The real challenge is what happens next: make something existing work better. That's where every other AI tool breaks down. Tempo aims to be the AI design tool that finally helps teams collaborate on code the same way they work together on designs.

In the early days, bringing this product to market proved challenging. As Tempo improved, competitors like Lovable and Bolt were always one step ahead. Under intense pressure to find growth, the company launched Agent+, an agency service where expert designers and developers built software for early-stage startups. They leaned on Agent+ as an extension of their runway and a learning engine to discover how teams could work together directly on code—tweaking, refining, building on what already exists.

To accelerate learning, we banned Figma—the tool every designer depends on—and committed to using our own AI product for all in-house and agency design work, even though it wasn't ready to replace Figma. This experiment revealed harsh truths: what should have unlocked amazing creative exploration showed serious problems instead. Tempo could implement existing concepts but couldn't support the creative exploration designers actually needed. This discovery shaped our hybrid approach, combining AI with direct visual control rather than relying solely on text prompts.

But we had another problem. Enterprise customers still weren't convinced about using Tempo for their daily work, even when we built features that looked and worked like tools they already used. At first, we thought they just needed more features to feel confident making the switch. But the real problem was much simpler: teams were afraid of letting AI touch their existing code, worried the AI might break something that already worked. The question became: how do you convince teams to trust AI with their most important work?