I replaced six SaaS products with systems I run
I replaced Basecamp, Frame.io, DocuSign, Attio, GoHighLevel, and Calendly with software I operate myself.
| Previous tool | Current system | Primary reason |
|---|---|---|
| Frame.io | FreeFrame | Own the workflow and storage layer |
| DocuSign | Documenso | Control the signing infrastructure |
| Attio / GoHighLevel | Twenty CRM | One self-hosted customer record |
| Calendly | Cal.com | Control scheduling and integrations |
Basecamp was replaced by a Markdown-native AI workspace with Git as the source of truth, an in-browser editor, and an AI coding agent for every teammate.
AI reduced the cost of custom glue
Companies historically bought SaaS because custom software required too much engineering time. AI has reduced the cost of writing integrations, adapting interfaces, documenting systems, and maintaining small internal tools.
That does not mean every product should be rebuilt. It means the old build-vs.-buy calculation is stale for workflows that are expensive per seat, hold sensitive data, or force the company into a process that does not fit.
The decision should be made workflow by workflow, with operating cost included.
The license fee is only one line in the model
Self-hosting moves work back inside the company. Someone must run Docker, reverse proxies, TLS, backups, monitoring, updates, and recovery. Integrations still fail. Storage still grows. Security still needs ownership.
For my internal operating system, one MCP gateway handles access, routes internal AI tools, and audit-logs every tool call for cost attribution. That control is valuable, but it also creates a system I am responsible for operating.
The honest comparison is total cost and strategic control, not subscription price versus zero.
Own the workflows that make the business different
Buy commodity capabilities when a reliable product fits. Own the layer that contains your operating logic, your sensitive data, or your margin advantage.
For a data- and process-heavy company, that may be the system that assembles estimates, reconciles accounts, manages precedent, prepares a lender package, or dispatches field work. The model is replaceable. The workflow and its history are the asset.
AI moved the build-vs.-buy line. The useful question is not “Can we build it?” It is “Which layer is worth owning?”