Marketing Discussions: July 14th, 2026
Ask an AI model for campaign ideas and receive fifty before lunch. Sure, a few are clever. Most are competent. None are obviously unusable. The team asks for fifty more, then requests alternate headlines, visual concepts, audience variations, landing-page structures, and social executions. By the end of the day, everyone has more material than they can properly review.
The next morning, the creative meeting lasts two hours because nobody knows which direction deserves to survive.
This is becoming a familiar AI marketing problem. Generation became faster than judgment.
Marketing teams once lost time while trying to produce enough viable options. Writers stared at empty documents. Designers waited for approved copy. Product marketers spent days creating the first coherent version of a campaign narrative. AI has reduced much of that initial friction.
The constraint has moved.
Someone still needs to recognize which idea understands the customer, which claim reflects the product, which expression sounds like the company, and which campaign will remain interesting after the novelty wears off. Someone needs to reject the plausible work that has no reason to exist.
Taste is becoming the marketing bottleneck.
What taste means in marketing
Taste in marketing is the ability to select and shape work according to the customer, product, brand, market, and moment.
It is more rigorous than personal preference. A marketer saying “I do not like this color” is expressing preference. A marketer explaining that a visual system makes an enterprise security product appear unserious, resembles three major competitors, and conflicts with the expectations of its buying committee is exercising judgment.
Marketing taste combines several forms of understanding:
- Knowledge of the customer and the language customers actually use
- Familiarity with the category and its repeated conventions
- A clear view of the product’s strengths, limits, and credible promises
- Sensitivity to timing, culture, format, and distribution
- An understanding of the company’s existing identity
- The ability to distinguish polished execution from a meaningful idea
- The confidence to reject work that is acceptable but undistinguished
Taste does not require one person to possess flawless instincts. Strong teams develop shared judgment through research, standards, critique, experimentation, and repeated exposure to good work.
The objective is not to turn every marketer into an elusive creative genius. The objective is to build an organization capable of making better choices than its competitors when both have access to similar tools.
AI expanded the possibility space faster than teams expanded their judgment
Generative tools can create campaign territories, draft copy, develop layouts, summarize customer research, classify feedback, produce images, and adapt approved ideas across channels. That abundance has changed the economics of creative exploration.
It has also created a review problem.
A team that once considered three concepts can now consider thirty. The additional possibilities can increase the chance of finding something exceptional. They can just as easily consume the team’s attention, blur distinctions between directions, and encourage endless generation instead of commitment.
The recent conversation surrounding Cannes Lions reflects this tension. McKinsey identified “creativity versus AI mediocrity” as one of the central signals from the 2026 festival, arguing that the industry has begun moving from AI experimentation toward harder questions of business value, operating models, human creativity, and differentiated growth.
Digitas North America CEO Amy Lanzi made the abundance problem more direct during a Cannes interview with The Verge. She argued that the advertising industry does not need an endless supply of additional content and questioned the platform-led idea that creative should function primarily as another form of automated targeting.
The concern is not that AI-generated work is always bad. Much of it is increasingly fine.
“Fine” is becoming plentiful.
A structurally complete article, plausible product image, reasonable headline, or clean social post no longer demonstrates much operational sophistication on its own. The useful question is whether the asset contains an idea customers can recognize, remember, trust, or act upon.
Competence is becoming less differentiating
For years, companies could outperform weaker competitors by reaching a basic level of executional competence.
They could publish more consistently, improve their landing pages, create clearer advertisements, build better email sequences, test more headlines, and maintain stronger campaign operations. Many companies still need that work. AI can help them reach the baseline more quickly.
The baseline will continue rising.
The average presentation will become cleaner. The average brief will be more complete. The average campaign will include more variations. The average content team will conduct more research and produce more formats with fewer people.
A company cannot assume those improvements will make it distinctive when its competitors receive the same advantages.
The behavior of AI companies themselves illustrates the point. Sensor Tower data reported by Marketing Brew found that U.S. AI brands spent approximately $424 million on advertising during the first quarter of 2026, more than three times the amount spent during the same period in 2025. Anthropic and OpenAI sharply increased spending across channels including television, streaming video, YouTube, and LinkedIn.
These companies possess some of the world’s most advanced technologies for generating and distributing information. They still need familiar brand-building mechanisms because technical capability alone does not guarantee recognition, trust, or preference.
AI can help a company explain more features. Taste helps the company decide which idea it wants people to remember.
More output creates more editorial debt
Marketing teams often discuss content debt as an archive problem. A company publishes hundreds of articles, landing pages, or campaign assets and eventually discovers that much of the material is outdated, duplicative, poorly connected, or inconsistent with the current product.
Generative production can accelerate that debt.
Every new asset creates future obligations:
- Someone may need to update the claims.
- Someone may need to verify that the product still works as described.
- Someone may need to maintain the link, design, tracking, or localization.
- Sales and support teams may need to correct the expectations it creates.
- Search systems and AI tools may continue surfacing it after the company has moved on.
- Customers may encounter contradictory explanations across channels.
A team generating ten times more material can create ten times more opportunities for inconsistency.
The quality-versus-volume trap becomes an information-governance problem as much as an editorial one. The cost of an asset includes the attention required to review, distribute, measure, refresh, and eventually retire it.
Good taste therefore includes the ability to decide that a piece of work should never be published.
Taste begins with customer proximity
Teams cannot judge whether an idea is useful when they have only a synthetic understanding of the audience.
AI can summarize reviews, sales transcripts, support tickets, survey responses, search behavior, and community discussions. These are valuable accelerants. The team should still spend time with the underlying material and, where possible, with the people themselves.
Customer proximity gives marketers access to details that generic generation tends to smooth away:
- The awkward phrase buyers use to describe the problem
- The internal political obstacle preventing a purchase
- The workaround customers have built themselves
- The moment a buyer becomes worried enough to seek a solution
- The feature customers praise that the company barely mentions
- The promise the market has heard too many times
- The objection that sounds irrational until the operating context becomes clear
Specific details create the raw material for distinctive work.
A campaign built around a real customer tension has a stronger center than one built around a demographic label and several generalized pain points. A content brief becomes more useful when it contains actual language, objections, proof requirements, and decision context instead of a fictional persona with a stock-photo name.
AI should help teams inspect more customer evidence. It should not become a substitute for customer evidence.
Taste requires constraints
Infinite possibility is rarely a useful creative environment.
Strong marketing work usually emerges inside defined constraints: one audience, one problem, one product truth, one desired response, one primary channel, and one reason the company has permission to speak.
Constraints make judgment possible.
Compare these two requests:
Create a memorable campaign for our B2B analytics platform.
Create a campaign for demand-generation leaders who receive conflicting attribution reports from paid media, SEO, and RevOps. Demonstrate that our product reconciles the full lead journey without claiming perfect attribution. The campaign should make an analytics leader feel relieved rather than threatened.
The second request gives both the model and the marketing team a field in which useful decisions can occur. The audience, tension, product truth, emotional outcome, and credibility boundary are visible.
Teams using an AI content strategy should document these constraints before generating a large number of outputs. Otherwise the system will optimize for plausibility and surface variation rather than strategic relevance.
A practical taste rubric
Marketing teams can make creative judgment less mysterious by evaluating work against explicit criteria.
Score each serious concept from one to five across the following dimensions.
Customer truth
Does the work reflect a problem, desire, behavior, or tension that the intended audience genuinely experiences?
A low score means the work relies on broad category assumptions. A high score means customers are likely to recognize themselves immediately.
Product truth
Can the company deliver the promise being made?
A strong concept should emphasize something the product, service, experience, or organization can credibly support. Creative exaggeration cannot become a substitute for operational reality.
Distinctiveness
Could a direct competitor publish the asset after changing the logo?
Category language is sometimes necessary. The core idea should still contain a recognizable perspective, proof point, visual choice, or product relationship that belongs to the company.
Memorability
Will the audience retain a sentence, image, term, demonstration, emotional turn, or useful framework?
Memorability does not require spectacle. A precise idea can remain longer than an elaborate execution.
Usefulness
Does the asset help the audience understand, decide, compare, explain, or act?
Useful work earns attention by reducing confusion or creating value before the transaction.
Coherence
Does the concept fit the larger brand, product, and customer experience?
A surprising idea can still be coherent. The audience should not feel that the campaign belongs to a different company from the product.
Channel fit
Does the execution use the strengths of its format?
A strong research article should not become a video script pasted into a blog. A useful executive post should not read like a compressed landing page. Adaptation requires more than shortening.
Business consequence
What should happen if the work succeeds?
The answer might be recognition, qualified traffic, category understanding, sales enablement, product adoption, retention, or direct conversion. The team should know which outcome it is trying to influence.
The rubric does not turn taste into arithmetic. It gives the critique a shared language.
Build critique into the operating model
Many teams use AI during generation and leave the review process unchanged. The result is a larger pile entering the same approval bottleneck.
A better workflow separates four activities.
Divergence
Use AI, customer evidence, competitive research, and human brainstorming to produce several genuinely different territories. Avoid generating twenty variations of the same safe premise.
Selection
Reduce the field using the taste rubric. Ask which territories deserve further investment before polishing every execution.
Development
Let human and machine collaborators improve the surviving ideas. Add proof, product detail, stronger language, visual logic, and channel-specific expression.
Critique
Review the work with people accountable for different truths:
- Marketing protects the customer and strategic objective.
- Product protects accuracy.
- Brand or creative leadership protects identity and distinctiveness.
- Sales protects buying context.
- Legal protects material risk.
- A clearly named owner makes the final decision.
Consensus should inform the decision without becoming the decision. A campaign assembled from everyone’s least objectionable preference will usually arrive smooth, safe, and forgettable.
Know which work to automate
Marketing teams should automate work where speed and consistency create leverage:
- Transcript processing
- Research organization
- Feedback classification
- Versioning and resizing
- Initial outlines
- Metadata
- Formatting
- Quality-assurance checks
- Performance summaries
- Localization support
- Reformatting an approved idea for additional channels
Teams should preserve explicit human ownership where judgment creates the value:
- Positioning
- Product promises
- Customer interpretation
- Strategic tradeoffs
- Creative territory
- Humor
- Cultural sensitivity
- Final selection
- Material claims
- Ethical boundaries
- Brand-defining language
The dividing line will move as tools improve. Accountability should remain visible even when the production method changes.
Measure the value of selection
Taste can improve business performance, but the relationship will rarely fit inside one attribution cell.
Teams can still measure whether better judgment creates stronger work.
Compare selected concepts against generic or control executions. Track:
- Attention and completion
- Recall and branded search
- Direct and returning traffic
- Engagement quality
- Saves, shares, and customer reuse
- Qualified conversion
- Sales acceptance
- Message accuracy during sales conversations
- Customer language adoption
- Creator, analyst, or community references
- Performance decay over time
Distinctive work may outperform immediately. It may also create value through slower recognition and repeated exposure. A useful testing program should include both conversion measures and memory signals.
The content marketing guide for 2026 makes a similar distinction between gaining visibility and earning conversion. Taste influences the connective tissue. It helps determine whether the material people encounter gives them any reason to continue.
An operating checklist for marketing leaders
Before approving the next AI-assisted campaign, ask:
- What customer evidence shaped the idea?
- Which product truth gives us permission to make the claim?
- What would remain if a competitor copied the format?
- Which part should the audience remember tomorrow?
- Why is this the correct format and channel?
- What did human judgment add after generation?
- What did the team deliberately reject?
- Who owns the final decision?
- Which business behavior should change?
- What future maintenance obligation does this asset create?
A team that cannot answer these questions probably needs another round of thinking rather than another round of generation.
The bottleneck moved upward
AI did not remove the hard part of marketing. It exposed where the hard part had been hiding.
Producing a plausible asset required enough time and labor that teams could mistake executional difficulty for strategic value. Once production became easier, the importance of customer understanding, product truth, editorial standards, cultural awareness, and decisive selection became more visible.
The companies most likely to benefit from AI will not be the companies that generate the largest number of options.
They will be the companies that know which option deserves to become real.