What Need Is There for a Web Agency in the Age of AI?

It's a fair question. AI can generate code, write copy, produce images, and scaffold a functional website in the time it takes to finish a cup of coffee. So why would you pay for a web agency?

I get asked some version of this regularly.

The Question Answers Itself

When someone objects to the need for a web agency by saying AI can just do it, I don't argue with them. I ask them a few questions back: Have you done it yet? Do you plan to? What's your plan? How will you know whether what you built is actually effective?

If you can answer all of those clearly, you probably don't need us. And that's fine.

Most people can't. The questions tend to trail off, and the real problem comes into focus. Building a website is the easy part. Building one that works requires knowing what "works" means, and being able to verify it. That's a different skill set from being able to prompt your way to a homepage.

The Floor Has Always Been Low

There has never been a meaningful technical barrier to building a website. HTML and CSS are learnable in a weekend. JavaScript took a little longer, but the tooling has only gotten friendlier.

AI raises the floor further. You can get further faster with less knowledge than at any point in the history of the web.

But this has happened before in adjacent fields. When digital cameras put a camera in every phone, everyone became a photographer. When desktop publishing put page-layout software on every laptop, everyone became a graphic designer. When the web went mainstream, everyone became a web designer. In each case, the technology democratized access to the entry level.

What it didn't do is democratize expertise.

Professional photographers that get noticed are still drawing on first-principles knowledge about light, composition, and the relationship between subject and frame. The designers producing work that moves people still understand visual hierarchy, contrast, and why a reader's eye goes where it goes.

Web work is no different. It has always been possible to build a site without knowing what you're doing. AI makes that even more accessible. And it doesn't change the question of whether what you built will actually work.

What We Actually Sell

The agency's job is to be the part of the process that knows the principles.

Information architecture: how people navigate, where they get lost, what a site's structure communicates before a user reads a single word. Usability and universal usability: who can actually use this, and on what devices, with what assistive technology. Principles of design: the science of attention and comprehension, the mechanics of where a reader's eye goes and why. SEO and accessibility: how search engines read a site, how screen readers encounter it, what technical gaps quietly undermine both. Effective user interfaces: the difference between a form that gets filled out and one that gets abandoned.

These are not things a client can check with a prompt. They require someone who has internalized the principles well enough to evaluate the output, catch what's wrong, and fix it.

That's the gap between someone who can build a website and someone who can guarantee the website does its job.

Three Things You Get

When someone asks what they're paying for, the answer has three layers.

First-principles expertise. We know what makes a site actually work. We're applying principles that took years to internalize, so you don't have to build that knowledge from scratch.

Your time back. Your time is worth something. Managing a digital presence, staying current with platform changes, troubleshooting accessibility failures, validating that your AI-generated code is correct: all of this takes more time than it looks like from the outside. We do that work so you can put your attention where it belongs.

Effectiveness assurance. When AI generates something for you, it generates something plausible. Plausible and effective are not the same. We can tell the difference. You get to ship with confidence.

The Grey Goo Problem

There's a deeper issue underneath all of this.

AI-generated web work at scale is converging. The aesthetics are blurring together. The copy is starting to sound the same. If you've spent any time reading AI-generated content, you know the texture. There's a kind of generically competent sameness to it, and it's spreading across the web at speed. (You may have already noticed glimpses of it in this article. ;-))

This is the grey goo of generative AI content. Everything technically works. Everything looks approximately right. Nothing is particularly yours.

The question for any organization, ministry, or business working in this environment is: how do you keep your voice, your brand, and your actual personality in your product or your messaging while still leveraging AI? It requires someone who understands both your particular identity and the principles of effective communication. The tools don't supply that.

We build, and we preserve what makes you distinctly you. We use AI where it saves time and adds leverage, and we do the work by hand where AI would flatten you into something generic.

Maybe you don't need us.

Maybe you don't need us.

If you know what you need to build, you have the time to learn what you need to learn, you can verify that the output is effective, and you're confident you can maintain it as the landscape shifts, then you're right. The tools are good enough that a capable, motivated person can build something that works.

If any of that is unclear, here's what working with us looks like: your time back, a team that knows the principles cold, and confidence that what you've built will actually do its job.

That's the offer.

How This Was Written

This article was partially written using AI.

I started with a verbal voice dump: a free-association, conversational capture of my thinking on this topic. My AI team then compiled that with other observations and knowledge I've accumulated in my personal knowledge base over time, pulling from scattered notes and connections I'd already made, and assembled it, and then I made a final pass by hand to edit it into the piece you just read.

This is exactly the distinction the article makes. My thinking, voice, expertise, and observations were the source. The AI assembled and articulated. That's what it looks like when AI amplifies the work you've already done. The grey goo problem exists when AI is the source. When you are the source and AI is the tool, the result is still distinctly yours.

The wrong version of this is easy to picture, because the shortcut is genuinely available. You open a chat window and type: "Write me a 1,000-word article on why businesses still need a web agency in the age of AI." Or maybe you've thought about it a little more: "Help me brainstorm some angles on why web agencies are still valuable, then turn those into an article." The AI produces something plausible, maybe even good-looking, and you publish it. The thinking that was supposed to happen before and during the writing never happened. The clearest version of this is the student who drops an essay prompt directly into an AI, accepts whatever comes back, and submits it. The assignment was to do the intellectual work, and the document was supposed to be evidence that the work happened. When the AI generates the document, the work gets skipped. What you have is output without formation.

That's the line. AI as assembly tool for thinking you've already done is one thing. AI as a replacement for the thinking is something else entirely.

There is a category of work I do entirely by hand and will always do that way: writing a pastoral letter to someone in a hard season, or preparing a sermon. In that work, the process is the point as much as the product. Something happens in me when I sit with a text or a person's situation and work through what to say: attention that sharpens, convictions that clarify, the slow work of being formed by the labor. Handing that off would skip something I don't want to skip. The letter or the sermon might look the same from the outside. The person writing it would be less formed for having avoided the work. Some tasks carry that weight. For those, the old ways are still the way.

One more thing worth saying: this article has AI tells. My teenage girls could spot them. I think I got rid of most of them in my final editorial pass, but that's why the transparency section exists. Without it, those tells would quietly undermine the credibility of the piece. If you can tell that AI generated the output, what's to say that it didn't generate idea and the content as well?

Matt Heerema is the founder of Mere Agency, a digital presence agency serving ministries and mission-driven organizations.