How I Actually Use AI in My Webflow Workflow Every Day.

It Started With a Client Call I Almost Forgot About

Last Tuesday I had a discovery call at 10 AM with a SaaS founder in London. I had prepped nothing. Zero notes. The call was in eight minutes. So I opened Claude, pasted the company's homepage URL and LinkedIn page, and asked for a quick brief: what they do, who their competitors are, and three smart questions to ask in the call.

Ninety seconds later I had a one-page brief that would have taken me 30 minutes to research manually. The call went well. The founder signed a $12,000 project two weeks later.

That is not a flex. That is just Tuesday. AI is not some futuristic concept in my workflow anymore. It is the equivalent of electricity. I notice it most when it is not there.

I am writing this because 92% of US developers now use AI coding tools daily. GitHub reports that 46% of all new code is AI-generated. Google CEO Sundar Pichai said over 30% of new Google code comes from AI. Gartner forecasts that 60% of all new code will be AI-generated by the end of 2026. This shift is happening fast, and I want to share what it actually looks like in practice for a Webflow developer working with real clients.

The Morning: Client Discovery and Content Strategy

Most of my mornings start with client communication. Briefs, proposals, content outlines, meeting follow-ups. This used to eat two to three hours of my day. Now it takes about 45 minutes.

I use ChatGPT Plus ($20 per month) for quick research and brainstorming. Competitive analysis, market positioning questions, content calendar drafts. The key is giving it enough context. I paste in the client's brand guidelines, their existing site copy, and their goals. Then I ask specific questions rather than vague ones.

For longer strategic work, like writing a full project proposal or a content strategy document, I use Claude. It handles nuance better and produces writing that sounds less like a robot wrote it. The 84% of Stack Overflow's 2025 survey respondents who said they use or plan to use AI tools are onto something. But the tool matters less than how you prompt it.

Meeting notes are another time saver. I record client calls (with permission), run the transcript through Claude, and get structured notes with action items in under a minute. Before AI, I spent 20 minutes after every call writing up notes. Now I spend two minutes reviewing what Claude extracted.

The Design Phase: Prototyping at Speed

I use v0 by Vercel ($20 per month for Premium) to generate UI components and page layouts from text descriptions. It is not a replacement for Figma or Webflow's visual editor. It is a starting point generator.

When a client says they want a pricing page with a toggle between monthly and annual billing, I describe it to v0, get a working React component in 30 seconds, and use that as a reference for the Webflow build. The 59% of agencies that report using AI to optimize web development operations are doing something similar. They are not replacing designers. They are eliminating the blank canvas problem.

For visual assets, Midjourney Standard ($30 per month) handles hero images, abstract backgrounds, and concept illustrations. A client needed custom illustrations for a fintech landing page last month. In the old workflow, I would have spent $500 to $800 on stock illustrations or hired an illustrator for $1,500 or more. With Midjourney, I generated 12 options in 20 minutes and refined the best three into final assets.

Generative AI cuts preliminary design discovery time by up to 70%, and that matches my experience. The creative direction still comes from me. The execution of exploration just got radically faster.

Development: The Code Editor Changed Everything

I switched to Cursor Pro ($20 per month) about a year ago, and it is the single biggest productivity change in my workflow. GitHub Copilot users report up to 55% more productivity writing code, and Cursor's AI-first approach takes that further for my use case.

When I am writing custom JavaScript for Webflow interactions, Cursor's inline completions save me from looking up syntax constantly. When I need to write a Webflow API integration, I describe what I want and Cursor scaffolds the function. I still review every line. I still test everything manually. But the first draft of code appears in seconds instead of minutes.

The real game changer is the Webflow MCP Server. This is an open-source tool (MIT license, completely free) that gives AI assistants like Claude and Cursor direct access to Webflow's capabilities through 10 agent skills. CMS management, accessibility audits, site health checks, publishing workflows. I use it with Claude Code to manage CMS items, run accessibility checks, and publish content directly from my terminal.

Before MCP, publishing a batch of blog posts meant logging into the Webflow dashboard, navigating to the CMS, creating each item manually, filling in fields, and hitting publish. Now I describe what I want, and the AI handles the CMS operations through the API. It is compatible with Claude Code, Claude Desktop, Cursor, and Windsurf.

100% of surveyed agency owners have included AI in their web design processes. That number sounds high until you realize that even using ChatGPT to write alt text counts. The spectrum of AI usage is wide, from casual assistance to deep workflow integration.

Content and SEO: Writing That Ranks

I use Claude for first drafts of client website copy, blog posts, and SEO metadata. Not because I cannot write. Because a first draft that is 70% there in two minutes lets me spend my time on the 30% that actually matters: the voice, the specifics, the parts that make it sound like a human wrote it.

For SEO research, I combine AI with traditional tools. I ask Claude to analyze a competitor's content structure, identify keyword gaps, and suggest heading hierarchies. Then I cross-reference that with Ahrefs or Semrush data. The AI gives me speed. The data tools give me accuracy. Neither works as well alone.

74% of developers report increased productivity with AI approaches. For content work specifically, I would put that number even higher. Writing meta descriptions for 40 pages used to take an afternoon. Now it takes 30 minutes, including my manual review and edits.

The Honest Truth About What AI Gets Wrong

I would be lying if I said AI made everything better. It does not. And the data backs this up.

The METR study from July 2025 found that experienced open-source developers were actually 19% slower when using AI tools, despite believing they were 20% faster. Read that again. They thought they were faster. They were measurably slower. The gap between perception and reality was nearly 40 percentage points.

Why? Because experienced developers spend more time reviewing, debugging, and correcting AI output than they save on initial generation. The AI writes code quickly, but that code is not always correct. A CodeRabbit analysis found that AI-generated code has 1.7 times more major issues and 2.74 times higher security vulnerability rates than human-written code. Roughly 24.7% of AI-generated code contains a security flaw.

I have hit this myself. Last month, Cursor generated a form validation function that looked perfect. Clean syntax, good structure, handled edge cases. But it had a subtle XSS vulnerability in how it processed user input. I caught it during review. A less experienced developer might not have.

This is the paradox. AI tools are most useful to developers who are experienced enough to catch their mistakes. The answer, for me, is selective usage. I use AI for boilerplate, scaffolding, and first drafts. I do not trust it for security-critical code, complex business logic, or anything that touches user data without thorough manual review.

Andrej Karpathy coined the term "vibe coding" in February 2025, and Collins Dictionary named it Word of the Year for 2025. The concept is real. But vibe coding without code review is how you ship vulnerabilities to production.

What This Stack Actually Costs Me Per Month

Cursor Pro runs $20 per month (credit-based since mid-2025). ChatGPT Plus is $20 per month. v0 by Vercel Premium is $20 per month. Midjourney Standard is $30 per month. Claude Pro is approximately $20 per month. The Webflow MCP Server is free and open source.

Total: roughly $110 per month.

For context, that $110 replaces what used to be $500 or more per month in stock assets, 10 to 15 hours of research and writing time (worth $1,000 to $1,500 at my rate), and significant context-switching overhead between tools and manual processes.

The ROI is not even close. $110 in, several thousand dollars in time savings out. Every month.

Why AI Makes Experienced Developers More Valuable

There is a fear in the industry that AI will replace developers. I do not see it that way. AI raises the floor of what anyone can produce. A founder can use Webflow's AI Site Builder to generate a multi-page site from a text prompt. That is genuinely impressive. But raising the floor does not lower the ceiling.

The ceiling, the quality of a site built by an experienced developer who understands brand strategy, conversion optimization, performance, and accessibility, that ceiling is actually higher now. Because I can move faster on the mechanical parts and spend more time on the strategic parts that AI cannot do.

A client does not hire me to write code. They hire me to solve their business problem through a website. AI helps me solve that problem faster. It does not replace the judgment, the experience, or the client relationship that makes the solution work.

If you want to talk about how AI fits into your next web project, or if you are curious about any of these tools, reach out. I genuinely enjoy these conversations. No pitch. Just a developer who is excited about where this is all heading.