Build or buy agents: Lindy, Make, n8n, Sana — what makes sense for an SMB
The build-vs-buy question changed shape in 2026. It’s no longer an IT decision — it’s an ops decision. Here’s the criterion I use with clients.
Over the last twelve months the agent market has split into two families:
- Ready-made agents — Lindy, Siit, Fin — arrive with the use case solved.
- Builders — Make, n8n, Gumloop, Sana — hand over the parts, you assemble the fit.
For an SMB, "build or buy" has stopped being philosophical. It’s a practical decision with four variables.
Buy when…
- The use case is industry standard. Customer support, scheduling, HR triage — if hundreds of companies do this the same way, someone has solved it better than your team will.
- You need to be running tomorrow. A ready agent installs in hours, not weeks.
- You have no internal technical capacity and don’t want to create that dependency.
Build when…
- The process is a differentiator for the business. If the way you serve customers is what sets you apart, don’t automate it with the same agent your neighbour uses.
- You need to cross proprietary systems (custom ERP, internal database, old integrations nobody else has).
- Volume justifies engineering. Past a certain scale, per-execution pricing on a ready agent starts to hurt.
The classic mistake: buying without a process
Many SMBs buy a ready agent before having the process written down. Result: the agent does what it knows how to do, the team reshapes the process around it, and three months later the process is worse — not better — than the original.
The right order: design the process first (enough for a new person to run it from a document), measure the before, and only then bring the agent in. Bought or built, but always with a reference to compare against.
The hybrid approach that became normal in 2026
The pattern that emerged: buy a ready agent for the reference process and use a builder (Make or n8n) for the connections around it — pull CRM context, push the result to the billing system, notify Slack when something goes off script.
It’s neither buying nor building. It’s knowing when to stop buying and start gluing.