The AI-in-marketing conversation has shifted from “are you using it?” to “prove it paid off” — and this week delivered hard numbers, uncomfortable agency admissions, and a quiet infrastructure shift landing on every iPhone by default.
Proaction Puts a Revenue Number on Agent ROI
Fleet management startup Proaction deployed OpenAI’s Codex alongside GPT-Live-1 and GPT-6 Astra and recorded a 60% sales lift plus 75+ hours of manual work eliminated — figures OpenAI published in a formal case study this week. That makes it one of the most concrete B2B outcomes attached to the agent stack so far, tying the tools directly to revenue rather than stopping at a productivity claim.
When pitching AI investment internally, lead with the revenue metric first and the hours-saved figure second — that sequencing is what moves budget decisions.
Brands Are Quietly Gaming the AI Shopping Shelf
Modern Retail writer Deana Burke got a completely fictional natural-deodorant brand successfully recommended by ChatGPT, Claude, and Gemini using nothing but structured product content and review-style language — no ad spend, no real market presence. It confirms these assistants weight specific claims, structured detail, and third-party review phrasing heavily, regardless of whether a brand actually exists at scale.
Audit your product listings for specificity and review-style language now, because AI shopping assistants are already pulling from those signals when they surface recommendations.
Agencies Have Quietly Stopped Hiring Entry-Level Staff
Agency executives at Digiday’s AI Marketing Strategies event admitted under Chatham House rules that AI tools have made them halt entry-level hiring entirely, with junior roles covering research, drafting, and QA now largely absorbed by software. The sharper problem they flagged: no one has built replacement training paths for the craft skills those roles used to develop in people over time.
If you’re mid-career at an agency, document the judgment calls you make daily — that institutional knowledge is exactly what AI can’t generate and what clients will increasingly pay a premium for.
A Copywriter Scores AI Output at 95% and 50× the Volume
Copywriter Neville Medhora of Swipe File published a self-graded comparison finding AI-generated marketing copy now hits roughly 95% of his own quality bar — and produces 50 variations in the time he writes one by hand. The gap is no longer idea quality; it’s volume, which means the real bottleneck on copy testing has shifted from writing time to review and judgment time.
Restructure your copy workflow so a human is spending most of their time evaluating and selecting variants rather than drafting them — that’s where the leverage is now.
Duer Is Betting Print Beats Algorithms for Owned Reach
Performance-denim brand Duer is launching a quarterly print magazine called Interval at 5,000 copies, distributed through stores, loyalty mailers, and wholesale partners — a channel that doesn’t depend on Google referrals or platform feeds. The timing is sharp: publishers at the Digiday Publishing Summit this week described exactly this kind of owned-audience move as their post-search survival strategy.
Map every touchpoint where you reach your audience directly without a platform intermediary, then invest in deepening at least one of them before Q1 planning locks in.
iOS 27 Makes Siri an Agent on Hundreds of Millions of Phones
Apple’s iOS 27 rebuilds Siri as an LLM-powered agent that maintains conversational context and executes tasks — drafting emails, surfacing project-linked messages — moving it from a voice command shortcut to a genuine assistant. That puts an agentic AI on several hundred million iPhones by default, eliminating the device-level adoption hurdle that kept AI tools niche for anyone who never downloaded a separate chatbot.
Treat the iOS 27 rollout as a consumer AI literacy event — your audience’s baseline expectations for what software should do for them just moved up significantly.
Stagwell Chose Palantir for Data Fusion, Not Dashboards
Stagwell chief AI officer John Kahan explained the holding company’s decision to build its Marketing Cloud on Palantir rather than larger legacy data vendors, citing Palantir’s ability to fuse disparate client and campaign data sources as the differentiator — not visualization features. The bet is that agencies competing against bigger holdcos need a genuinely different data infrastructure to make AI-driven media buying credible, not another reporting layer.
Before evaluating any new AI marketing platform, ask specifically how it handles data from sources that currently don’t talk to each other — that integration capability is where the real competitive edge lives.
More from Rafal Reyzer
For deeper dives on AI and marketing strategy, visit my YouTube channel →
Rafal Reyzer
Hey there, welcome to my blog! I'm a full-time entrepreneur building two companies, a digital marketer, and a content creator with 10+ years of experience. I started RafalReyzer.com to provide you with great tools and strategies you can use to become a proficient digital marketer and achieve freedom through online creativity. My site is a one-stop shop for digital marketers, and content enthusiasts who want to be independent, earn more money, and create beautiful things. Explore my journey here, and don't forget to get in touch if you need help with digital marketing.