How One Marketing Leader Reimagined Her Team With AI

A Conversation with Raginee Mamgain, Director of Marketing, Constellation Dealer Group.

Artificial intelligence has become a fixture in marketing conversations, but for most teams, its use remains limited to drafting emails or brainstorming ideas. Raginee Mamgain has taken a different approach.

When her content manager resigned several months ago, she saw an opportunity to rethink how work got done. Rather than backfilling the role, she built a team of AI agents, each designed to support a different stage of the marketing workflow. Today, her four-person team lets agents handle the research, the ideation, and the writing producing more with fewer resources

At a recent marketing roundtable, Raginee shared how she built her AI-powered marketing team and the lessons she’s learned along the way.

Q: What made you turn to AI in the first place?

In the beginning, we started embracing AI out of necessity rather than strategy. My team is four people, and we handle marketing for two business units. We cater to more than 25 sales and product people, so you can imagine the volume of requests we take on. Emails, social posts, webinars, blogs, publication partnerships – all of it flows through 2 people on our team.

When we first got started, we were trying to figure out what are the things we need right now that are repeatable, that can be automated.

Q: Walk us through how you built the agents.

The turning point was a hackathon where I learned Cursor, the AI coding environment I now run my whole operation from.

In my experience, the real craft is in the prompt that comes before the build. I used a tool called Prompt Cowboy to draft the master instructions. I wrote 20, 30 lines of description explaining exactly what I need: this is the marketing team structure, these are the industries we cater to, this is the kind of content we come up with. That output became the prompt I pasted into Cursor to spin up the agents, each with a name and a job.

It’s important to emphasize that none of this landed on the first try. I spent about a week building, testing, and troubleshooting before everything started working the way I wanted. There was plenty of trial and error along the way, but we ended up with a system, and I’m still tweaking and improving it as I learn.

Q: Everyone’s worried about “AI slop.” How do you keep your content from sounding like everyone else’s?

This is one of the most important parts of the system for me. Every morning, I have a research agent scan key industry publications, associations, and competitor activity to identify recurring themes and emerging insights. But I also have to assume my competitors are doing something similar, so I built a contrarian streak into the prompt.

The agent generates five content ideas each day, but two of them have to challenge the prevailing conversation. If everyone in the industry is writing about a shortage, for example, the agent is instructed to explore the opposite perspective. Otherwise, everyone ends up publishing the same AI-generated content.

Q: Where is AI still falling short for you?

Graphic design isn’t quite there yet. I’ve connected a Canva-powered designer agent and trained it on our brand guidelines, but it’s not ready to replace my human designers. It still drifts outside the guidelines, changing colours, fonts, and other brand elements. You end up spending a lot of time correcting it.

Instead, I changed the agent’s role entirely. Rather than asking it to create finished designs, I prompt it to produce a detailed design brief for a human designer. That one shift has saved me hours every week because I’m no longer starting from a blank page or writing creative briefs from scratch.

Q: A lot of us struggle to work as a team in these tools. How do you share what you build?

This is something a lot of teams are still trying to figure out. Many people have built impressive agents, but they’re difficult to share or hand off to someone else. I’ve had much better success building Claude Skills instead of artifacts.

You can copy an entire Skill and share it with anyone using Claude. I’ve created a Skill for each of our products, and for enterprise teams there’s an even cleaner approach. You can have IT deploy those Skills across the organization so everyone is working from the same foundation.

It’s made collaboration much easier. If someone on my team is overloaded, I can hand them the relevant Skill and know the tone, messaging, and product knowledge will be consistent.

Skills are making me rethink my own approach. I’m starting to question whether agents are always the right answer. Skills are easier to maintain, easier to share, and you can run them whenever you need them.

Q: Let’s talk AEO (Answer Engine Optimization). Everyone’s trying to figure it out, what are you doing?

The fundamentals still come back to producing high-quality content. You need FAQ pages, visibility across trusted third-party publications, and a website that answers questions conversationally instead of just listing product features.

We’ve started experimenting with ChatGPT Ads, which work very similarly to Google Ads. The search volume in our niche is still relatively small, but we’ve already started seeing referral traffic from ChatGPT show up in GA4.

One of the biggest lessons I’ve learned is that citations matter. There’s a big difference between simply being mentioned and being cited as a source. One of the places AI models reference most often is YouTube, so we shifted our strategy. We now have an offshore resource repurposing every webinar and customer story into YouTube content.

Q: After all your experimentation thus far, what’s been the real payoff?

My work has become more fun. The constraints that used to define my job have loosened, and it’s made marketing exciting again. AI has opened up so many new possibilities.

I’ve also made peace with the fact that marketing attribution will probably always be a little murky. I’ve started thinking differently about what success looks like. Sometimes the biggest win isn’t a flashy campaign, it’s saving time, reducing costs, or creating capacity.

Back to Insights