Why AI made good agencies more expensive

There's a widespread belief that AI has made web agencies obsolete. What we see in our client conversations tells the opposite story.

The promise that didn't deliver

Over the past eighteen months a narrative has been everywhere. ChatGPT writes the copy, Cursor generates the code, Webflow AI assembles the page, Midjourney produces the images. So do agencies still matter? The implied answer was no.

That narrative produced a generation of websites that share a precise trait. They are technically correct, visually acceptable, and entirely unable to convert. Sections that follow each other without a reading hierarchy. Headlines that describe the product instead of selling an outcome. Responsive layouts that work on every device and convert on none.

This is not a tool problem. The tools did exactly what they were supposed to do. They generated output. The issue is that generating output was never the value of an agency. The value was the judgment that decides which output to produce, why, and when to stop iterating.

What's actually happening to prices

The numbers tell a story most people are reading wrong. According to Clutch, average hourly rates for U.S. web agencies moved from a 100 to 149 dollar bracket in 2022 to a 150 to 199 dollar bracket by 2024, with premium tiers consistently above 250 dollars per hour. Over the same period generative tools became dramatically more capable and cheaper.

If AI had genuinely commoditized agency work, we would expect the opposite movement. Falling prices, compressed margins, consolidation toward whoever charges least. The opposite is happening. Agencies that raised the level of strategic judgment they bring have raised their prices. Those that simply produced output are disappearing, but not because AI replaced them. They were replaced by clients who now produce that output themselves, and then turn to a more capable agency once they realize that isn't enough.

The twelve months of silence

The pattern has become predictable. A founder decides to launch a new product. They choose not to involve an agency. They use some combination of Webflow templates, ChatGPT for copy, Midjourney for images, a few no-code tools for integrations. In two weeks the site is live and it feels like a win.

For the first few months the metric they watch is completion. The site exists, it's fast, it's responsive. All good. Then they start watching the metric that matters. Traffic comes but doesn't convert. Pages get indexed but don't rank. The newsletter has more subscribers but fewer opens. The product is excellent but the website doesn't communicate it.

At this point two paths. The first is to keep iterating with the same tools, getting the same results. The second is to call an agency. When the call comes, the conversation has changed. They no longer ask "how much does it cost to build a site." They ask "why isn't my site working." It's a different question, and it carries different commercial weight.

We are receiving this kind of call at a frequency we had never seen before. Not from cold prospects. From founders who twelve months ago decided to skip the agency, and are now coming back because they understood empirically what they had been buying.

What AI is genuinely good at, and what it will never do

This part deserves to be said plainly, without defensive rhetoric. AI today is very good at specific things. It generates first-draft copy that only needs editing. It produces layout variations from well-formed prompts. It writes functional code for recurring patterns. It speeds up technical QA. It automates repetitive tasks that used to require hours of junior work.

All of this is real, and at Eclipse we use these tools every day. Webflow has a direct Claude connector, we are actively testing it, and in many cases it concretely accelerates the work.

But there is a list of things AI doesn't do, and not because it isn't capable enough yet. It doesn't do them for structural reasons.

AI doesn't decide what to remove. A site that converts is almost always a site where someone had the courage to remove ninety percent of what the client wanted to include. That decision means saying no to someone who is paying you, and explaining why the no is in their interest. No generative model will ever do that.

AI doesn't understand brand positioning before the brand even exists. Emerging brands have no conversion data, no benchmarks, no history of customer behavior. They have a thesis. Translating that thesis into a digital experience requires understanding intentions that haven't been articulated yet, not even by the client. This is exactly the kind of work generative tools cannot do, because they have nothing to generalize from.

AI doesn't take responsibility. When a site goes down during a launch, when a migration costs SEO ranking, when a Shopify integration breaks the checkout, someone has to answer for it. Not a tool. A person who put their professional reputation into the project. That is a service you buy, not one you generate.

The real shift for agencies

None of this means agencies can keep working the way they did ten years ago. The pressure is real. But the pressure isn't coming from AI itself. It comes from the fact that AI made visible to everyone the difference between producing output and making decisions.

Agencies that mainly sold execution (build this Figma into pages, develop this template, configure this integration) are struggling. That value has genuinely decreased, because a client with a decent prompt and a few days of patience can produce a passable version of that output on their own.

Agencies that sell judgment (what to build, why, in what order, with what trade-offs) have seen their value rise. That value cannot be replicated by a tool, and it becomes scarcer every month, because the volume of generic output keeps growing and the need for someone who can tell signal from noise grows with it.

We arrived at this conclusion empirically, not as a marketing choice. We looked at which projects we closed with healthy margins and which we didn't. Projects where we brought only execution became harder to close and tighter on margins. Projects where we brought strategic decisions, architectural choices, and brand positioning on digital, grew in budget and duration.

What informed clients are actually buying

When a founder with a growing product calls us today, they are not buying a website. They are buying a precise sequence of difficult choices.

They are buying the professional who decides whether this feature belongs on the homepage or on a dedicated page, and can defend the decision with reference to how early-stage visitors actually behave. They are buying someone who looks at the Shopify catalog structure and says half the categories need to go, because they dilute SEO authority without driving qualified traffic. They are buying someone who refuses to copy the layout of the successful competitor, because that layout worked for that brand, in that moment, with that audience, and replicating it just means arriving second.

They are buying, in other words, the judgment of someone who has seen hundreds of websites go well and dozens go badly, and can recognize the patterns before they become problems. This is exactly what AI makes more valuable, not less. Because anyone can produce an output now, but almost no one can tell whether that output is the right one.

The question that matters

The conversation we hear in first calls has shifted. It no longer starts with "what does your agency do." It starts with "I already tried doing it myself, I understood what I actually needed, now I want to do it properly."

This is the market position experienced agencies occupy in 2026. Not in competition with AI tools, but in continuity with them. Clients arrive after understanding what those tools can and cannot do. They arrive better informed, more demanding, and willing to pay more for a service they recognize as structurally different from what a sequence of prompts can produce.

AI didn't kill agencies. It raised the bar, eliminated those who only sold execution, and made strategic judgment the only real commercial asset left. For those who built that asset over the years, there has never been a better time.

See how we work with our clients.