Marketing teams and UX researchers face mountains of conversational data (interviews, focus groups, campaign comments, and creative tests) and no consistent way to turn it into decisions. The Wharton podcast series Where AI Works (see full series at www.knowledge.wharton.upenn.edu/where-ai-works-show/) surfaces three practical realities: (1) early AI ROI shows up in conversion and engagement, not just cost cuts; (2) brand-specific models and human-in-the-loop processes reduce hallmarks of “sameness”; (3) agentic systems and scale require data & change management. In this post we translate those signals into a repeatable ai qualitative analysis workflow you can run in two weeks with Evidano (www.evidano.com), from ingest to themes, cross-segment comparisons, and stakeholder-ready evidence.
Fast Take: Why the Wharton Season Matters for Qual Research
Wharton’s Where AI Works (season published August 25, 2025) collects practitioner case studies from Mondelez, Accenture and creative leaders that show tangible adoption patterns for marketing and creative teams.
- Mondelez reports ~40 campaign AI activations over two years with measurable lifts in awareness and revenue.
- Accenture reports $5.6B in generative-AI revenue over the prior 18 months and >2, 000 gen/agentic-AI projects (Lan Guan).
- Common threads: brand-specific models, human-in-the-loop review, and platform-first engineering to industrialize use cases.
Read the source series at www.knowledge.wharton.upenn.edu/where-ai-works-show/.
Findings Snapshot
| Date / Season | Episode / Guest | Signal (metric / quote) | Implication for Qual Research |
|---|---|---|---|
| 2025 (Season) | Jonathan Halvorson (Mondelez) | 40 campaign AI activations; top-performing Cadbury and localised creative programs | AI enables personalized, scalable creative; need to analyze creative feedback at scale (n=millions) |
| 2025 | Jill Kramer (Accenture) | 1, 000 marketers onboarded with agents; 30–50% speed-to-market gains; ~35% reduction in steps | Workflows shift from linear to agentic, qualitative pipelines must support iterative, autonomous follow-ups |
| 2025 | Lan Guan (Accenture) | $5.6B gen-AI bookings over 18 months; 2, 000 projects; 86% of C-suite plan to increase AI spend | Enterprise adoption is real but only ~8% have fully scaled, snapshot your data and talent readiness first |
| Academic signals | MIT / IIM; Univ. of Hamburg / Emory (cited) | 6–9% ad CTR lift; ~0.75 point lift on 7‑point purchase-funnel scores | Quantitative signals complement thematic analysis; combine theme prevalence with engagement metrics |
How It Works: From Podcast Signals to Research Questions
The Wharton episodes distill three operational patterns useful for qualitative teams:
- Start small, industrialize common building blocks: vectorized archives, brand playbooks, and a single-source-of-truth SOP for language assets.
- Human-in-the-loop is not optional: teams use AI to draft briefs, generate concepts, and then iterate via creative directors and rapid testing.
- Agentic workflows change the cadence: autonomous agents can run parallel creative tests and trigger follow-up interviews, so qualitative pipelines must support iterative re-ingestion and re-coding.
Methodologically, convert those patterns into research questions you can answer with ai qualitative analysis: Which creative variants lift attention? How do segments talk about brand distinctiveness? Which phrases predict higher CTR or conversion?
Implications for Researchers: ai qualitative analysis in marketing
For UX / Qualitative Researchers
Use thematic + frequency analysis to detect whether AI-created creative reduces idea diversity (Wharton research shows AI increases quality but can lower diversity).
Design cross-segment comparisons (by region, persona, product) to surface divergence: not just aggregate themes.
Instrument rapid A/B creative tests with embedded open comments, then run co-occurrence networks to link language patterns to outcomes.
For Product & Marketing Leaders (CMOs / PMs)
Prioritize data & governance first: inconsistent SOPs or multiple playbook versions will break model performance (Lan Guan example).
Measure themes alongside behavioral KPIs (CTR, conversions). Use mixed-methods reports that combine thematic prevalence with ROI signals.
Treat brand distinctiveness as a codified asset: upload playbooks and archives to train brand-specific models that avoid the vanilla-vanity problem described by Halvorson.
For Policy, Privacy & Research Ops
Expect change-management costs: interviews in the series emphasize demos, cohorts, and repeated communications as the dominant adoption levers.
Plan PII redaction and consent workflows before ingesting enterprise customer data.
Demand vendor transparency on training data and model governance when choosing tools or partners.
Do More, Faster with Evidano
Problem: Fragmented transcripts, loose codebooks
Solution: Evidano ingests interview transcripts, survey spreadsheets and campaign comments, applies consistent AI-assisted coding, and exports hierarchical code → subcode visualizations for stakeholder review.
Problem: Brand-specific nuance and multilingual inputs
Solution: Upload brand playbooks and use Evidano custom dictionary + translation to train consistent labels; retain human feedback loops to refine model outputs.
Problem: Measuring theme impact on outcomes
Solution: Link thematic frequency and co-occurrence outputs to campaign metrics (CTR, conversions) for combined thematic + quantitative analysis in one dashboard.
Problem: Need follow-ups at scale
Solution: Use Evidano AI avatar interviewers to run autonomous follow-ups, then auto-ingest responses and update themes, all encrypted and never used to train third‑party models.
Two‑Week Pilot Checklist: Reproducible ai qualitative analysis
Run this short pilot to validate value quickly and in a research-safe way:
- 1) Collect inputs: 20–50 interview transcripts, 2–3 creative test comment datasets, and 1 spreadsheet of survey responses.
- 2) Prep: Upload brand playbooks / SOPs; enable PII redaction and custom dictionary.
- 3) Transcribe & translate (if needed) with Evidano automatic pipeline.
- 4) Auto-code + import a seed codebook; run AI-assisted refinement with human reviewers (2 rounds).
- 5) Produce thematic, frequency and cross-segment analyses; create co-occurrence networks and top quote exports.
- 6) Link themes to metrics (CTR, conversion lift) and produce a 2-page decision memo with top 3 actions.
- 7) If gaps appear, deploy AI avatar interviewer for targeted follow-ups and re-run step 4–6.
Expected outputs in 2 weeks: clean codebook, thematic dashboard, representative quotes, and a stakeholder slide set tied to business outcomes.
Wrapping Up & Next Steps
Wharton’s Where AI Works makes the practical case: AI already moves metrics for marketing, but the real win for researchers is coupling rigorous qualitative workflows with brand-specific modeling and disciplined change management.
If you want to pilot ai qualitative analysis on a live corpus (interviews, transcripts, survey responses or campaign comments), Evidano can run the two‑week checklist above and deliver a stakeholder-ready synthesis. Start with a focused objective (e.g., why did ads in region X out-perform region Y?) and we’ll help map inputs → themes → actions.
Ready to try a pilot? Learn more or schedule a demo at www.evidano.com.
Keep reading
- Commentary on NewsTwo Definitions: Climate Change Acceptance for UndergradsHow a PLoS One Delphi study (Aug 25, 2026) defined climate change acceptance for undergraduate science students, and how AI-enabled qualitative analysis applies it.
- Commentary on NewsResearcher-in-the-loop: AI-enabled UX researchHow the researcher-in-the-loop model governs AI-enabled UX research. Learn practical governance, stats from the August 2026 piece, and how Evidano supports this workflow.
- Commentary on NewsResearcher-in-the-Loop: Governance for AI UX ResearchGovern AI in qualitative UX research with the researcher-in-the-loop model from Jennifer L. Bowie (Aug 25, 2026): practical rules, risks, and tool mappings.