You run five discovery agents. Every week they hand you five reports: support signals, NPS drivers, interview themes, journey maps, segment updates. Then what?
Then you sit with all five and try to find the connections by hand. Does the interview theme match the journey friction? Is the NPS driver the same thing the support tickets keep hitting? How do these rank against each other, and which one blocks the others? That is a couple of hours you rarely have on a Friday. The Opportunity Prioritization agent does it for you, every Friday at 11 AM.
Three layers of synthesis
It reads all five DISCOVER outputs and runs three passes over them.
The first is cross-dataset validation. If three reports point at the same issue, support signals and interview themes and journey friction, it is a high-confidence opportunity. One report mentions it and confidence drops. Strongest signals surface first.
Then impact estimation. Using segment size, willingness to pay, and engagement data, the agent estimates how many customers an opportunity touches, how much churn it would take out, whether it opens new revenue. Rough numbers, but better than the guess you would otherwise make.
Then dependency mapping. Does fixing A unlock B? Does solving slow onboarding depend on first fixing the confusing data mapping underneath it? The agent lays out the sequence so you prioritize in the right order.
Every item lands OST-ready: a 1 to 2 sentence problem statement, the evidence (which reports, which data points), the affected segments, an estimated impact, a rough effort size. Impact is concrete. Not "customers want X" but "40 customers want X, and it would reduce churn by 3 percent." It ranks the top 10 by (impact times segment size times willingness to fix) divided by effort, explains the rationale for the top 3, and flags the surprising ones, the patterns no single report caught on its own.
What it changes
You stop reconciling five reports by hand and start with one prioritized list: confidence scored, impact attached, sequence mapped. Consensus opportunities surface early, because three reports pointing at the same thing is real pain, not a hunch. You stop building things nobody asked for. And roadmap planning turns from a data-gathering slog into a decision meeting. You walk into Monday with the opportunities already validated across signals and the order already worked out.
One prerequisite: all five DISCOVER agents have to be running, because this one eats their weekly outputs. Any missing, and the synthesis has holes. Get those live first. Then schedule this agent for Friday and let it hand you a stack before Monday planning.