Use Frontier Models as Planners, Not Typists
The product-builder move is not to replace every model with the newest frontier release. It is to route each part of the workflow to the model that should own that job.
The strongest practitioner signal around Fable was not "it writes code faster." It was that builders were using it as a planner and orchestrator: decomposing ambiguous work, deciding what matters, and supervising cheaper execution paths.
This is how AI products become economically sane. Frontier models are best used where judgment, decomposition, and long-context tradeoffs matter. Routine implementation, extraction, formatting, and verification can often be handled by smaller or specialized models if the plan and acceptance criteria are strong.
Split your next agent workflow into four roles: planner, implementer, reviewer, and fallback. Write the expected output for each role. Then decide which model should do each job based on measured quality, latency, and cost instead of brand preference.
Measurement decision record
For one workflow, create a routing table with columns for task, model, cost ceiling, quality check, and fallback. The exercise will show whether you are designing a system or just calling an API.
Open the related Product Analytics exercise →Keep Going