Hello Reader
When I was the director of marketing for an events management company, we ran demand programs the way you pack for a trip at 5am, throwing things in and hoping you didn't forget something important.
Every Monday we'd look at the pipeline number, panic, and launch something.
By Friday we'd already moved on to the next thing without checking if the last one worked. If we missed our number, budget was on the line (and so was my job).
And that was years ago.
Now the pressure is even higher, because the CEO wants us to wave our AI wand and do it faster, with less budget and fewer people.
But AI on top of an unstable system just compounds the instability.
So before we talk about where to add AI, we need to talk about what powers your programs, and that's workflow design.
Where We Are Today
Before we get to workflow design it’s helpful to understand the new rules of demand generation.
As marketers we’re operating under three key shifts:
#1 Your Buyers Have Left the Funnel
Your buyers aren't sitting in one channel waiting for your email sequence to arrive. They’re bouncing between AI search tools, peer communities, private Slack groups, podcasts, LinkedIn threads, and conversations you’ll never see.
Obviously this is the short version. If you want more detail on the buyer journey and how it's changing, check out this edition from a few weeks ago.
#2 More Content Won't Save You
AI made content production cheap and fast.
Every competitor on your list can now publish more posts, send more emails, launch more ads and spin up more landing pages in a week than your team used to produce in a quarter.
#3 One Touchpoint Isn't Enough
Today, budgets are tighter, committees are larger, and risk tolerance is lower.
A single white paper download no longer earns you a demo. Buyers want steady proof across multiple touch points before they'll even take a meeting with your sales team.
We Were Productive and Still Losing
Let's go back to my prior events job.
Every quarter at that company started the same way: A whiteboard full of ideas, a rush to execute, and no conversation about how any of it connected to what we ran last quarter.
We were always building something new.
Here's a campaign strategy workflow I wish I'd taken the time to build back then.
In this workflow I'm responsible for aligning strategy and setting the KPIs before I collaborate with AI to build out the campaign themes and channel mix. My favorite part of this workflow is asking Claude to act as my target persona so I can test my assumptions.
Here's another one for content creation:
When I hand off to my team, they're responsible for starting with our brand guidelines to set the creative direction, then they collaborate with AI for drafting and scaling, but they're always close out the loop for final review and approval.
Adding AI to the Recipe
Once you've plotted out your workflows and cleaned up the process it's time to decide what you'll delegate to AI.
One of the first workflows I tackled was campaign analysis. If you don't know how your programs are converting, you have no way to know if you're helping the sales team hit their goals.
What you need is a performance loop that runs on a shorter cycle.
Here's what mine looks like.
After a campaign runs for two weeks, I export the performance data from Meta and Google and pull screenshots of every creative variant. I load it all into Claude (the data and the visuals together, because Claude can connect creative decisions to performance patterns).
Then I run it through a sequence:
First, I ask for a creative analysis. What's working, what's underperforming, and what patterns it sees across the variants.
Because it can see the actual ads, it can connect performance to creative decisions. In one recent campaign, it identified that the headline with a customer proof point was outperforming the office imagery variants by a wide margin.
Second, I have it dig into demographics. Which age cohorts are converting, where the spend is being wasted, and whether there's an anomaly it finds surprising. Last month it flagged that interactive content was outperforming the static ads.
If you want to get really fancy you can even ask Claude to build you an interactive dashboard.
The whole review takes about 30 minutes. Before this workflow, it took half a day and I still missed things.
Optimize all Your Workflows
You can apply this same approach to any workflow in your demand engine (or any other program!).
The pattern is the same:
Standardize the inputs, decide what you and your team handle, assign the rest to AI and run it on a cadence.
If we're talking about topic validation, that means scraping your sales call transcripts and Gong notes, feeding them into your LLM (minus the PII), and clustering the objections and recurring questions.
Every Monday, marketing reviews the top three objection clusters and assigns one to a campaign theme for the month.
For content production, that means maintaining a single reference table that maps your topic clusters to your ICP segments and active campaign themes.
When it's time to produce, you feed that table into your LLM with your brand guidelines and let it draft the full content kit (article, social post, email etc.) matched to the right audience and message. Your two-week production cycle collapses into an afternoon.
Here are a few more workflow ideas from marketers I follow on Reddit:
EntreprenuerRideAlong combines ICP lookalikes, LinkedIn profiles and ChatGPT to identify high converting leads.
Either_Bunch is layering AI agents on top of playbooks.
If online directories matter for your business you can try AntiqueDark’s approach.
Before You Add Anything New
When I think back to that events job, effort was the one thing we never lacked. Every campaign started from zero because we had no foundation and no clear workflow.
If I could go back, I would take the time to document our workflows, score them honestly and then decide where to add AI.
Does this feel familiar?
Use the scorecard below to see where your workflows are breaking. It's a great place to start and will show you exactly where to focus.