Winter 2023. I’d been reading Paul Roetzer and Mike Kaput’s stuff from the Marketing AI Institute and thought, “okay, let’s try this at work.” I was in account management at an agency, and I figured there had to be ways AI could make our lives easier.
Turns out, actually implementing this stuff is way harder than reading about it. Here’s what worked, what didn’t, and the stuff nobody tells you about.
The AI tool that had the most immediate impact wasn’t ChatGPT or Midjourney’s image generation capabilities.
It was Canva’s background remover.
I realize that sounds anticlimactic, but here’s where context comes in.
The account management team spent considerable time preparing client presentations. We often needed to show what projects could look like—compositing images, stacking objects, creating visual concepts. Previously, we had three options: skip the visual entirely, use stock photos that never quite matched our needs, or brief the creative department.
The third option meant briefing cycles, revision rounds, and for complex work, outsourcing to external 3D designers. This process was expensive and time-consuming, reserved mainly for major tenders.
With Canva’s tools, account managers could create early-stage mockups independently. We weren’t producing final creative assets, but we could visualize concepts for initial client conversations. The time savings were substantial.
Canva’s magic eraser served a similar purpose, allowing us to remove logos from reference images without creative department involvement. Midjourney helped us visualize unconventional ideas—clients understood these were AI-generated, but having visual references made proposals much clearer than descriptions alone.
ChatGPT and Claude became valuable for brainstorming (from client event ideas to test drive routes), asking better questions during project planning, and handling copywriting for presentations and emails.
These weren’t revolutionary applications. They were practical productivity improvements that compounded over time. The key insight: it’s not about technological sophistication, it’s about understanding your workflow and identifying genuine bottlenecks.
Initially, I created comprehensive PowerPoint presentations outlining AI use cases and implementation strategies.
Nobody read them, Im sure. To be honest, I wouldn’t either.
What actually drove adoption was identifying a few curious team members, helping them see results firsthand, and letting them share their experience with colleagues organically. Top-down mandates don’t work for this kind of change. You need genuine champions.
Resistance was inevitable. Some people were skeptical, others proposed alternative solutions. I learned not to argue. Instead, I’d ask them to research their alternative and present a business case.
Most of the time, they didn’t follow through. Occasionally, they’d surface something genuinely valuable I hadn’t considered.
The budget reality was straightforward: without demonstrable ROI, funding disappears. This isn’t experimental—every use case needs clear ties to time or cost savings. I presented specific problems we’d solve, secured licenses for Canva, ChatGPT, Claude, and Midjourney, and needed to prove their value.
We followed a structured rollout: two weeks of preparation, two to three months piloting with a small team, two weeks of evaluation, then gradual implementation over four to six months. Rushing this process would have created chaos.
Someone needs to own the implementation process. That responsibility fell to me as the AI Coordinator.
The role involved identifying problems AI could address, evaluating different solutions, optimizing prompting methods, and sharing best practices with relevant team members. It required continuous feedback gathering and technique refinement.
The ongoing work was less visible but equally important: monthly internal newsletters on AI developments, maintaining a shared folder with best practices, creating and updating a prompt library, and testing new tools before introducing them to the team.
I joined the Marketing AI Institute’s Slack community to learn from other professionals managing similar implementations. Staying current with AI trends wasn’t optional—it was essential to the role’s effectiveness.
This coordination function matters. Without it, teams fragment across different tools, learnings aren’t shared, prompts remain unoptimized, and many people never discover available resources.
It’s not particularly glamorous work, but it’s the difference between having AI tools available and having a team that uses them effectively.
By the time I left the agency, I regularly heard about new AI applications team members were exploring independently. That organic adoption was the actual success metric.
The initial skeptics converted when they observed their colleagues’ productivity gains. People stopped seeking permission and started experimenting on their own initiative.
For me, that’s what made the implementation successfull. People started to identify real problems, apply simple solutions, build momentum through early believers, and let adoption spread naturally.