For Ash·Privately built·May 2026 Sanskai·Dev Batra

Tech Labs Capital, on its own terms.

InDebted has the phrase Third-party AI sitting on their own stack page. Almost nobody in SaaS writes it that way. Five others here are on the same line. Here's where.

From indebted.co — published
Machine Learning · Strategy · Third-party AI
indebted.co ↗
Precedent Self-built ML, layered Latent · open surface

Same shape as what we'd build for the portcos — but the customer is us.

Two layers running in the open on our own studio domain. Same build pattern: AI inside one workflow, custom software around it.

Same build pattern we'd run for any portco: a small AI layer doing one job well. Custom software around it. Both live on theucu.com — our studio side, separate from client work.

The layer founders are right not to build themselves.

There's a layer most B2B SaaS founders are right to leave off the core roadmap. The customer-facing AI surface — voice, generative copy, conversational match, churn-risk. It sits between the product and the moment of use. A real category, just one most founders shouldn't put on the critical path.

InDebted drew that boundary themselves. Publicly. They wrote Third-party AI into their own stack page — kept the ML decisioning in-house, marked the customer-experience layer as something they outsource on purpose. Five others on this list are on the same side of that boundary.

One workflow first. Software built around it. If we did the first part well, the next layer gets asked for. The founder keeps building the engine; we build the surface around it.

If any of these read real to you, I'll run the discovery with the founder. Whichever portco.

Dev
dev@sanskai.com