Generating 100+ pieces of enterprise content using AI
The Weekly Signal is a weekly letter about doing good marketing work as AI changes how we do it. Expect ideas, people to learn from, and things I’m figuring out in my own work.
Inside my AI content engine
I get a lot of questions about how I build enterprise-class content generation systems that turn a video into an article, social posts and graphics. So I recorded the process, including the parts where I stop the agent and make a decision.
A long-form input like a podcast transcript is just one input. The agent also needs the way I want to write, examples of the designs I want it to use, and somewhere to save the work so other people can review it. In the walkthrough, you’ll see me choose an angle, review the newsletter, work from my Figma library and save the finished package for final review by another teammate. It starts with the recording already done. If you’re wondering what sits behind the output, this is that part.
Does every person producing content need this setup? No. But I’m typically building agentic systems for clients that need consistently on-brand, on-voice content in large volumes modified across many voices or profiles, in a process where multi-person review and editing is expected.
If that’s the kind of system you’re operating, watch the demo and reach out if you’d like to chat about it!
Count the work someone actually uses
Ruben Dominguez · The AI Corner · October 14, 2026 · Article
A busy agent can produce a very impressive activity log. Ruben’s piece asks the less comfortable question: what did the business get for the entire cost? His proposed unit is a complete workflow, including setup, review, repairs and unsuccessful runs, measured against the old way of doing the job. For a content team, that means looking beyond drafts generated to pieces accepted and used. The number to watch is the cost of work your team accepts and uses.
The relationship work still matters
Kyle Poyar and Maja Voje · Growth Unhinged · October 14, 2026 · Survey
Their September survey of 300 software GTM leaders puts introductions and intimate events at the top of respondents’ channel ratings. AI search is also gaining ground. Those findings can coexist: use AI to improve preparation and execution, then decide where an actual conversation would move the relationship forward. Compare its company-size and deal-size mix with your own before borrowing the benchmarks.
Before deciding AI images are the problem, look at the image
Rachel Banawa · Nielsen Norman Group · August 21, 2026 · Research
This older study is relevant to the design discussion in the walkthrough. Seventy-seven U.S. participants evaluated versions of a fictional consulting firm’s page with different hero images. They weren’t told which images were generated. In that setting, the AI images did not carry an inherent trust penalty compared with stock photography. The study did not test disclosure. For your own creative, look closely at representation, believable interactions, visual errors and the final crop.
“There are two ways to win right now: being exceptional at relationships and being exceptional at AI.”
Kyle Poyar and Maja Voje, Growth Unhinged, “What comes next in GTM”, October 14, 2026.
Turn a complaint into an ad hypothesis
Cody Schneider’s documented workflow starts with problems people describe on Reddit, matches them to product features and develops ad scripts. You don’t need to hand over an ad account to try that first part.
Start here: Collect ten relevant public complaints. Ask your agent to group the underlying problems, keep the source links and match each group to something your product can actually do. Pick one defensible match and draft three different ad concepts. Have someone who knows the product challenge the promise before you choose a concept to test. You’ll have a focused idea to test and a clear reason for testing it.
Find the competitor change that calls for a response
Corey Haines’s competitor-profiling skill gathers positioning, pricing, product pages and reviews into a dated, comparable record. Keeping that record lets you ask what changed, rather than generating the same competitor summary again.
Start here: Save one competitor’s homepage, pricing and changelog today. Repeat next week, then ask your agent to show only evidence-backed differences and propose one response for your team to consider. A new headline might merit a sales-enablement update; a missing page might simply be a site change. Check the source before treating either as a strategic move.
Make a short list of customers to call
Shopify’s segmentation guide describes using Sidekick to find valuable, engaged customers. In its Clubhouse Skin example, Ben Attwood used the analysis to reach customers individually. That is a more interesting use of a segment than immediately sending it another discount.
Start here: With data you’re authorized to use, ask your agent for a small group of repeat customers whose purchase interval has recently stretched. Verify the records, then have the account owner choose five people for a personal check-in. Bring what you learn back to the product or retention team. The aim is to find a customer problem you can do something about.
Ads can follow the next question in AI Mode
Google · October 28, 2026
Google says it is launching ads in follow-up AI Mode queries in the U.S. and expanding AI Mode ads into ten more markets, including Australia, Canada and India. For paid-search teams, the practical question is whether an ad and its landing page answer the more specific need that emerges during a conversation. Check availability in your account and judge the results against your own campaign.
The shopping assistant may arrive inside the browser
Commerce and Cloudflare · October 21, 2026
The companies announced WebMCP support for eligible BigCommerce storefronts. Compatible browser agents can use exposed shopping tools to find products and work with a cart; the shopper still completes checkout. The current documentation distinguishes the browser connection from the store’s server-side shopping tools. If you run a supported store, have your commerce team check what is actually exposed and test product variants and cart behavior. Try a few real products, including variants, before relying on the experience.
A more explicit say in where Google spends
Google · Rollout began October 15, 2026
Google introduced channel priorities and, for supported campaign types, budget allocation controls. One important detail: Performance Max groups Google-owned surfaces together. This is not a separate budget dial for every placement, and Demand Gen does not support the budget-allocation feature. Ask your paid-media lead which controls have appeared in your account and what decision they would change. Judge a test by qualified demand or revenue, not simply a tidier spending chart.
An ad idea hiding in the collection
Peabody Essex Museum · Campaign story by Whitney Van Dyke · July 24, 2026
Image: Peabody Essex Museum. Campaign artwork from its July 24 story.
PEM built its campaign around the strange, specific stories in its collection. One ad’s “We’ve Got That Drip” line refers to a preserved sea turtle that really did leak oil. The visual and the joke lead to an object you can actually visit. It’s a tourism example, but the lesson travels to a crowded B2B category: find the detail only your business can substantiate before reaching for another broad claim.
Something to borrow: Ask the people closest to your product for an odd, memorable fact that customers rarely hear. See whether that detail could lead a campaign and whether the destination delivers on the curiosity it creates.
Where does AI-produced work still get stuck on your team: choosing the idea, getting the draft right, making the design consistent, or getting everyone’s approval?
Reply with one recent example. What did you have to change before the work was actually usable?
Know someone doing great marketing with too much on their plate? Send them The Weekly Signal. One less thing for them to dig through.
Someone sent you this? Good friend.
