The AI tools we actually use to build faster (and which ones we dropped)
Not every AI tool lives up to the hype. Here's our honest breakdown of what made the cut and what didn't.
We've tested dozens of AI tools over the past year, integrating them into our design and development workflows. Some have become indispensable. Others were impressive demos that fell apart in production. Here's our honest assessment.
The Tools That Made the Cut
Claude Code is our primary coding assistant. It handles everything from component scaffolding to complex logic, and it's remarkably good at understanding the context of a full codebase. We use it for rapid prototyping, debugging, and even writing documentation. It's saved us an estimated 15-20 hours per week across the team.
For design, we use Figma with AI plugins for generating placeholder content and exploring layout variations. But we've found that AI-generated designs still need significant human refinement to meet the quality bar our clients expect. The tools are great for exploration, not for final output.
Honorable Mentions and Dropouts
v0 by Vercel is excellent for quickly generating UI components. We use it as a starting point, then customize extensively. It's particularly useful for standard patterns like dashboards, settings pages, and data tables where the structure is well-established.
What we dropped: several AI copywriting tools that produced generic, unmemorable content. An AI-powered project management tool that added more friction than it removed. And a code review bot that flagged too many false positives and slowed our workflow down.
Our Philosophy
Our philosophy: AI tools should remove friction, not add it. If a tool requires more time to manage than it saves, it's not worth it — no matter how impressive the demo was. We evaluate every tool against one simple question: does this make our clients' projects better or faster? If the answer isn't a clear yes, we move on.
Share this article
