A website redesign quote from three years ago read like a construction estimate. Hours of design, hours of development, a chunk for copy, a line for QA. Project managers padded the copy line because somebody had to write 40 pages of fresh prose. Designers billed for mockup rounds because every variation was hand-pushed in Figma. Developers quoted boilerplate build-out hours the same way a framer quotes studs: so many for the nav, so many for the footer, so many for form components.
That quote looks different now. Some lines are a fraction of what they were. Others haven't moved an inch. And a handful of lines didn't exist on anybody's rate card two years ago.
If you're the one approving the budget, the real question is which pieces moved, and why — the total dropping is almost beside the point.
Why Does the Quote Look So Different Now?
Short version: AI absorbed the mechanical work, and the judgment-heavy work got more expensive. First-draft copy, first-pass mockups, and boilerplate component code used to eat real hours. Those hours now compress into prompting, editing, and review. The strategic work around them — figuring out what the site should do, who it's for, and how it should behave — carries more weight because everything around it got cheap.
MotoCMS has a useful walkthrough of how AI rewrote the website redesign budget, laying out where savings land and where new line items are showing up. The pattern matches what buyers see on real quotes. The total may be flat, but the composition is scrambled.
Which Line Items Are Shrinking?
Three categories are getting cheaper on almost every quote. Knowing why helps you push back when a vendor hasn't passed the savings along.
- First-draft copy. Writing a serviceable first pass for a product page used to be a copywriter's billable afternoon. Now it's a prompt and a cleanup. The copywriter still earns their fee, but the fee is for voice, accuracy, and the parts a model can't invent.
- First-pass mockups. Designers can generate layout options, hero variants, and component states in minutes rather than hours. The exploration phase that used to require three review cycles often collapses into one. Expect the design line to drop, not disappear.
- Boilerplate build work. Standard components — forms, nav patterns, card grids, footer blocks — are the sweet spot for code generation. A senior developer still assembles and reviews, but the keystroke work is largely done by the tool.
Which Lines Are Holding Steady?
A different set of line items hasn't moved, and in some cases they've grown. These are the places where AI doesn't help much, or where it introduces risk that someone experienced has to catch.
- Strategy and UX research. Interviews, journey mapping, information architecture, and the hard conversations about what a page is for — none of that compresses. Buyers now expect it to be sharper because the production phase is faster.
- Accessibility. Compliance isn't a generated artifact. WebAIM's 2025 analysis found that 94.8% of home pages failed WCAG 2 conformance, and shipping AI-generated markup without a real audit tends to make that worse.
- QA and performance tuning. Automated tests catch regressions, but they can't tell you whether the site feels fast on a mid-range phone over spotty Wi-Fi. Core Web Vitals work — LCP, INP, CLS — is still a specialist line item, and this primer explains why page experience signals keep mattering to search visibility.
- Discovery and scoping. The hours spent deciding what to build before anyone builds anything have gotten more important. Faster production means a vague brief gets you a lot of polished, wrong work.
What's Brand New on the Quote?
A few line items would have been incomprehensible on a 2022 quote. They're the ones most likely to be missing from a cheap bid, and the ones most likely to blow up a project in year two if they're skipped.
- AI features in the product itself. A site search that answers questions, a chat assistant on a product page, generated recommendations — each one carries its own scope, model, prompt, evaluation, and failure modes. Scope it as a build inside the build.
- Content governance. If marketing is going to generate drafts weekly, somebody has to decide what gets published without a human read, what the voice guardrails are, and how corrections propagate. That's a document and a workflow, and it belongs in the redesign scope.
- Structured data for AI search. Answer engines read a site differently than a 2015 crawler did. Schema, clear entity markup, and clean semantic structure now feed both traditional results and AI-generated summaries. Teams that skip this watch their referral traffic erode as answers get served without the click.
- Evaluation and monitoring for AI outputs. When a feature generates text or decisions, something has to grade it over time. A test suite for a chat endpoint isn't the same artifact as unit tests for a form handler, and it needs its own budget line.