Build vs. Buy: Integration Strategy in the Age of AI
Coding agents made writing code cheaper. But did they actually make building integrations easier?
What you'll learn
A year ago, building a native integration to a payroll provider took weeks of engineering. Now an agent drafts it in an hour. But the hard part of unifying employment data was never writing code. It was the walled gardens, data mapping, and auth configurations within deeply fragmented ecosystem.
These challenges are still at the heart of the build-vs-buy debate, even with AI in the mix.
This session examines the benefits and tradeoffs of each approach by walking through what it takes to build an integration from scratch, step by step.
Takeaways
- Securing API access comes before code, and is often the hardest part of building an integration
- The domain specifics that makes mapping HR, payroll, and benefits data uniquely difficult
- Why "shipped" is the wrong definition of done
- An implementation engineer's perspective on the factors that drive teams to buy (and a live demo of Finch's Integration Skills)
Who should watch
Founders, CTOs, and VPs of engineering and product at B2B software companies whose products depend on employee, payroll, or benefits data.
"We could build these ourselves, but if it's going to take us two to three years to build decent market coverage and we can do it in two months with Finch, why wouldn't we?" — James Fletcher, Thoropass