Set performance targets for AI agents the same way you do for human salespeople. The AI then refines its own prompts and code to hit those targets, but a human reviewer has to approve every change before it goes live.
Obviously, the targets, targets are quite cool here as well because we can also give the AI targets. And based off that, the AI can improve itself in terms of the prompts it has or in terms of the structure or what it does, its actual code. And the C-suite or whoever you appoint to that can also, before it does that, obviously approve or deny those, those changes, those things that it wants to improve. With that, we create a self-improving loop, meaning we not only set targets for the actual SDRs, but we also set targets for our AI. And based on those, based on those targets and what it sees, it can improve, improve itself pretty much.
