In 2023, one of the biggest calls I made in marketing was introducing AI into the workflow. It sounds simple now. At the time, it wasn't obvious which tools were worth the effort and which ones just looked good in a demo.
I talked through this on camera for The MarTech Summit's Lightning Interview series. Here's the version with the "um"s taken out.
Why bother
The market doesn't hold still. Staying competitive meant adopting the trends that actually mattered instead of watching from the sidelines. So the team looked at AI and machine learning, mapped where they could realistically help, and checked that against real business objectives and the resources we had. No point picking a tool because a competitor uses it.
That mapping pointed to two clear time sinks: content generation and data analysis. Both ended up being where AI paid off.
Start basic, then get specific
A lot of the early research turned up tools that looked impressive but didn't fit our goals or weren't practical to run. So I started with the basics: Grammarly, ChatGPT. Once those were embedded in daily work, I moved on to more specialized tools: Jasper, Scalenut, SEMrush.
The lesson that stuck: not every well-reviewed tool fits every business. Cost, ease of use, and scalability matter more than the feature list. Test small, then scale what actually works.
The result was a 37% jump in marketing efficiency and effectiveness. That's the number that made 2024 planning easier, because it gave us a reason to go deeper into AI across other parts of the business.
The part that doesn't get talked about enough
Here's where the "AI output nobody trusts" problem shows up. Teams tend to land on one of two extremes: avoid AI because you don't trust it, or ship its output straight through because it's faster. Neither holds up.
AI has to be supervised, not just switched on. Someone has to check that the final product holds up on quality, and that it doesn't step on anyone's data, security, or legal footing. That responsibility sits with the marketer running the tool, not the tool itself. Nobody else is going to catch it before it goes out.
The job isn't "use AI." The job is "use AI, then take responsibility for what it produces." That's the whole difference between speed that helps and speed that creates a mess to clean up later.
Where this goes next
This is the same principle I still run on: build the workflow so AI does the heavy lifting, then put a human on the last mile before anything ships. It's why I don't just automate content production, I review it. Same logic applies to nurture emails, campaign copy, and reporting.
