Progressive
personalized genAI creative at scale (scripts, voice, music) optimized by segment
On an audio campaign, Progressive had generative AI produce 96 variants (scripts, voice, music) approved in 14 days, optimized by segment with Claritas, for a 31% gain in quote starts and a 197% lift over the control group.
Objective
Produce a large number of personalized audio variants in a few days instead of several weeks, then deliver only the ones that perform, to pull up quote starts at lower cost and lead time.
The deployment
Progressive handed generative AI the writing of the scripts and the creation of the voices and music beds for an audio campaign, then let an optimization engine sort the variants by their real performance. The team had 96 variants approved in 14 days, versus a handful of spots produced over about seven weeks with the classic method. The variants were delivered in streaming (SiriusXM, Spotify, iHeart, and others), then activated and optimized by segment through the Claritas tool. Progressive and Claritas report a 31% gain in quote starts, a 197% lift over the control group, and 98% listen-through to the end. The operational lesson comes down to one sentence from the Claritas side: 96 versions was perhaps too many, a sign that the volume of creatives should stay driven by measurement rather than maximized for its own sake.
Results Proof C
Campaign reported by name in the marketing press (The Drum) with quotes from the Progressive lead and the Claritas executive, corroborated by specialist audio press on the volume of variants and the effect on quote starts. Figures declared by both parties to the campaign, not an audited financial result.
How it works
Documented architectureThe stack in detail
- plateforme Claritas AI Creative Optimization Learning-based creative optimization: activating audio variants by segment and stopping the weakest performers.
- llm Outils genAI texte, voix et musique Generated the scripts, synthetic voices, and music beds for the 96 variants; the exact tools are not publicly named.
- plateforme SiriusXM Delivery of the variants in audio streaming.
- plateforme Spotify Delivery of the variants in audio streaming with targeting by segment.
- plateforme iHeartMedia Delivery of the variants in audio streaming.
How it runs, concretely
For ops teams-
1Generating the variants AI and marketing
The AI writes the scripts, generates the voices and music beds; the team gets the selected variants approved.
-
2Multi-platform delivery marketing and agency
The variants go out in audio streaming (SiriusXM, Spotify, iHeart) with targeting by segment.
-
3Optimization by segment AI and agency
The Claritas engine activates and prioritizes variants by their real performance and cuts the least effective.
Quote starts and engagement per variant. If the per-segment performance signal is missing, you no longer know which variants to cut and the volume of creatives becomes noise.
How your customers perceive this type of use
Sourced studiesLe paradoxe est documente des deux cotes : 71% des consommateurs attendent des interactions personnalisees et 76% sont frustres quand elles manquent (McKinsey, 2021), mais 75% declarent ne pas acheter aupres d'organisations auxquelles ils ne confient pas leurs donnees (Cisco, 2024). La « creepy line » est localisee : messages recus quelques secondes apres une recherche et suivi de localisation sont les pratiques qui mettent le plus mal a l'aise (Periscope by McKinsey, 2019).
Acceptance conditions
- La confiance dans le traitement des donnees precede l'achat : 75% ne achetent pas sans elle (Cisco 2024)
- Un cadre legal protecteur rassure : 59% des consommateurs disent que des lois fortes sur la vie privee les rendent plus a l'aise pour partager des informations dans des applications IA (Cisco 2024)
- La personnalisation elle-meme est attendue quand elle est consentie : environ la moitie des consommateurs (US 55%, UK 52%) disent s'inscrire souvent ou parfois a des services personnalises (Periscope by McKinsey 2019)
Red lines
- Le message declenche quelques secondes apres une recherche ou un achat : deuxieme ou troisieme cause de malaise selon les pays (Periscope by McKinsey 2019)
- Le suivi de localisation percu comme de la surveillance : 40% de malaise en Allemagne et au Royaume-Uni (Periscope by McKinsey 2019)
- Le mesusage des donnees personnelles par l'IA, devenu la premiere inquietude des consommateurs, a 53% et en hausse (Qualtrics 2025)
Sources: McKinsey & Company 2021 · Periscope by McKinsey 2019 · Cisco 2024 · Qualtrics 2025
How to replicate
Inference, not sourcedData prerequisites
- Usable audience segments
- Reliable measurement of quote starts per variant
- Rights to the brand voice or a licensed synthetic voice
Org prerequisites
- Accelerated creative validation chain
- Media optimization partner or tool
- Brand rules to frame genAI
Possible stack
- GenAI text/voice/music tools + creative optimization platform (the Claritas route)
- Audio DSP with dynamic creative
The plan, step by step
- Step 1Secure voice rights (licensed synthetic or brand voice), set the brand rules for genAI, and connect quote-start measurement per variant.Deliverable: Legal framework and per-variant measurement operational.
- Step 2GenAI production sprint (scripts, voice, music) and validation through the accelerated creative chain.Deliverable: Batch of 10 to 20 approved audio variants for the pilot.
- Step 3Deliver as a test with a control group on a streaming channel, targeting by segment.Deliverable: Live pilot campaign with per-segment performance read.
- Step 4Cut the weak variants, measure the lift vs control, and decide the next production volume (volume stays driven by measurement).Deliverable: Lift vs control review and scale plan.
First step: Launch an A/B test of a few genAI variants on an audio channel and connect quote-start measurement before scaling up the volume.
The same pattern, proven in other industries
Sources
- S1 Progressive Insurance tests limits of AI-generated ads - and learns when to pull back Established press archive pending
- S2 How AI Is Changing the Audio Ad Market Secondary archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.