The red pool is where my content engine started. In June, a reviewer sent back the first page it ever produced with two-thirds of it rewritten: pure red ink. Every step of that staircase is a page that came back cleaner, and the top step is July's review, quoted in full: "no changes."

Zero-click searches hit 68%, and top-ranked pages now convert clicks at less than half their old rate. If your growth model still assumes Google sends you customers, rebuild it now — direct relationships, email, communities, anything you own outright. Publishers blocking Google entirely isn't fringe behavior anymore; it's a live strategy at major outlets. Plan for search traffic as a bonus, not a foundation.

The build-vs-buy question isn't about cost anymore, it's about who keeps what your team learns. Rent infrastructure. Own your intelligence because that's your actual moat. Before signing anything new, ask where the learning lives and whether you can export it. If you can't, you're renting your own institutional knowledge back from a vendor.

I've sat in meetings where the CMO nods along about "leveraging AI" while the junior strategist is quietly running three tools nobody bothered to mention. Turns out that's not just me, 47% of marketers say leadership lags everyone else on AI skills.

I keep hearing "we optimized for AI search" like it's a secret formula. It's not, it's just... good marketing? Answer real questions, back up claims with evidence, stop writing "industry leader" copy nobody believes. Turns out the AI just wants what your actual customers wanted all along.

Stop hiring for "AI experience". It's a weak signal. Test whether candidates can spot bad AI output and connect it to business goals. And protect junior roles: if AI eats all the entry-level execution work, you lose the training ground where people learn enough to eventually direct AI instead of just operating it. That's a talent pipeline problem, not just a productivity win.

Defaulting to the priciest frontier model for every task is the new "we'll figure out cloud costs later." If you're running agentic workflows without routing logic, you're burning 20-30% on intelligence-saturated tasks that a cheaper model handles fine. Before scaling agent deployments, audit which steps actually need frontier reasoning versus which are just habit. The bill won't wait for you to notice.

The gap between AI users isn't tool access—it's what they're automating. Task-level prompting caps growth at your personal bandwidth. System-level programming (your objection-handling logic, your decision patterns, your SOPs) decouples revenue from your hours. Audit your AI usage this quarter: are you speeding up your own work, or building something that runs your judgment without you? Only one of those scales.