PwC builds on the open CrewAI framework instead of buying off the shelf

PwC used the open-source CrewAI framework to orchestrate agent crews for internal generative AI work. Vendor-reported outcomes: code-generation accuracy climbed from around 10% to over 70% and turnaround on complex documents dropped sharply, which is what convinced skeptical consultants to adopt the tooling. The signal for smaller teams weighing the build-vs-buy choice is that even a firm that could buy anything chose an open multi-agent framework when it needed control over how agents collaborate. CrewAI reports its framework now powers over 12 million daily agent executions in production, so the build-your-own lane is not a niche.