibly have been organised around broad capability taxonomies such as creativity, critical thinking, and ethical reasoning, or around functional marketing stages such as strategy, execution, and evaluation. However, these alternatives proved less explanatory under AI-mediated conditions. Capability taxonomies capture important cognitive attributes but do not differentiate the specific forms of agency practitioners described when working with generative systems. Stage-based models describe sequences of activity but do not explain how responsibility is exercised once AI can generate viable outputs at scale. By contrast, the Voice–Judgement–Taste framework aligns more closely with how practitioners articulated responsibility in practice—through authorship, decision authority, and comparative evaluation. While alternative framings remained analytically viable, the framework was retained as the most explanatory of several plausible interpretations, as it provides the clearest account of how practitioners organise and defend non-delegable human responsibility in AI-accelerated marketing work. Table 3 synthesises these analytical boundaries while preserving their interdependence.

OPEN IN VIEWER

Table 3.

Conceptual Differentiation of Voice, Judgement, and Taste.

Construct	Core role	Conceptual description	Distinguishing features	Regulates	Mechanism type	Guiding question
Voice	Authorship and identity expression	Voice refers to the human capacity to author meaning through stylistic, tonal, and identity-based expression. It reflects the ability to produce communication that signals organisational or personal perspective, narrative coherence, and relational authenticity. Voice operates through expressive authorship, shaping how ideas are communicated rather than determining which ideas should be selected or evaluated.	• Concerned with expression rather than evaluation• Anchored in identity, perspective, and narrative positioning• Governs how communication feels and signals ownership	Meaning expression	Semiotic/linguistic	Who is speaking, and how is meaning being expressed?
Judgement	Decision Authority and Normative Governance	Judgement refers to the human capacity to determine whether and how AI-generated or human-generated outputs should be used. It involves contextual evaluation, ethical consideration, strategic alignment, and accountability for decision outcomes. Judgement governs the legitimacy and appropriateness of actions, particularly under conditions of uncertainty and technological delegation.	• Concerned with appropriateness and responsibility• Anchored in decision authority and accountability• Evaluates alignment with goals, ethics, and context	Action permission	Normative/governance	Should this be done, used, or trusted?
Taste	Resonance and Distinctiveness Discernment	Taste refers to the human capacity to recognise and justify what makes communication compelling, distinctive, or resonant within specific cultural, audience, and marketplace contexts. It involves evaluative discernment of symbolic meaning, experiential quality, and differentiation under conditions of content abundance. Taste governs why certain ideas or executions feel impactful, persuasive, or memorable.	• Concerned with resonance rather than correctness• Anchored in cultural, experiential, and symbolic interpretation• Evaluates distinctiveness and communicative effectiveness	Option selection	Evaluative/curatorial	Why does this work, connect, or stand out?

What to Assess: From Outputs to Defensible Processes

Rather than evaluating artefacts in isolation, practitioners described three observable dimensions of contribution:

Table 5 translates these dimensions into rubric-ready indicators to support calibration and reliability. Table 6 then translates these indicators into concrete learning activities and assessment tasks, showing how voice, judgement, and taste can be made observable and evaluable in practice.

OPEN IN VIEWER

Table 5.

VJT Definitions and Indicators.

Capability	Definition	Observable indicators
Voice	The capacity to convey identity, intent, and relational stance in ways that align with a brand, audience, and communicative situation.	Idiom control; on-brand syntax; narrative shaping; alignment with audience register; tonal consistency.
Judgement	Context-aware, ethical, and strategic decision-making that governs when, how, and why to use AI, ensuring outputs are accurate, appropriate, and defensible.	Workflow placement decisions; risk identification; verification steps; escalation behaviour; contextual rationale.
Taste	The ability to discern quality and distinctiveness, selecting and refining the most resonant options from among multiple acceptable AI variants while maintaining cross-asset coherence.	Criteria articulation; distinctiveness thresholding; cross-touchpoint coherence; selection and refinement of AI variants.

OPEN IN VIEWER

Table 6.

Translating VJT Learning Activities Into Observable Teaching and Assessment Practices.

Learning activity	VJT dimension	What this means in practice	What students do	Evidence instructors look for
Stage 1: Foundational drills
Indicator fluency	Integrated (V/J/T)	Recognising what “good” voice, judgement, and taste look like in practice	Evaluate sample outputs using VJT criteria to identify strengths and weaknesses	Accurate identification of tone issues, risks, and quality differences; appropriate use of VJT terminology and criteria
Before–after voice repairs	Voice	Revising an AI-generated draft to improve tone, identity, and audience alignment	Compare an AI-generated version with a revised version and explain what was changed and why	Clear improvement in tone and audience fit; removal of generic phrasing; explicit explanation of how revisions restore identity, stance, or intent
Rapid taste selections	Taste	Choosing the strongest option from multiple AI-generated outputs	Select one option (e.g., headline, caption, concept) from several alternatives and justify the choice	Clear comparative reasoning; justification of distinctiveness, relevance, or audience alignment; ability to explain why one option is stronger than others
Micro judgement logs	Judgement	Short records documenting how decisions about AI outputs were made	Briefly document sources checked, prompt strategy, risks identified, and why the final version was selected	Evidence of fact-checking, risk awareness, contextual adaptation, and clear reasoning for accepting, adapting, or rejecting content
Stage 2: Applied briefs
Brand-register rewrites	Voice	Adapting content to match a specific brand or communication style	Rewrite generic AI output using a provided brand voice guide or exemplar	Alignment with brand tone, vocabulary, and level of formality; consistency with brand or communicative guidelines
Channel-specific microcopy	Voice (with Judgement support)	Adapting tone, structure, and messaging to fit specific platforms and formats	Rewrite or generate content for different channels (e.g., social caption, email subject line, UX microcopy) based on the same brief	Appropriate tone and format for each channel; awareness of platform conventions; consistency of underlying message across formats
Cultural-fit checks	Judgement	Ensuring outputs are appropriate for a specific audience, market, or context	Review AI outputs and adapt or flag issues (e.g., tone, phrasing, assumptions) for a given audience or region	Awareness of cultural/contextual misalignment; appropriate localisation or adaptation; identification of potential risks
Workflow placement and risk mapping	Judgement	Deciding where and how AI should be used within a task and identifying associated risks	Map a workflow showing where AI is used, where human intervention is required, and what risks (e.g., factual, ethical, reputational) arise at each stage	Clear identification of decision points; awareness of risks and mitigation strategies; justification of where AI is appropriate or inappropriate
Micro style guide development	Voice (with Taste support)	Making implicit expectations about tone, style, and positioning explicit	Create a short guide (e.g., tone, vocabulary, do/don’t examples) based on a brief, persona, or brand	Clear articulation of stylistic rules; usability of the guide for consistent application; alignment with audience and purpose
Stage 3: Integrated/capstone
Cross-touchpoint coherence activity	Taste (with Voice support)	Ensuring consistency and intentionality across multiple campaign elements	Align several outputs (e.g., social post, email, landing page) so they form a coherent campaign	Consistent tone, message, and positioning; evidence of coordination across assets; sense of intentional integration
Multichannel campaign project	Integrated (V/J/T)	Developing a coordinated set of outputs across multiple touchpoints with consistent voice, judgement, and taste	Produce a campaign including multiple assets (e.g., social, email, landing page) with supporting rationale for decisions	Coherence across touchpoints; alignment with audience and purpose; evidence of integrated decision-making across voice, judgement, and taste
Judgement checkpoints (AI-restricted tasks)	Judgement	Verifying independent reasoning without AI support at key points	Complete parts of a task without AI (e.g., justification, critique, or selection decisions)	Ability to explain and defend decisions independently; evidence of reasoning not reliant on generated content
Portfolio-based integration (VJT portfolio)	Integrated (V/J/T)	Demonstrating how voice, judgement, and taste operate together over time	Submit a portfolio including drafts, revisions, logs, and final outputs across tasks	Clear links between decisions and outcomes; evidence of development over time; coherence across artefacts; defensible and well-articulated reasoning
Reflective VJT synthesis memo	Integrated (V/J/T)	Articulating how voice, judgement, and taste informed decisions and developed over time	Write a structured reflection explaining key decisions, trade-offs, and how outputs improved across iterations	Clear linkage between decisions and outcomes; evidence of critical reflection; articulation of learning and capability development across VJT dimensions

Surfacing Voice: Assessing Expressive Authorship

Practitioners positioned voice as most visible in revision rather than initial drafting. While AI may accelerate production, identity becomes legible through deliberate human re-authoring (Tan et al., 2025). Instructionally, this translates into compact evidence bundles—an AI draft, a tracked human revision, and a concise rationale linking specific linguistic, narrative, and, where relevant, multimodal choices to a declared authorial stance, persona, strategic position, campaign voice, or organisational identity.

In practice, participants described making voice explicit through conventions and reverse-engineering exercises. Liam described teaching junior marketers’ narrative stance as role encoding—“blog articles are written in the first person, biographies in the third”—and emphasised shaping reader movement as a “journey” rather than presenting “lots and lots of facts.” Others described training junior marketers to analyse communicative exemplars for characteristic syntax, idioms, humour registers, visual cues, and formality thresholds, then reapply these patterns when revising AI drafts. In brand-based tasks, this may involve tone-of-voice guides and brand assets; in non-brand tasks, it may involve declared audience, genre, persona, or strategic intent. Jack framed this as “eradicating” generic phrasing to restore organisational specificity.

Several practitioners cautioned that beginning with AI can shortcut immersion in idea development, reinforcing the need for structured revision practices that foreground identity restoration and intentional stance-taking. Collectively, these approaches position voice as observable alignment between communicative choices and a declared identity, stance, and audience relationship (Tan et al., 2025), consistent with research conceptualising voice as self-representation constructed through textual and contextual cues (Tardy, 2012).

Making Judgement Auditable: Evaluating Decision Authority

Judgement was framed as accountable deployment authority: when outputs can be generated instantly, professional responsibility rests on whether someone can defend their use. Practitioners emphasised active interrogation and contextual verification, highlighting hallucinations and contextual misalignment as routine risks. Warren described reviewing materials assembled through “Google and AI and YouTube,” requiring students to explain the reasoning behind their inclusion, while Elliott resisted what he termed a “spinner in 10 seconds,” warning against organisational pressure to treat instant output as publication-ready.

To make judgement assessable, participants advocated lightweight documentation rather than exhaustive process capture. A concise judgement log—recording sources consulted, prompt strategy, verification steps, risks identified, and rationale for final selection—was commonly described as sufficient. As Jonathan put it, AI is “amplified intelligence” that “still needs to be guarded by humans.” In curricular terms, evaluation therefore extends beyond output quality to the defensibility of deployment decisions under AI acceleration.

Cultivating and Operationalising Taste: Criteria-Based Selection Under Abundance

As discussed in RQ1, practitioners described taste as structured evaluative selection under conditions of output abundance, enacted through curation, originality sensing, and coherence construction. Practitioners emphasised the need to externalise tacit quality expectations into explicit criteria. Warren described building a “codified system of goodness versus badness,” while Chloe warned that without developed taste, practitioners risk allowing the tool to dictate decisions rather than directing it themselves.

Instructionally, this suggests designing studio-style tasks in which students generate variants, apply articulated quality criteria, justify comparative selection decisions, and synthesise a coherent execution. The assessable unit is the reasoning underpinning selection as well as the quality of the chosen output.

Ethical and Cultural Fit as Routine Checks

Across interviews, voice, judgement, and taste were linked to reputational exposure and trust. Assessment criteria therefore extend beyond creative quality to include cultural sensitivity, regulatory awareness, and anticipatory risk evaluation. Embedding these checks within routine coursework aligns evaluation with professional accountability expectations. At the same time, briefs and brand guidelines are themselves culturally situated artefacts rather than neutral evaluation anchors. Research on taste and valuation shows that evaluative criteria are embedded within socially organised systems of meaning that shape how quality and appropriateness are recognised and legitimised (Arsel & Bean, 2013; Smith Maguire, 2018; Warde, 2014). In diverse classrooms, instructors may therefore need to surface these assumptions and encourage discussion of how expectations vary across audiences, markets, and communicative traditions.

Workflow-Aligned Learning Progression and Implementation Constraints

Interviewees proposed a staged, learn-by-doing progression that develops voice, judgement, and taste through iterative practice while maintaining manageable evidence requirements. Figure 1 translates these practitioner insights into a curriculum model that progressively increases contextual complexity and accountability. Stage 1 develops indicator fluency through low-stakes drills that isolate core VJT practices (voice revision, judgement logging, and basic taste comparisons). Stage 2 introduces applied client-style briefs in which students must justify decisions within defined brand and audience constraints, making judgement and taste criteria explicit and documented. Stage 3 shifts the emphasis from task execution to accountable integration: students produce a portfolio demonstrating cross-touchpoint coherence while defending how voice, judgement, and taste informed campaign-level decisions. The progression therefore moves from skill recognition, to situated application, to integrated professional accountability, enabling instructors to scaffold increasingly defensible decision-making under AI-assisted production.

OPEN IN VIEWER

Figure 1.

Practitioner-informed curriculum progression for developing voice, judgement, and taste in AI-mediated marketing work.

Practitioners cautioned against over-engineered documentation that attempts to capture every procedural step. As Warren observed, “stage one, stage two, stage three. . . doesn’t scale,” reinforcing a preference for concise rationales that surface decision logic without imposing unsustainable grading burdens. Participants also distinguished between surfacing capability and manufacturing dispositions: as Noah noted, curiosity “has to come from the self,” suggesting assessment should evidence inquiry-in-action rather than attempt to grade personality traits.

To address integrity risks in AI-assisted coursework, practitioners recommended periodic AI-restricted checkpoints. As Eve described, examinations show that “it’s just you and a piece of paper,” functioning as complementary verification of independent reasoning alongside AI-permitted process evidence. Together, these design principles frame VJT as a scalable assessment architecture aligned with professional expectations of defensible judgement and discernment.