SystemC is the most common language used for modeling transaction level (TLM) behavior of hardware. From the beginning of TLM users have been exploring how to perform functional verification at the fast TLM level, and hopefully reduce functional verification at the RTL level. The timing might have been a bit premature and theoretical, since it's only very recently that the majority of the market has adopted sophisticated RTL functional verification approach such as the Open Verification Methodology (OVM - see www.ovmworld.org) and the Vera Methodology Manual (VMM). Now that users are fully embracing these advanced testbench approaches for corner cases at the RTL level, and designs are growing (sigh - same old issue!), the topic of leveraging TLM for functional verification is pressing.
There are several ways people I talk to approach this topic: tool budget, verification project schedule, coverage, use cases, technologically, and methodologically. These are all valid considerations, and there is no single way to consider the problem. However, one philosophy seems to be commonly held by our customers. If you approach the question of balancing TLM and RTL functional verification using a "verification goals" point of view, existing methodology and technology can be employed.
A recent article talks about top level verification planning. This is a notion first introduced in a commercial EDA product several years ago by Verisity. Since then the industry has realized the need to approach functional verification from the perspective of what is being accomplished, rather than from a tool-centric or language-centric perspective. This same goal-orientation will play into the eventual industry best practices that will be developed for employing more and more TLM functional verification in the project flow. The speed of TLM verification is too valuable to avoid for long. And verification planning and management approaches are needed to not only make the decision about verification goals up front, but keep projects on track through their complicated milestones to achieve high quality closure.