A lead that messages at 11pm doesn't wait until morning. Failed contacts and after-hours enquiries leak conversion quietly, and no advisor roster can cover them. Rather than a big platform bet, I scoped a contained pilot: AI-driven SMS engagement on new leads, measured properly.
Two vendors made the shortlist. The mature option had the stronger security posture. The challenger was more than 10× cheaper, with a more flexible API and faster iteration. We compared them across ten technical dimensions with our engineering lead, and I made the uncomfortable call: take the challenger and close the security gap in the contract. I led the vendor through our security review and worked the protections into the agreement alongside legal.
The measurement design is the part I'd defend anywhere. All new leads across two partner programs, a randomised holdout control assigned at the record level, treatment versus control booking rate as the primary metric, downstream conversion as the check. The pilot is still running. When it reads out we will know exactly what the channel is worth, and the same evidence base feeds a build-vs-buy decision on an in-house engagement platform.