Build vs buy: when to license an agent, when to build one

For every agentic use case on your roadmap, ask: (1) Is the workflow this agent automates a source of differentiation, or is it table stakes? (2) Is the underlying data proprietary to us, or industry-shared?

If the workflow is table stakes (case triage, lead routing, meeting prep), buy. The Agentforce skill catalog and partner ecosystem will out-iterate your internal team within a quarter. If the workflow is genuinely differentiated (a pricing agent that reflects your commercial model, a service motion that's a brand promise), build.

The most common overbuild we see is service triage. Teams scope a custom triage agent because their categorisation 'is unique'. It almost never is — and twelve months later, the buy path would have outperformed.

Reserve build capacity for the two or three agents your CEO would put in an earnings call. Buy or configure the rest.

The mirror failure is buying an off-the-shelf 'sales agent' that does generic deal coaching when your commercial motion is genuinely complex. If MEDDPICC, your discount matrix, or your channel rules are how you compete, the agent has to know them. That's a build.