We propose a new framework for ant-based routing algorithms which addresses the trade-off between effective and timely discovery of optimal routes, and the desire to keep data traffic flowing smoothly. The approach utilizes a database of stored delay estimates on routes to build separate routing tables for ants and data packets. This decoupling results in better control of responsiveness to load or topology fluctuations since the tables can evolve at different rates, according to different rules. The presented architecture also features guaranteed loop-free routes for data packets.