How Should Brand Marketing and Communications Teams Train on AI? Lessons From Our Six-Week Certification

Greenough, soon to be The Marka Group, certified our full team on AI through a six-week program with Pete Pachal of The Media Copilot. We did it to make sure everyone is living out our agency charter: AI-powered, human-led. Here’s what we learned and what the team built.

This week we finished with a Demo Day where our full team recorded Loom videos on something they built and discussed successes, challenges, and how they plan to use it in their everyday work. Our favorite part may have been hearing everyone’s voice at 1.75x speed. 😁

What we took away from the experience:

• Decide what AI's job is before you start. AI should have a clear job description and deserves a place in your roles and responsibilities documents. It can advise you, give you a single pane view of priorities, synthesize data, draft content, and more, but each role calls for a different level of review. Decide what role you want AI to play before you start, and make sure it’s consistent, well fed, and that a human leads and vets it.

• Make AI interview you first. Our team learned to have AI ask questions, one at a time, before it writes anything. It usually surfaces a decision or POV that needs to be made, or it becomes clear the proper level of detail isn't provided. Talking instead of typing helps too, because dictation captures more context and your intended tone faster.

• Remain human-led. Whether the feedback on what you prepared is a call that is coming from inside the house or not, don’t just have AI regenerate an entirely new draft. Understand the feedback and address it, as regeneration can leave some good material on the cutting room floor! And because no one likes making the same mistake twice, keep your project instructions up to date to reflect the executive's and/or brand's preferences!

• Check your own files before the internet. In one exercise, a draft release from a made-up company claimed to be the first wearable to score sleep regularity. A few minutes of checking turned up several products already doing it. The habit is to list every claim, check it against the company's approved materials first, and trace every statistic back to the original study. It's the same check a skeptical reporter will run, so let’s make sure we do it first.

AI also can't see everything. Many publishers (ahem, New York Times) block AI tools, so our team routes trade newsletters to their work inboxes, where Claude can read them alongside a web search.

Keep what you know separate from how you work

Pete's test for organizing AI tools comes down to one question: is this about one of our brand partners, or about the process? What we know about a brand partner should be hosted in its own workspace. A repeatable process, like a news scan or a claim check, becomes a shared tool anyone on the team can run, with no client details inside it.

Give agents less access than you think they need

Once AI can send, file, and post on its own, the biggest risk is no longer a bad actor. It's an agent with broad permissions that misreads the goal. Start every new automation with a person approving each step, watch a few runs, then loosen it. Our rule is that anything leaving the building stays manual.

What the team built

Every project at Demo Day came from someone's own account work, so we all now have ideas, skills, artifacts, and more to build from. Three dozen of them! A few highlights:

• A healthcare or cybersecurity roundup of the most important news and trends, updated daily at 8 a.m., with the top story of the day highlighted, why it matters to us and our brand partners, the angles we can take, the right spokesperson, and a first-draft pitch. All ready for human editing and vetting!
• A research assistant for 300-page industry reports. It first checks whether a newer edition exists and whether a competitor published a similar report, then pulls the statistics and a trendjacking hook.
• A pitch reviewer that reads a pitch the way the reporter would, against what that reporter has covered recently.
• A media training summary that color-codes where an executive landed each core message during a mock interview and where they could have pushed further. This has been a brand partner favorite!
• An onboarding document built from SharePoint and email history, so new team members get up to speed in one place.
• Monthly expense reports split out for every employee and drafted for review, with Excel handling the math. Because we also have to do math sometimes! Ha!

So, essentially, we’re living the dream over here!

The project I keep thinking about came from a new employee who has been with us for only a week. She built a project that briefed her on a new brand partner in about five minutes: what they sell, who buys it, the competitors, the people, and the recent coverage. She moved from onboarding to action in five business days vs. five weeks!

What this means if you work with us

Every Demo Day project kept a person in the driver's seat, and no decision or automated action happens without their approval. The team built this in on purpose because we know it’s our judgment that keeps quality high.

What comes next

The team is now certified and starting from the same place, which makes it easier to take the full team to the next level. We’re building out what our next certification will look like now, and it may include some coding. 👀

This was a real investment for us, and an important one. Our mission of being AI-powered, human-led means no one should be left behind, and we feel it’s our responsibility to work together on how and when to use it, beyond handing over the tools and mandating use.

As the saying goes, “If you want to go fast, go alone; but if you want to go far, go together.”

‍