Three Critical Lessons in AI Strategy for Marketing Leaders
There is only one reason it’s hard to build an AI strategy: it requires a new mental model.
AI has introduced a paradigm shift to outcomes-first-tech. When CMOs try to use the same methods and frameworks they’ve used in the past to include AI in their strategies, they meet some tough challenges.
We’re used to thinking about marketing tech as broadly fitting into one of two categories:
1) Storing data
2) Managing workflows
Just about every tool claims to drive outcomes, but few actually do. At least, not directly or completely.
So there’s really no example or framework we can pull from to wrap our heads around a new type of technology that — we believe — can finally deliver those outcomes we’ve been chasing.
Not in our tech stack, and not in the systems we use to plan our teams and processes.
So as a CMO, how should you think about building an AI strategy? Broadly speaking, there are three strategic pillars to consider:
- People - your team: the people doing the work.
- Processes - the series of steps and workflows they use to complete their work.
- Technology - somewhat obviously, the tools or platforms involved in those processes.
I’ve been fortunate to know people who I believe are leading the way through this paradigm shift for each of these pillars. Here’s what I’ve learned from them.
1. PEOPLE → Hire and Train for Understanding, not Mastery
Jason Widup said it best: you don’t need everyone to be a prompt engineer. You need them to recognize where AI fits, when it’s valuable, and when it’s not.
Prompt engineering is a great skill for now. But the fact that it’s needed at all is more a highlight of the shortcomings of the current technology than a long-term skill development opportunity.
When I was a kid, a school librarian taught me how to use search operators to build better google searches.
But nobody types “restaurants near me -fast food”. They just search.
We should expect prompt engineering as a skill to follow suit for most use-cases. It’s a band-aid for how early LLMs behaved, not a path towards using future versions well.
So don’t just hire expert prompt engineers. Hire people who understand how AI can be useful now and where it’ll be useful in the future.
Key takeaway:
Train your employees to think of their work as processes where any given piece could be delegated to an AI, and teach them how to decide for themselves whether each step should be.
2. PROCESS → Remove Friction, Don’t Just Add Automation
Ram Krishnan, Chief Commercial Officer at PepsiCo has a great blog called Ramalytics. He covers a breadth of topics, but often talks about AI as a way to elevate human capability, not replace it.
In the past, tech adoption was often efficiency-focused, like a digital production-line for the many tasks we need to complete to do our jobs as marketers. Like automation-based tools, AI can amplify efficiency. But it also offers more.
The paradigm shift requires us to think beyond the “doing more with less” approach of automation- or rules-based systems that helped us execute faster.
AI isn’t just something we can use; it’s something we can collaborate with.
That’s not something we could have said about marketing automation platforms, but is certainly true of AI.
Key takeaway:
Start involving a simple AI tool in your daily work and life. Choose any LLM and consult it about decisions you need to make. There’s no wrong choice: they’re all great.
Ask it to use frameworks you like (like jobs-to-be-done, or cost-benefit-analysis) to help you solve problems. Give it context and ask it for ideas, not just answers to your questions. Role-play with it by telling it to consult with you as if it were David Ogilvy or Ann Handley.
3. TECH → Solve the Context Gap
The latest version of Scott Brinker’s Martech Map lists 15,000 martech tools.
Most of them don’t directly integrate with each other. And why would they? It’d be absurd to build that many integrations.
But it does mean that each time we’ve added a solution to our marketing stacks, we’ve also added a bit of a problem: there are big decision-making spaces between our tools. Currently, people fill those gaps with csv exports and vlookups.
But some AI tools are built to establish the missing context, bring structure to it, and act on that unified picture of the truth in real time.
Read more: My Co-Founder and Pixis CEO, Shubham Mishra, just published an article last week digging into the context gap.
Pixis, for example, can understand ad performance data from both Meta and Google in real time, and use that data to act as a strategic partner, hone targeting, adjust bids, optimize ad spend across channels, and even generate new creative to test.
Key takeaway:
When you’re looking at AI platforms, consider whether they simply replace an existing point-solution, or if they would help all your existing point solutions become exponentially more valuable by filling the context gaps between them.
To be clear, there’s nothing wrong with the former, but seeing the difference between incumbent replacement and system uniter can be helpful when you’re evaluating new tools.
Shift Your Lens, Then Your Stack
Early adopters of AI will have an advantage. But balance urgency with a thoughtful approach.
Specifically, you’ll need to shift your mindset away from thinking, “how do I automate tasks?” to, “how can I orchestrate outcomes?”
Break these questions down further into people, process and leave tech for last. With so many existing martech solutions on the market, and new AI-powered tools popping into existence every single day, it’s easier to first establish clarity on your context gaps. Then set about testing tools that might help stitch them closed.
Do that, and you’ll stop asking whether AI belongs in your strategy, and start asking which outcomes you’d like to unlock next.
