The best predictor of getting a second shot isn't success — it's the story you tell about your failure.
The Summary
- Ian Crosby raised $10M from Khosla Ventures for Synthetic, a fully autonomous AI bookkeeping service, despite his previous company Bench collapsing.
- Synthetic aims to replace human bookkeepers entirely with AI agents, targeting the startup market Crosby already knows intimately.
- The bet signals that VCs see founder learning curves as more valuable than clean track records when AI automation is the product.
The Signal
Ian Crosby's first act was Bench, a bookkeeping service for small businesses that raised over $60M before it fell apart. Now he's building Synthetic, which does the same job but replaces the humans with AI agents. Khosla Ventures wrote a $10M check to make it happen.
The timing matters. Bench failed because unit economics in human-powered services are brutal. You scale headcount, not software. Margins compress. Quality control becomes a nightmare when you're managing hundreds of bookkeepers across time zones. Crosby lived that pain for years.
"The best founders don't avoid their previous mistakes — they automate around them."
Synthetic is the answer to every problem Crosby couldn't solve at Bench. Fully autonomous AI bookkeeping means:
- No hiring ramp as you add customers
- Consistent quality across every account
- Margins that look like SaaS, not a services business
The product targets other startups, the customer segment Crosby knows best. That's pattern recognition from someone who spent a decade in bookkeeping hell. He's not building for dentists or law firms. He's building for founders who need clean books for their next fundraise and don't want to think about QuickBooks.
The Implication
Watch for more "second act, same vertical, AI-native" companies. Founders who failed at human-powered services have a decade of domain knowledge and a clear vision of what breaks at scale. They know where the work is repetitive, where mistakes cluster, where customers actually need speed versus hand-holding. That's the map for where to deploy agents first.
If you're building in a space you've already lost in once, you're not starting over. You're starting with scar tissue that looks a lot like product roadmap.