Website Redesign for a YC-Backed AI Startup

The Problem

Riveter had built something genuinely useful. Their product is an all-in-one web search agent that replaces the messy stack most AI teams end up building themselves (scrapers, SERP APIs, proxy rotators, the whole tangle). Solid product, real problem, YC backing.

The website wasn't doing any of that justice.

People were landing on the site, not really getting what Riveter did, and leaving. The demos that did come through were mostly the wrong fit, so the team was spending time on calls that went nowhere.

Here's what was actually broken:

1. The colors and typography felt flat and forgettable, which is a real issue when you're selling infrastructure and buyers are quietly asking "is this a real company?"

2. Use cases and industries were crammed onto the same page, so neither audience got a clear answer to "is this for me?"

3. Nothing on the site visually explained how the product works. A pretty technical proposition was being carried entirely by text.

4. The homepage read like a features checklist. There was no story pulling a visitor from "curious" to "book a demo."

For a YC startup trying to build trust with developers and product teams, the site was the weakest part of the funnel.

Our Approach

We started by figuring out who actually needed to be convinced. Two groups kept showing up: engineers evaluating a web data API, and product teams shopping for something to plug into their AI stack. The old site was trying to talk to both at once and landing with neither.

Once that was clear, we worked through the site in phases.

Rebuilding the IA. We pulled use cases and industries apart into two separate page systems. Use cases now speak to what someone is building. Industries speak to who they are. Same product, two different entry points, and the messaging on each finally fits.

Rewriting the homepage as a story. We laid out a new sequence for the homepage that walks people through the problem, why the usual stack breaks, what Riveter replaces, how it works, and proof it holds up. Every section is there to earn the next scroll.

Designing a visual system that feels serious. New color palette, new type system, and a set of custom 3D and 2D illustrations built specifically to show how Riveter works. We added light animation to bring the "one API, many sources" idea to life without needing a paragraph to explain it.

Building in Webflow. We developed the whole thing in Webflow with a proper CMS setup, so their team can add blog posts, new use cases, and industry pages without waiting on a developer.

What does a website redesign for AI startups actually involve?

A website redesign for AI startups is more than a fresh coat of paint. It means restructuring the site around how technical buyers actually evaluate a product, translating a complex backend story into something a non-engineer can grasp in a few seconds, and rewriting the flow so a visitor moves from curious to booked demo. For AI infrastructure companies especially, the site has to earn technical credibility and commercial trust in the same scroll.

Key Design Decisions

We chose custom illustrations over stock diagrams. Riveter's core value (collapsing a five-tool stack into one API) is basically invisible. You can't show it with icons pulled from a library. So we built a custom illustration system that shows scrapers, proxies, and search APIs being replaced by a single flow. Animation reinforces it. A visitor who spends ten seconds on the homepage now walks away with a picture of what Riveter actually does, not just a claim about it.

We split use cases and industries into their own systems. This was the single biggest fix in the IA. A founder building an AI research agent and a compliance team at a fintech have nothing in common except the API they'd use. Forcing them onto one page meant everyone got a watered-down message. Splitting them let us write sharper copy, use relevant proof, and place the right CTA for each group.

We treated color and typography as trust signals, not decoration.For infrastructure buyers, "does this look like a real company?" is a real evaluation step, even if nobody says it out loud. The old palette was working against Riveter on this. We rebuilt it with a deeper brand color, stronger typographic hierarchy, and more confident spacing. The point wasn't to look pretty. It was to look like something you'd trust with a production workload.

The Outcome

The founders told us the site did what it needed to do. The real win wasn't just more demo requests. It was better ones. The team stopped burning cycles on prospects who were never going to convert, and the site started doing the qualifying work for them.