A speculative near-future architecture, not yet built or actionable per Chandra, in which separate AI agents divide the Present Forward: A Five-Phase Process for Problem-Focused Strategy pipeline: a strategy agent, a roadmap/feature agent, and an engineering agent that communicate with each other to iterate on strategy automatically rather than through a human working group.
Framed as sitting on the far side of a 'crossover point where human judgment will be inferior to something that's able to process multiple signals simultaneously' — a threshold Chandra expects but doesn't think has been crossed yet for strategy work. Flagged explicitly as a medium-term direction to watch, not a current recommendation.
Relates to AI Era Bottleneck Shift: From Building to Deciding and Orchestrator Layer for Narrow Agents.