Every marketing team I talk to right now is using AI for something. Content drafts, ad copy variants, chatbots, Claude skills or routines to automate repetitive work. Almost none of them can tell me whether it’s making them money.
I’m not trying to harp on the tools. What I really want to do is talk about a measurement problem.
The gap nobody wants to name
Here’s what I see over and over when I step into a business as a fractional CMO: the marketing team’s tech stack has quietly absorbed four or five AI tools over the last eighteen months, usually one at a time, usually because someone on the team found something useful and just started using it. Nobody made a formal case for the spend. Nobody set a baseline before flipping it on. So when leadership asks “Is this working?” the honest answer is “It increases my productivity!” but no one can quantify a return on investment that the CFO would be happy with.
The irony is that AI tools are often more measurable than some of the marketing channels we’re using. We’ll happily run a sponsorship or a conference booth for years on vibes, but we hold a $200/month AI tool to a standard of proof we never applied to anything else. That double standard is worth noticing, but I digress; let’s get back to the topic at hand.
While most marketing teams have a really great AI tool list. They fail to set up the right framework to measure the results.
A framework that actually holds up
If you’re going to defend AI spend in a board meeting, or decide whether to renew the monthly expense, you need to be able to answer three questions for every tool:
1. What is it replacing or accelerating, specifically? Not “content creation” in general but the exact task. A tool that cuts first-draft time on blog posts from four hours to forty minutes is replacing a measurable amount of labour. A tool that “helps with strategy” is not, and you should be suspicious of any AI ROI claim that can’t point to a specific task. From there, what are you now doing with those three hours and twenty minutes to improve the business?
2. What did the baseline look like before? This is the step everyone skips, and it’s the one that makes the rest of the math possible. Before you adopt, capture: cost per unit of output (per email, per ad variant, per qualified lead), cycle time, and the downstream metric that unit was supposed to move (conversion rate, reply rate, MQL-to-SQL rate). Without this, “it feels faster” is the best you’ll ever get.
3. Did the downstream metric actually move, and by how much relative to the cost? I’m not asking about vanity metrics but the ones connected to revenue. A tool that produces twice as many ad variants isn’t a win if your conversion rate on those ads didn’t budge, because you’ve just increased production cost with no output change. A tool that shaved two days off a campaign launch cycle is a win if speed-to-market is genuinely tied to a revenue outcome for your business. Has it been for yours?
Where this breaks down in practice
The teams that struggle here usually have one of two problems. Either their attribution is already broken before AI enters the picture. If that’s the case, you won’t have any clean baseline to compare it against. Or you’re measuring AI tools against each other instead of against the pre-AI process. Comparing Tool A to Tool B tells you which AI tool is better. It tells you nothing about whether AI was the right call in the first place.
If you don’t have a working marketing mix modelling (MMM) or even a decent multi-touch attribution setup, fix that first, please. AI ROI measurement is downstream of general marketing measurement maturity.
The honest version of this conversation
I’d rather tell a client “We can’t prove this is working yet, here’s what we need to fix first” than let them keep renewing a tool on faith. That’s an uncomfortable conversation in the short term and a trust-building one in the long term, and it’s usually the conversation that turns into a real marketing operations rebuild, not just a tooling decision.
If your team is running AI tools without a clear before-and-after on the metrics that matter, that’s usually the first thing I look at in a marketing tech audit. It’s rarely the tool that’s the problem. It’s usually that the scaffolding hasn’t been built before the workers show up on Monday.
Need an outside set of eyes on your marketing stack and what it’s actually returning? Schedule a call to talk through a marketing tech audit.
