AI for marketing: Stop automating the fun parts of your job (ft. AI use case discovery template)

Hands up if you (and the rest of your marketing team) have been here:

A team gathers, someone asks "what would you want an AI agent to do for you?" and within minutes the room is full of ideas. People want to be shouting their ideal marketing agent (and often, some non-marketing agents too) so the world can hear it.

And this totally  f e e l s  like a productive session. Except, nothing changes.

Enthusiasm behind dream agents has never been an issue. But when you're getting started with agents, you need to rethink the question.

When you ask people what they want AI to do, you're going to get wish lists; ideals or solutions in search of problems. This leads to agents being built that demo beautifully, but don't actually deliver anything for your team. But you want to start with the problem-fighters first.

Here at Optimizely, we've spent the last several months running AI agent discovery workshops across our marketing organization, and we've learned this the hard way. The highest-value use cases are almost never the ones people volunteer first.

The solution-thinking trap: Dream agents aren't necessarily the high-impact agents

When our team sat down with our field and events function, the first idea that surfaced was an agent to generate event booth messaging and giveaway ideas. Creative, useful-sounding, easy to demo. We almost built it.

...then we kept asking questions.

It turned out that after every event, the team was spending hours in Excel — manually tracking down lead conversation data from sales, enriching company information through research, reformatting spreadsheets, and uploading everything into downstream systems. Every. Single. Event.

The booth messaging idea was the art and writing. The Excel process was the laundry and dishes.

That's solution thinking, and it's one of the most common reasons AI pilots stall.

Introducing AI agents into your workflow: The right question to ask instead

Instead of asking "what should AI do for us?", ask: "where does work slow down, break, or fall on one person's shoulders?"

That reframe changes everything. It moves you from possibility space into problem space — which is exactly where high-value AI use cases live.

The framework we use to get there has three phases:

1. Map the work neutrally

   
   Pick one frequent, repeatable campaign or deliverable. Walk through every step it takes to get it from idea to execution to measurement. No editorializing yet — just map it. (Teams often tell us this part alone feels revelatory. Most of us rarely stop to look at our own workflows end to end.)

2. Find the friction

   
   Go back over the map and identify where something is too time-consuming, too effortful, or consistently below the quality bar. For each one, ask why. Then ask why again. Keep going until you hit the root cause — the actual problem, not the symptom.

3. Turn the root cause into a brief

   
   That problem statement becomes the foundation of your agent brief and product requirements document. A well-defined brief doesn't just help whoever is building the agent — if you're building in Optimizely, it gets you halfway there before you've written a single prompt.

What this surfaces in practice

Running this process across our marketing organization, we've now built over 100 agents actively in use — and a significant backlog of validated use cases we're working through.

Here's a sample of what came out of our of agent discovery sessions:

GA4 reporting agent: Every marketer has a complicated relationship with Google Analytics 4. Most either avoid it or create a bottleneck by relying on one person to pull reports. This agent pulls key metrics, generates executive summaries, and recommends next actions — including a version that automatically runs 30 days post-publish to flag whether an article is hitting benchmark and what to do if it isn't.

AI search visibility agent: Executives kept asking how the brand was showing up in LLM-generated answers. The information existed, but it was scattered and time-consuming to surface. We built an agent that connects to Profound, audits pages, and delivers recommendations for improving citation visibility in AI search — on a schedule, without anyone having to chase it down.

Competitive intelligence agent: A two-person team trying to monitor a crowded competitive landscape. No capacity for daily research. Now an agent runs weekly, scans competitors, summarizes recent news and activity, and delivers a digest every Monday morning so the team can plan the week from a position of actual awareness.

None of these would have surfaced from the question "what do you want AI to do?" The GA4 agent came from asking why reporting was always late. The competitive intel agent came from asking why the team felt perpetually behind. The Excel enrichment agent came from asking what happened after the event ended.

The real 'wow' moment in our agent discovery process

The best AI use cases in your organization are almost certainly already there — embedded in the slow, manual, person-dependent work your team has quietly normalized. They're not hiding. They're just not what people reach for first when you ask them to dream.

Change the question. Map the work. Find the friction. That's where the ROI is.

You can do it too with our AI use case discovery template (download it for free).