Writing my book was one job. Getting anyone to read it was a separate one.
I recently published a collection of poetry and essays. When it came time to promote it, I was the whole marketing department: strategist, copywriter, designer, and video editor, with no budget and a few spare hours a week.
If you run or market a small service business, that probably sounds familiar. Maybe you have one marketer, or the marketer is you, fitting it in between client work. You know you should be posting, emailing and making videos, and you can't get to all of it.
So I gave AI a copy of my book and asked how to promote it. I used ChatGPT and Gemini to brainstorm drafts for posts, captions and emails, and Canva to make the scenes for a book trailer. On one of my social media accounts, the trailer got 991 views, my posts got more comments, and I sold 8 copies that month.
I found four marketing jobs AI could help me with: finding a reason people should care, turning one piece of content into several, making videos, and planning a campaign for different audiences.
Here's an example of a campaign that brings all four together.
One campaign that did all four
During the pandemic, Cadbury in India ran a Diwali campaign to help struggling local shops. Shop owners entered their store's name on a website and got a version of the ad in which Bollywood star Shah Rukh Khan named their store. People then saw the versions for shops near them. According to WPP, the agency group behind the campaign, more than 2,000 stores took part, 130,000 ads were made, and the ads drew 94 million views on YouTube and Facebook. The campaign's case video shows how it worked.
In one campaign, Cadbury:
- Gave people a reason to care beyond "buy chocolate": help the shop down the street.
- Turned one ad into thousands.
- Made videos without filming a single shop.
- Showed each viewer the version that fit them.
Cadbury had a film star and a global ad agency. I had a laptop and a few spare hours a week. The same four moves were open to me, and they're open to a firm of 10 to 50 people. You won't need 130,000 versions. One for each type of client you serve might do.
Here's what I tried and how you could use each one in your firm.
Find the reason people should care
"Buy my book" isn't a reason to buy a book. I gave AI the book description and asked for several reasons a reader might care. It came back with more than I expected. I kept the ones that were true to the writing and threw out the rest.
Try it: paste ten of your public Google reviews and a short description of what you do into ChatGPT or Claude, and ask:
Here are 10 reviews from our clients and a description of our firm. Give me 5 reasons someone would hire us, in our clients' own words. Next to each one, say which review it came from.
Use reviews that are already public, or remove names first.
Turn one piece into ten
From one book description, I got drafts of emails, book club discussion questions and posts shaped for each platform. I still rewrote plenty, but I was editing instead of staring at a blank page.
Try it: take your best article, webinar or client presentation and ask:
Turn this into 5 LinkedIn posts for [who you serve], each making one point from the piece, plus a short client newsletter. Use only facts in the piece. Flag anything you're unsure of.
If the drafts don't sound like you, our guide Make AI writing sound like you shows you how to give AI samples of your writing so its drafts sound more like you.
Make videos without a studio
Video was the thing I knew I did not have the time for. So, I built the trailer with AI, then trimmed and finished it in Adobe Premiere Pro. It got more views than anything else I posted for the book. That's about five times the 200 views a post of mine usually gets.
Try it: write down a story you already tell on sales calls about one client problem you solved. Ask AI to turn it into a 60-second script. Record yourself reading it on your phone, then use a tool like Descript to cut the pauses and add captions.
Plan the campaign, then tailor it
Instead of asking for one post at a time, I gave the AI agent an audience and an offer, then asked it to map out a full week. Seeing the week at once showed me where posts repeated each other, so I could make each one add something new.
Try it: plan one month around a single offer, then ask:
Rewrite each of these posts for [client type A], [client type B] and [client type C]. Keep the facts the same. Change the example and the reason it matters to each one.
An accounting firm might write one version for restaurant owners and one for contractors, each about the problem that client actually has.
What made the drafts usable
Tell AI who the reader is. When I described the reader and what I wanted them to feel, the drafts got better. Vague instructions got generic drafts.
Have someone who knows the subject check every claim. As an apprentice, I can say this with some authority: AI drafts like a keen new hire. Fast, eager, and in need of someone who knows the work to check it. In an accounting firm, that means a partner approves any tax point before it's posted.
Track what each post produces. Some posts took me longer with AI than they would have alone, and faster isn't always better. In the Content Marketing Institute's B2B Content and Marketing Trends: Insights for 2026 report, a survey of 1,015 B2B marketers published in October 2025, 87% of those using AI for content said their productivity improved, but only 58% said their content quality did. I now track replies, clicks and sales against the time each post took, so I repeat what earns its time.
Write the setup down once. Here's an illustrative example for an accounting firm:
Reader: owners of trades and construction businesses, 5 to 50 staff.
What they care about: cash flow, surprise tax bills, time.
Voice: plain and direct, the way our partners talk on a call.
A good post: one real problem, one thing to do about it, under 150 words.
Never: client names, or tax advice a partner hasn't reviewed.
In ChatGPT or Claude, create a Project (a saved workspace in the sidebar) and paste this brief into its instructions. Every chat in that Project starts from your brief, so you stop explaining yourself every time. To bring in your logo, fonts and approved wording as well, see our guide Make every AI deliverable feel like your brand.
One professional services firm we work with uses this kind of setup. Its brand guidelines, post templates and past posts went into its AI tool once. A recurring social post, the kind that goes out at the same time every year, used to take about an hour: find last year's version, update the facts, rebuild the graphic, write the caption. By the firm's own estimate, it now takes about 5 minutes, including a person's review before it goes out. How much time you'd save depends on the task, your firm and who's doing it.
Not every marketing task belongs with AI. Our article Should AI do this work? gives a quick way to decide, and Is my data safe with AI? What will IT say? covers what to check before client information goes anywhere near it.
For eight marketing jobs AI can prepare, and an example of checking one expert answer across four pieces, see What AI can do for marketing in a small business.
Want help setting this up?
Building that setup is part of what Sidekick Orchestration does. We coach leaders to use AI on the work that fills their week, then help them roll it out across the company. For marketing, that means one brief, one set of templates and one review step for whoever posts, even if that's an office manager with two hours a week. If your staff already use AI on their own, we give them one shared brief and set of templates so every post sounds like the firm, with a named person approving anything a client sees.
Book a free 30-minute call. We'll go through your marketing tasks together, and you'll leave the call knowing which two or three marketing jobs to hand to AI first.
Sources: WPP, #NotJustACadburyAd (the agency group's own case study); Ogilvy and Wavemaker, case video. Content Marketing Institute, B2B Content and Marketing Trends: Insights for 2026, October 8, 2025.


