Reporting period: October 26–November 1, 2026
This week’s signal is not another race to make a bigger chatbot. The more useful shift is happening inside real workflows: AI is recommending products against live inventory, operating computers at a lower cost, and turning prompts into interactive worlds.
For marketers, the common thread is clear. The value is moving from what a model can generate in isolation to what it can do with good data, tight rules, and a measurable job to finish.
Fanatics puts AI recommendations inside the buying journey
What happened
On October 26, Fanatics and OpenAI announced an AI-powered version of Shop the Look. It recommends coordinated, in-stock merchandise on product pages using enriched catalog, team, and inventory data, while Fanatics’ merchandising rules constrain the results. The looks refresh daily as the catalog changes (source).
The companies say they tested the experience across roughly 30,000 MLB products and saw 10% higher engagement than the previous Shop the Look experience. That is a vendor-reported result from this deployment, not an independent finding or a promise that the same lift will transfer to another catalog (source).
Why it matters
This is a practical example of AI working at the decision point instead of producing one more top-of-funnel asset. The recommendation is useful because it is grounded in current inventory, connected to a shopper’s context, and bounded by business rules.
GrowthGary’s take
The model is only one part of the system. The harder—and more reusable—work is cleaning product data, defining compatibility rules, deciding what “good” looks like, and measuring the result against the experience it replaces.
That is the same discipline behind a dependable AI marketing workflow: clear inputs, a constrained output, a reviewer, and a comparison point.
What to do next
- Choose one customer decision where people regularly need help comparing or combining options.
- List the live data and deterministic rules the recommendation must use.
- Test against the current experience and track both engagement and the downstream business outcome.
- Keep a human review path for edge cases, unavailable inventory, and brand-sensitive combinations.
GPT-6.1 Terra lowers the cost of computer-use workflows
What happened
OpenAI released GPT-6.1 Terra on October 26 for ChatGPT Work, Codex, and the API. The company positions it as a faster model for visual reasoning, writing, and computer use, including tasks such as filling forms and editing video. Standard API pricing is $1 per million input tokens and $5 per million output tokens; OpenAI says that is 50% below GPT-5.6 Terra and GPT-6.1 Sol pricing (source).
OpenAI also reports better speed, task cost, and benchmark performance than several of its own prior models. Those comparisons come from OpenAI’s evaluations, so they are a reason to run a controlled test—not a substitute for one (source).
Why it matters
Computer-use models become more interesting when the economics support repetitive, high-volume work. Marketing teams can test them on jobs such as compiling campaign evidence, checking page elements, moving approved information between systems, or preparing recurring reports.
GrowthGary’s take
Lower token cost does not automatically mean lower workflow cost. Preparation, retries, permissions, and review still count. The useful metric is not the model’s price or how many steps it attempted; it is the cost of an accepted result.
If you are still deciding where an agent belongs, start with the practical distinction between an assistant and an AI agent that can take actions.
What to do next
- Pick one repeatable computer task with stable inputs and a visible finish line.
- Record the current time, error rate, and review effort before introducing AI.
- Limit permissions and require approval before publishing, spending money, or contacting customers.
- Run a small sample and compare total cost per accepted result, including correction time.
Genie 4 previews a new kind of interactive creative
What happened
Google DeepMind introduced Genie 4 on October 27 as a research preview for trusted testers. The world model generates explorable scenes with native audio and interactive dialogue, and users can change elements such as weather, camera position, people, or objects with natural-language instructions while the scene runs (source).
Google says creative testers are using it for pre-visualization and rapid concept exploration. The company also identifies current limits, including visual drift, imperfect object permanence, legible text, spatial accuracy, and long-session consistency. Generated video and audio include SynthID watermarking (source).
Why it matters
Creative AI is starting to move beyond making a flat image or clip. An interactive world can become a fast draft for an event concept, product environment, training scenario, or experience that a team can explore before committing to production.
GrowthGary’s take
Genie 4 is a preview, not a production-ready campaign platform. The near-term opportunity is concept validation: helping a team see and challenge an idea earlier. Brand accuracy, product truth, continuity, accessibility, and rights review still need people and a production process.
What to do next
- Identify one customer experience that would be easier to evaluate as an interactive prototype than as a written brief.
- Define the brand, product, and factual details that must remain fixed.
- Use current accessible tools to prototype the decision flow now; treat world models as an emerging production option.
- Keep the experiment inside a focused 90-day marketing plan so it competes against real priorities.
The takeaway: The strongest AI marketing systems are getting more grounded, more operational, and more interactive. Start with a measurable customer or workflow problem, connect the right data, add rules and review, then judge the result against what you do today.
Want to turn one of these signals into a practical workflow? Explore GrowthGary AI Marketing Automations.

