Marketing AI used to mean help writing a post. A caption here, an ad draft there — and the owner still did the real work: deciding what to sell, to whom, and whether any of it worked. That era is ending. The new pattern is a working loop. The system studies your real sales and customer data, drafts the ads and messages, a person you name approves them, the campaign runs, and you get a plain daily report on what sold — not just what got clicked. This week's Scale with AI episode walks through the loop, the evidence behind it, and the deadline sitting ten days out.

The Shift

The difference between a tool and a loop is what gets judged. A tool is judged on output — how many posts, how many drafts. A loop is judged on finished, approved campaigns and revenue. In a loop, the machine does the analysis and the drafting, a named human approves anything that spends money or reaches a customer, and the report at the end of the day answers the only question that matters: what sold? The evidence that this works is starting to arrive. A working paper published this month studied small online sellers using AI built directly into the software they already ran their business on. Sellers who adopted it saw monthly sales an estimated 51.6 percent higher than similar sellers who did not, with orders up about 47 percent — and the largest gains went to the smallest firms. Honest limits: that is one platform, one group of sellers, and a research estimate, not a promise. But it matches the pattern we see up close. AI built into your daily work beats another subscription you have to remember to open, and the advantage lands hardest exactly where an owner has the least time to do marketing by hand.

The Law

On November 25, ten days from today, new Federal Communications Commission rules take effect for how customers opt out of your automated calls and texts. The plain version: if a customer opts out after a purely informational message, like an appointment reminder, that opt-out can apply just to that kind of message, so customers keep the alerts they still want. Marketing is stricter — an opt-out after any marketing message ends all of your marketing calls and texts to that person. Every business must clearly state its designated opt-out method, and that method has to actually work. You must honor an opt-out within ten business days at the legal maximum; the smarter standard is to honor it the same day, log which category it was, and keep the proof. Underneath all of this sits the older rule: since February 2024, an AI-generated voice counts as an artificial voice under federal law, so AI marketing calls carry the same consent duties as any robocall. Test your own opt-out this week — text stop to your own business number, tell one of your own team members to stop your messages on a live call, and confirm both actually stopped and were logged. This is general information, not legal advice — and it is exactly the kind of setup detail worth getting right once.

Who's Doing It

Consider the pattern from a seafood shipper selling online across the country. An AI system analyzed its store and advertising data, wrote the ad ideas and creative briefs, produced the ads, and sent every one to the owner in a chat message for approval before a dollar was spent. In the first week, the owner approved about 20 percent of what the system made. By the end of the test, he was approving about 90 percent with no changes, because the system had learned his taste from his corrections. The reported test result was three and a half times the revenue of the company's usual ads, on about twice the spending — roughly 63 percent more revenue per advertising dollar. The honest limits matter: one company, the figures are the platform's own report, and the test budget was provided for the trial. Do not expect those numbers. Do copy the structure — machine drafts, owner approves, nothing spends unapproved — and watch the approval rate itself, because it tells you whether the system is actually learning your business. Now compare a survey of 1,400 online sellers released this week: 59 percent use AI for marketing and social media, but only 15 percent use it for sales insight, 13 percent for customer service, and 4 percent for operations. Most businesses are using AI to make noise. The few using it to run the loop — from data to draft to approval to sale to report — are the ones pulling away.

Your Move

Three moves, in order. First, fix the funnel before you buy traffic. We saw a case this month with over 111,000 ad views and zero sales, because the page and the checkout failed on phones. AI that sends more people to a broken page just loses money faster — open your own site on your phone and buy your own product before you spend anything on reach. Second, pick exactly one marketing loop to automate: product photos and drafts for one product line, or follow-up messages for one kind of lead. Write the approval line before you start — the system may draft anything, a named person approves anything that spends money or reaches a customer, and nothing with a price promise or a contract ever leaves unapproved. Third, run the opt-out test described above, because November 25 does not move. We walk through the whole pattern in today's episode of Scale with AI: listen to the full episode.

That is what we build at The Agency: one clean loop, with the approvals, the consent records, and the daily reporting set up properly for your business. Start with a free AI Audit. We will look at your marketing and your funnel, and tell you plainly which loop to automate first. No hype — just a clear picture, and a team that builds it with you.