How AI Is Reinventing Advertising and the Future of Brand Creativity | Robin Taconet x Shekhar Natrajan

Overview

This episode of Tomorrow, Today features Robin Tauquedet, founder of Taximetry, discussing how AI can help brands produce advertising assets at scale. His company addresses creative fatigue—the decline in an ad’s effectiveness after audiences encounter it repeatedly—by generating and testing numerous on-brand variations.

The conversation moves beyond production efficiency to examine AI-generated “slop,” authenticity, fraud, disclosure requirements, and the limits of synthetic-content identification. Tauquedet argues that verifier agents, human review, regulation, and public education must work together because technology alone cannot reliably contain misuse.

Host Shekhar Natarajan repeatedly challenges the assumption that greater personalization is inherently beneficial. He asks whether optimization can dilute brand identity, weaken human agency, worsen attention problems, and substitute activity metrics such as token consumption for meaningful business outcomes.

The episode concludes with a broader assessment of AI’s hype cycle. Both speakers question whether current adoption has produced enough measurable value, while identifying medicine, cross-organizational coordination, and robotics as areas where AI may eventually deliver more profound benefits.

Segmented Summary

[01:11] Introducing Taximetry and the Creative-Fatigue Problem

[Fact] Tauquedet describes his engineering background and experience at Salesforce, MongoDB, and Meta.

[Fact] He founded Taximetry to help enterprises and brands produce enough creative assets to address creative fatigue.

[Fact] Advertising platforms want numerous assets for experimentation, while audiences become less responsive when repeatedly shown the same material.

[Fact] Producing conventional campaigns at this pace is costly because it can require models, photographers, influencers, and substantial coordination.

[06:31] Customers, Brand Assets, and Distribution Platforms

[Fact] Taximetry works with companies of different sizes, although most current customers are smaller jewelry and apparel brands with limited budgets.

[Fact] The company also works with agencies, providing asset creation while partners retain responsibility for branding or customer relationships.

[Fact] Its stated objective is to create high-quality assets at scale while respecting each brand’s identity.

[Fact] Meta, Google, and other advertising channels remain responsible for distribution; Taximetry aims to remove friction in supplying those channels with more creative material.

[08:18] Novelty, Attention, and the Economics of Social Media

[Fact] Tauquedet attributes creative fatigue partly to the gap between limited human attention and the rapidly growing volume of information.

[Fact] He argues that new formats can regain attention through novelty when familiar ads no longer produce a response.

[Fact] Natarajan questions whether generating more content merely intensifies information overload, polarization, and addictive scrolling.

[Fact] Tauquedet distinguishes organic content from paid advertising and argues that ads finance free social platforms while helping consumers discover relevant brands.

[Fact] Because platforms increasingly control audience targeting, he identifies the creative asset as one of the last major conversion levers available to a brand.

[13:00] A/B Testing and Campaign Measurement

[Fact] Taximetry uses experimentation and A/B testing, with the appropriate volume and performance metrics varying by customer and industry.

[Fact] Tauquedet says jewelry advertisers may need approximately 50–70 asset variations each week and may wait about two weeks for useful performance evidence.

[Fact] New brands may optimize for click-through rate and awareness, while established brands may prioritize return on investment and conversion.

[Fact] Campaign results are fed back into the system to guide the next set of assets and compare performance with industry benchmarks.

[15:39] Controlling AI-Generated “Slop”

[Fact] Natarajan describes AI-assisted brand management as frightening because generated images, voices, and videos can be low-quality or inconsistent with the brand.

[Fact] Tauquedet divides creative work into research, strategy, asset production, and performance feedback, with AI usable at each stage.

[Fact] Brands can choose different levels of AI involvement; some permit AI in research, selection, and analysis but reject fully generated public-facing assets.

[Fact] For brands with extensive existing libraries, AI can select, crop, and combine human-produced material rather than generate replacements.

[17:25] AI Disclosure and Regulatory Ambiguity

[Fact] Tauquedet says some European rules require AI-generated media to be identified, but partially AI-assisted workflows create unresolved gray areas.

[Fact] His company currently labels AI-generated voice, music, images, and video, but does not apply the same label when AI is confined to research or strategy.

[Fact] He says the regulatory landscape is changing quickly and that the company must adapt its practices accordingly.

[Fact] The discussion contrasts AI generation with older forms of image manipulation, while acknowledging that AI makes synthetic production dramatically easier to scale.

[22:29] Synthetic Media, Fraud, and Fact Versus Fiction

[Fact] Natarajan cites examples of manipulated product evidence, fabricated medical documents, insurance fraud, cloned voices, and impersonation scams.

[Fact] The speakers discuss digital signatures or synthetic identifiers that could mark generated content.

[Fact] Natarajan argues that these mechanisms have limited value unless upstream and downstream systems share standards and can preserve and validate the identifiers.

[Fact] He frames synthetic content as capable of creating both a false economy and a fraud economy when representations cannot be trusted.

[26:25] Verifier Agents, Human Review, and Public Literacy

[Fact] Taximetry uses verifier agents to detect visual defects such as malformed hands or extra limbs.

[Fact] Human reviewers remain in the loop, particularly before an asset is released as a representation of a brand.

[Fact] Tauquedet separates quality control for legitimate production from the harder problem of detecting deliberate fraud.

[Fact] He argues that fraud prevention requires regulation, education, public exposure to AI risks, and cybersecurity tools.

[Fact] He describes cybersecurity as a continuing cat-and-mouse contest and says the human being remains the most critical link in AI-enabled defense.

[31:00] Scaling Human Oversight and Defending the Startup

[Fact] Natarajan observes that people may surrender agency when asked to approve thousands of generated assets.

[Fact] Tauquedet proposes dividing reviews among people and placing control points earlier in the production process, while acknowledging that human oversight may eventually diminish.

[Fact] Taximetry orchestrates multiple agents and models rather than depending on a single model provider.

[Fact] Its workflow may use different systems for video, voice, images, reasoning, and storytelling, supplemented by skills and prompt engineering.

[Fact] Tauquedet acknowledges that agents will likely assume more prompt-generation work and that foundation-model providers remain a competitive threat.

[35:15] Business Models and the Risk of Creative Commoditization

[Fact] Taximetry offers fully delegated creative services priced similarly to an agency, licensed platforms with usage credits for marketing teams, and white-label services for agencies.

[Fact] Natarajan asks whether serving competing companies in the same industry could make their advertising increasingly uniform.

[Fact] Tauquedet says the system ingests each customer’s existing assets, voice, color palette, history, and other elements of its brand DNA.

[Fact] He argues that outputs should differ unless two brands have very similar identities and objectives.

[37:49] Personalization Versus Shared Brand Identity

[Fact] Natarajan argues that brands derive power from shared cultural experiences, using widely discussed Super Bowl commercials as an example.

[Fact] He questions whether individual-level personalization and repeated optimization could gradually transform a brand into “everything to everyone.”

[Fact] Tauquedet accepts that performance feedback could pull a brand away from its identity and says guardrails may be necessary.

[Fact] Taximetry currently keeps its campaign-performance feedback loop separate from the brand hub, so a failed campaign does not automatically rewrite the underlying brand.

[Inference] The exchange identifies brand consistency as a governance problem, not merely a generation-quality problem.

[42:10] Distribution Power and Human-Centered Metrics

[Fact] Natarajan argues that distribution may offer a larger opportunity than asset generation because broader reach strongly affects campaign success.

[Fact] Tauquedet agrees that distribution contains substantial economic value but says directly entering it would mean competing with major platforms.

[Fact] Taximetry is initially focusing on creative assets while leaving open the possibility of expanding into distribution later.

[Fact] Natarajan proposes evaluating advertising with “human” or psychological metrics, not only clicks, conversions, and return on investment.

[Fact] Tauquedet says the company is currently prioritizing evidence of scalable business value and expects ethical metrics to become more practical after it accumulates customers and performance history.

[Inference] Their disagreement centers on whether human safeguards should be designed from the beginning or added after commercial viability is established.

[49:27] AI Hype, Token Maxing, and Outcome Measurement

[Fact] Natarajan says many established brands remain at a very low level of AI maturity and sometimes use AI only for basic functions such as translation.

[Fact] Tauquedet agrees that AI is hyped and expects early adopters to enter a period of disillusionment as they question the returns on rising expenditure.

[Fact] The speakers criticize “token maxing,” in which organizations celebrate high token consumption without demonstrating corresponding outcomes.

[Fact] They discuss circular investment incentives in which companies invest in AI providers while also purchasing their products, making consumption beneficial to both parties’ reported growth.

[Fact] Tauquedet says organizations may nevertheless encourage broad experimentation to discover valuable use cases and ultimately replace labor costs.

[53:20] Efficiency, Labor, and More Transformative Uses of AI

[Fact] Both speakers agree that business outcomes are harder to measure than token consumption.

[Fact] Natarajan argues that fear of appearing technologically behind may encourage executives to exaggerate their AI adoption.

[Fact] He distinguishes simple cost reduction from business enablement and says AI’s deeper value may lie in creating coordination across companies and industries.

[Fact] He questions labor-replacement logic for very small businesses that already operate with only one or two people.

[Fact] Medicine is presented as a strong positive example, with AlphaFold and drug discovery illustrating how AI can combine biological, chemical, and mathematical knowledge.

[Inference] The discussion favors AI that reorganizes capabilities or enables new discoveries over AI justified solely by headcount reduction.

[60:05] A Five-Year Outlook for AI

[Fact] Tauquedet hopes Taximetry will remain viable and serve many customers five years from now.

[Fact] He expects models to become substantially more capable while governments impose more constraints and guardrails.

[Fact] He anticipates possible synergies between AI and increased computing capacity, including developments associated with quantum technology.

[Fact] He identifies robotics as a likely next implementation wave, with potential applications in elder care, physical services, and reducing burdensome labor.

[Fact] He also expects greater use of AI in defense, misinformation, cyberattacks, and psychological manipulation.

[Fact] Drawing an analogy with jobs created after running water displaced water carriers, he predicts that AI will ultimately be a net positive despite disruption.

Podcast Review/Summary

The episode’s strongest feature is its willingness to challenge the standard AI-advertising narrative. Rather than treating higher production volume as an unquestioned success, it asks who benefits, how authenticity is protected, and whether personalization undermines the shared identity that makes a brand culturally meaningful.

It also provides a practical view of AI creative operations: campaign metrics, feedback loops, selectable degrees of automation, verifier agents, human approval, multi-model orchestration, and several possible business models. The recurring distinction between generating assets and governing their use is particularly valuable.

Its main limitation is that many commercial and technical claims remain qualitative. The discussion offers few verified performance figures, comparative results, or detailed evidence showing that AI-generated campaigns outperform conventional production after accounting for cost, brand risk, and distribution.

[Inference] The episode is best suited to founders, marketers, agency leaders, brand managers, and AI-governance professionals who want a critical introduction to both the commercial promise and the unresolved human consequences of automated advertising.