“Can we make this ad hit harder? Make it more relevant?”
There’s probably a version of that sentence somewhere in your creative feedback. Perhaps you wrote it. Perhaps someone sent it to you, and you’re now trying to work out what to ask the designer for.
The ad looks fine, the branding is right, the offer is visible, but something about it feels too generic, too busy, or “too easy to scroll past.”
You can recognize the problem well enough to want a change—explaining that change might be where things get a little foggy.
In his article on ad creative testing with AI tools, Shivansh explores visual changes we can test, including making a setting more locally relevant and removing distractions from an image. Those possibilities are useful. But if you’re managing the work, there’s another question sitting underneath them.
How does this fit into what the team is already doing?
Someone still needs to decide what to change. Someone has to make it, review it, and work through the comments. If AI adds another folder of images and another round of approvals, I can understand why you might be reluctant to introduce it.
The promise of faster production only helps if it also makes the work around production easier.
That’s the part I want to walk through here. We’ll take one existing ad, work out what we want to learn, and use an AI mockup to help explain the change to a creative team. There’s a prompt you can copy along the way, and a short brief you can return to when you’re ready to try it.
Making room for a smaller first attempt
Trying something new at work rarely happens when your calendar is empty. Campaigns are running, deadlines are approaching, and people already have ways of getting things done.
A team may also have perfectly reasonable concerns about AI. A designer might wonder whether they’ll inherit an image that takes longer to repair than to create. A marketing manager might wonder whether learning to prompt is about to become another part of their job.
Those concerns belong in the conversation.
For a first attempt, I would keep the scope small enough that everyone can see what they’re agreeing to. One image that already exists. One change we think might help. One rough reference to make that change easier to discuss.
It could start with a conversation like this:
“I’m having trouble explaining the background change I’m picturing. Would a rough AI reference help? I’d like to try it on this image and see whether it makes the brief clearer.”
That gives your designer a chance to say what would actually be useful. Perhaps a mockup would help. Perhaps an annotated screenshot would be enough. Either response tells you something about how to support the work.
There’s room to try this without deciding, in advance, that it needs to become the new process for everything.
Getting from a feeling to something we can test
Let’s imagine we’re working with a plumbing company. The current ad shows a technician repairing a kitchen sink. Appliances sit on the counter, patterned towels hang, and the usual collection of things seems to accumulate in kitchens regardless of anyone’s intentions.
The technician is visible, but so is everything else.
Our first reaction might be that the ad needs more impact. Looking more closely, though, we notice that our attention keeps going to the counter before it reaches the repair.
That gives us a place to start.
“I’m wondering whether there’s too much happening around the technician. Could we try simplifying the background so the service is easier to recognize?”
We haven’t established that the background is causing poor performance. We’ve identified something worth testing. That distinction matters because it leaves us open to being wrong.
It also gives the designer more to work with than “make it pop.” (We still shudder reading those words…)
At this point, I’d spend a moment on what should stay the same. It’s easy to skip this because we tend to explain a revision through the things we want changed. But the technician’s position, lighting, camera angle, and offer all help define the comparison we’re trying to make.
If those change too, we may end up with an appealing new ad and very little understanding of which adjustment helped.
For this version, the question is narrow: does simplifying the background make a useful difference?
That’s enough to work with.
Giving the idea a shape your team can see
Even now, there’s room for interpretation. “Simplify the background” could mean removing a toaster. It could also mean turning the kitchen into a beautifully empty room where no one has ever made lunch.
An AI mockup can help us find the middle ground.
If you have permission to upload and edit the existing image, it makes a useful starting point. Working from it keeps the discussion connected to the asset the team already knows.
For our plumbing example, this is the prompt I’d start with.
A prompt to copy and try
Edit the attached plumbing-service image to create a rough visual reference for our creative team.
We want to explore whether reducing background clutter makes the technician and sink repair easier to notice.
Remove countertop appliances, patterned towels, and loose household items that compete for attention. Keep the kitchen believable and lived-in, with a simpler background rather than an empty room.
Preserve the technician’s appearance, pose, clothing, tools, sink, lighting, camera angle, and framing. Keep any existing branding unchanged, and do not add text.
The only intended change is reduced background clutter. Please create one version.
For your own ad, the parts to replace are the subject, the change you’re exploring, and the details that need to remain consistent. The useful structure is already there: what we’re trying to understand, what should change, and what should stay put.
Then we get to see what the tool understood.
It may be close. It may also remove the appliances and quietly remodel the entire kitchen. The background could look better while the technician acquires a rather questionable hand.
A quick comparison with the original helps us notice those departures before they become part of the brief.
This is also where we can lose the time we hoped to save. One more prompt. One more adjustment. Perhaps another version with slightly different lighting.
I’d keep returning to the purpose of the image: can the designer now see what I mean?
If the answer is yes, a note may be more useful than another generation.
“The countertop is close to what I’m picturing. Please keep the technician and lighting from the original; the AI has changed them here.”
The reference can be imperfect and still help the conversation. If it’s so far off that you need several paragraphs to explain what to ignore, it probably needs another attempt, or a simpler way of showing the change.
Turning the reference into a brief people can use
There’s a small but important difference between sending an image and explaining why you’re sending it.
An AI mockup arrives with plenty of detail. Some of it is intentional. Some of it may be the tool making decisions on your behalf. Your creative team shouldn’t have to guess which is which.
I’d send the original alongside one selected mockup, with a note that makes its role clear:
“This is a rough reference for the amount of background cleanup I have in mind. The original is still our source for the technician, tools, lighting, and composition. I’d appreciate your judgment on keeping the finished scene natural.”
Take a look again at that last sentence. The designer may see a problem with the reference that you haven’t noticed and they might have a better way to achieve the same thing. The brief gives that judgment somewhere useful to go.
Once we’ve worked through the idea, recording it doesn’t need to become a separate writing exercise. Most of what belongs in the brief is already in the conversation.
The creative brief to come back to
Campaign goal
Test whether simplifying the image helps people recognize and respond to the plumbing service.
Asset to update
[Link to the current kitchen repair image.]
What we’re wondering
The appliances and household items may be competing with the technician for attention. A simpler background could make the service easier to identify.
Requested change
Reduce the background clutter around the technician and sink.
What stays the same
The technician, pose, tools, sink, lighting, camera angle, framing, headline, offer, and overall brand tone.
How to use the AI reference
The mockup illustrates the amount of background cleanup we’re considering. Please use the original for everything we’ve identified as staying the same, and flag anything in the reference that would be difficult or unhelpful to reproduce.
First deliverable
One revised image in [required dimensions], due [date]. Once the direction is approved, we can confirm the remaining placements.
Feedback owner
[Name] will bring the team’s comments together.
How we’ll assess it
[Paid media owner] will confirm the comparison and primary success measure before launch. For this example, we’re interested in click-through rate while also watching cost per qualified lead and lead quality.
There’s no need to introduce a new form if your team already has a good brief. These details can live in the task you’re already using. The part worth carrying forward is the connection between the question, the proposed change, and what you’re asking the team to preserve.
You can save this section for the next request. Next time, much of it should feel familiar.
Keeping the review connected to the reason we started
The first version comes back. The background is quieter. The technician is easier to see.
Then someone suggests warmer lighting.
Someone else thinks the kitchen should look more premium. Another person wonders whether we should change the headline while we’re in there.
I can see how each suggestion happens. Once there’s a new image in front of us, we start noticing possibilities. The difficulty is deciding which of those possibilities belongs in this revision.
This is where the brief earns its place in the task.
We can return to what we intended to change, check that the other elements stayed consistent, and assess whether the image works for the brand and placement. Those questions give the review a purpose beyond whether everyone likes the new version.
They also make feedback easier to explain.
“It still doesn’t pop” leaves the designer to interpret it. We can get closer to the issue by describing what we’re seeing:
“My eye still goes to the bright appliance on the left before I notice the repair. Could we remove that and keep the original lighting? I think that would get us closer to the background change we discussed.”
The more premium kitchen may still be worth exploring. We can keep it as a possible next test without asking the current image to answer another question.
Having one person gather the comments helps too. If one reviewer wants the background emptier and another wants more household detail, that disagreement needs a conversation before it becomes a set of instructions.
Otherwise, the designer ends up trying to resolve a decision the rest of us haven’t made.
Finding out whether it helped
By this stage, we can learn two things.
First, about the working process. Did the mockup help the designer understand the request? Did it reduce the amount of explanation needed? Or did the AI introduce enough confusing detail that it made the handoff harder?
Those are worth asking directly. A useful answer might be, “The image helped, but next time just circle the part you mean.” That gives us something specific to improve.
The second is about the ad itself.
An approved image gives us something to test. Before it goes live, the paid media owner needs to confirm how the comparison will work, which metric will guide the decision, and what evidence would be enough to support a conclusion. Simply uploading both images doesn’t guarantee a controlled comparison.
Once results come in, a short note in the original task can preserve what happened. The new image may have performed better on the measure we chose. Perhaps it didn’t. Perhaps the evidence is still inconclusive.
More impressions alone won’t tell us that the change worked better.
Keeping that note alongside the brief means the next request can start with what we learned. The designer gets to understand what happened to the work, and we’re less likely to revisit the same idea months later without realizing we’ve already tried it.
Giving it a first attempt and a fair second one
If you’re still unsure whether this belongs in your workflow, I’d start with a revision you already need to request. Something small enough that a rough reference might help explain it.
You can copy the prompt above, adapt it to that image, and see what comes back. If it communicates the idea, the brief is here when you need it. If it doesn’t, that first attempt will at least show you where the explanation or the tool fell short.
I’d leave room for a second try when there’s a clear adjustment to make. Perhaps the first prompt asked for too much. Perhaps the designer needed a note about what to preserve. Trying again with that feedback gives you a better basis for deciding whether the practice is useful.
It also gives your team a say in how it develops.
If someone else shares responsibility for creative requests, this could be a useful article to send them before you try it together. You’ll have the same example, prompt, and brief to work from, and something concrete to discuss afterward.
The question I’d bring to that conversation is simple: did this make it easier for us to understand one another and move the work forward?
That’s a worthwhile place to begin.

