How do you decide which parts of a marketing job AI should take? Barry Moline, president of BJM Solutions and a CEO for 29 years, gave me an exercise on Unscripted Small Business that I haven’t been able to stop thinking about. Update the job description with the person who holds the job. Paste it into your AI. Then ask two questions: “what tasks can AI help with? And what tasks should remain human?”
Barry speaks on leadership, teamwork and AI, and he isn’t a marketer. That’s why his version is useful to you. It skips the tool debates and starts from the work your team already does, which is where most AI rollouts in agencies and in-house teams go sideways.
Watch the full episode on YouTube or listen on Castos.
Start the AI rollout from an accurate job description
I asked Barry how a CEO can bring AI into a company without building resistance. His first step isn’t a tool. It’s paperwork most teams skip: “take their job descriptions and with their employees who have those job descriptions and update them to make sure they’re accurate.” Job descriptions drift. After six months or a year, tasks get added and dropped, and nobody writes it down.
Once the description is current, “enter that job description into your favorite AI and say, what tasks can AI help with? And what tasks should remain human?” Barry’s verdict: “It’ll give you a great list.”
Here’s why that order matters for a marketing team. If you start with a tool, you get a list of things the tool can do. If you start with the job, you get a list of the work, sorted. Try it on your own roles: the content writer, the SEO analyst, the account manager, the person who builds reports every Monday. Then sit down with each of them and argue about the list. The arguing is the point. The employee is in the room when the line between AI work and human work gets drawn, so nobody hears about it secondhand. If you want to turn one of those tasks into something repeatable, our SOP on building a Claude skill that does your pre-work is a good first build.
The industrial laundry test shows what the list looks like
Barry ran the exercise with a group of industrial laundry operators, which he freely admits is a topic he doesn’t know much about. That’s a feature. He identified five categories of workers, from maintenance techs and safety technicians to supervisors, floor workers and drivers who visit customers.
Some of the work stays human no matter how good the model gets. “Somebody’s got to turn the screwdriver. Somebody’s got to replace the belt.” Two of the AI ideas surprised the room:
- A safety course built from the manuals. Upload the operations and safety manuals, have AI build a course with test questions, and translate it for workers who are stronger in Spanish. “You really want people to understand it at their core in their native language.”
- Searchable machine manuals. A maintenance tech searches every operating manual on a phone at the machine, so AI can “move you faster through the technical materials and then allow you more time to fix the machine.”
He said “their mouths were dropping open, like, I never thought about that.” Both ideas use documents the company already owned. Your version is the brand guidelines, the past audits, the client onboarding docs and the call recordings sitting in a shared drive.
Your best AI input is the expert down the hall
That’s the same lesson Jeremy Moser of uSERP, a link building agency, gave me on his Unscripted SEO episode. A cold prompt asking for marketing ideas gets you generic output. Feed it recorded sales calls and transcripts, and interview the people who do the work, like the installer or the salesperson, and you uncover the pain points customers actually bring up.
I’ve done this. I interviewed a concrete-wall installer and asked what he wishes every customer knew before he shows up. His answer, as I told Barry: “measure it three times because it’s concrete and it’s not changing dimensions. There’s no wiggle room.” That went into the site copy, and the next job went smoother because the customer had already read it. The way I put it on the show: “AI can help human expertise surface in appropriate context more easily.” If you record your client calls, every interview is a link-building engine too.
Shared context beats everyone running their own AI
I brought up my conversation with Daniel Hebert of Oleno, which builds marketing context for Claude and ChatGPT. Here’s the problem we talked about. Everybody got thrown into the deep end with AI, and everyone uses it differently. Your Claude instance or your GPT builds its own layers of personalization, so the way I use a tool “isn’t at all the same way that somebody else is going natively think about it.” Getting a team to share skills in Claude takes SOPs and a deliberate plan.
Barry’s fix is cultural. He tells leaders to “get cross-departmental teams together, anybody who wants to, twice a month, and share how people are using AI.” Somebody projects their screen and walks through it. Some people use AI for research, some for coding, some for graphics. “No one has a full grasp of everything. It’s an elephant.”
He also made a point about cost and stress that agency owners should hear. Peer sessions have no test and no deadline, so “the stress level is lowered significantly and our minds are open to learning from each other.” Compare that with “an outside consultant that we’re spending five thousand dollars a day for.” For marketing teams, I’d add one rule: when someone shows a workflow that works, write it down as a shared skill or SOP that week, so the show-and-tell turns into context everyone uses. Dr. Gleb Tsipursky’s research on why marketing teams quietly resist AI tools explains what happens when you skip the people part.
Hire marketers for get-along ability and train the tools
The other half of the episode was hiring, and it applies to marketing teams more than most. For most of his 29 years as a CEO, Barry hired the most skilled candidate. Then he found “the most skilled people were not always the best teammates.” After one bad hire, he changed the rule: “We’re no longer hiring for skills, we’re hiring number one for get-along ability.” They trained more, and “our results were better.” The tools your team uses will keep changing, and Barry says he sees “something new in AI at least once a week.” To me, a teammate who learns in public is worth more than a specialist who hoards what they know.
AI job description exercise: questions to run on your marketing team
- Is each job description on your team accurate today, and did the person doing the job help update it?
- When you paste it into AI, which tasks does it say it can help with, and which should stay human? Do you and the employee agree?
- Which documents does your team already own, like calls, audits and guidelines, that AI could turn into training or answers?
- Who on your team are the experts you haven’t interviewed yet?
- Is there a standing, low-stakes time for people to show each other how they use AI?
The full episode recap is Barry Moline on hiring for get-along ability on Unscripted Small Business, and my own take is on being the host, not the hero. Barry gives away one-to-two-page guides on building teams and influence at barrymoline.com/resources, his book is Connect! How to Quickly Collaborate for Success in Business and Life, and you can find him on LinkedIn.
This episode is supported by our sponsors: ClearLead Digital, which does property management seo for property managers, along with uSERP and Oleno, both mentioned above.
Sponsor disclosure: uSERP, Oleno and ClearLead Digital are paid sponsors of the Unscripted podcast network. Links to them are marked as sponsored.