A few weeks ago, we did something most agencies never talk about publicly: we handed a business their money back in full, walked away from a project we were proud of, and ended a ten-year relationship with a client who liked the design.

That sentence alone probably raises an eyebrow. Let us explain.

A client who trusted us for a decade

Our client is a photographer based in London. We first worked together ten years ago, when we created а complete brand identity: a logo drawn from а handwritten signature, bespoke stationery, an enduring tagline, and a website that went on to receive a Webby nomination and multiple international awards.

Ten years later, the site still looks modern. That’s rare in web design, where “current” often has a shelf life of three years at best. But the business had evolved: our client had grown, added new services, and needed a site that reflected where she is today.

She came back to us. Which, honestly, is the best compliment a design agency can get: a client returning a decade later because they still trust your eye.

It was love at first sight. Then came the feedback.

Our team put together a new design. We presented it. She lit up. Her words, more or less: "This is different. No photographer site looks like this. This is wonderful."

That's the reaction every designer hopes for: genuine surprise, not polite nodding.

Then, after the call, we received a feedback document.

Seventy pages of it.

Not seventy pages of our client's thoughts. Seventy pages of AI-generated critique, feedback that had been run through an AI tool and expanded into an exhaustive, granular list of things that could theoretically be different. Not necessarily better. Different.

Round two, and a plot twist nobody wanted

To our team's credit, we didn't dismiss the feedback. We went through the document, pulled out what seemed genuinely useful, and built a second design based on it.

The design was... fine. Technically competent. But it had lost the thing that made the first site so exciting, that unmistakable, "no photographer site looks like this" feeling. It looked like a polished design stifled by a checklist. Which, in a sense, was exactly what had happened.

Our client emailed back another forty pages of AI-generated design feedback.

Why we said no

Here's the thing about AI feedback tools, and it's worth sitting with: they are built to always find something. Ask an AI to critique a design, and it will critique the design — every time, without fail, regardless of whether the design is genuinely flawed or genuinely great. AI doesn't know how to simply say "this works." It doesn't get tired. It doesn't run out of suggestions. It will happily generate its ninetieth micro-improvement with the same confident tone as its first.

A human collaborator eventually says, "Yeah, actually, this is good." AI doesn't have an "actually, this is good" setting. It always has one more note. It never lets you have the last word.

And that's a fundamentally different dynamic than feedback. Feedback, at its best, is a conversation, someone bringing their taste, their gut, their lived experience of the client's world into the room.

What we were getting back wasn't a conversation. It was an infinite scroll of hypothetical improvements with no editorial judgment behind it, no sense of when "better" stops being better and starts being noise.

We could have kept revising. Somewhere around round four or five, we'd have burned through the budget, the goodwill, and almost certainly the very quality that made the first design great in the first place, because iteration without judgment doesn't refine work; it erodes it.

So we made the call: refund everything, and step away, rather than keep producing rounds of work we no longer believed in.

The part that's actually about our client, not the AI

We want to be careful here, because it would be easy to write this as "AI ruined a client relationship." That's not quite it, and it's not fair to our client.

She's a brilliant photographer and a genuine artist. And artists, more than almost anyone, know the difference between a critique that sharpens the work and a critique that just... keeps going.

We think what happened is something a lot of thoughtful, successful people are quietly running into right now: AI is an incredibly seductive sounding board. It's always available, it's articulate, it sounds authoritative, and it never makes you feel like you're being a pain by asking for one more pass. So it's easy to lean on it more than you mean to, right up until the tool that's supposed to sharpen your judgment starts quietly replacing it.

We think, given some space, our client will realize that too.

The lessons, for anyone hiring a design team

If you take one thing from this story, take this: great design isn't produced by exhausting every possible alternative. It's produced by knowing when to stop.

A few more specifics, for what they're worth:
● AI is a mirror, not a second designer in the room. It's excellent for stress-testing an idea you already believe in. It's dangerous as the sole voice deciding whether an idea is good, because it has no stake in the outcome and no instinct for "enough."

● Endless revision isn't rigor; it's often avoidance. At some point, more feedback stops improving a design and starts diluting what made it distinctive. The best creative decisions usually involve resisting a good idea's ninety-first tweak, not chasing it.

● If you're hiring an agency for their taste, let them use it. The whole value of a design partner, especially one with two decades of award-winning work behind them, is that they bring judgment you don't have in-house. Overriding that judgment with an algorithm that has no context of its own defeats the purpose of the hire.

● A good agency should be willing to walk away. Not out of ego, out of respect for the work. We'd rather refund a project than let it get watered down into something that doesn't represent either of us well.

Where this leaves us

We've had versions of this conversation with several clients lately — teams like Timmons, Perrill, and Marvell — about what actually differentiates a business in a crowded market. Our answer hasn't changed: it's design. Real, considered, human-judgment design. Not more design. Not endlessly optimized design. Just design that someone with taste and experience stood behind and said, "this is right."

AI has a place in that process. It doesn't have a seat at the head of the table.

Ready to take your brand further? Our New Jersey team is here to help. Tell us what you’re working on, and let’s build something great together. Say hello.