Your agent being "99% accurate" is why the client said no

​ I've seen a lot of agent demos go perfectly and still lose the deal with energy, pharma or banking clients coz they ask the one question most builders skip, what happens when the agent is wrong? the builder says 99% accurate, and the client hears one in a hundred. To them, an agent that's right 90 times and flags the other ten beats one that's right 99 times and quietly wrong once, because the second one is a liability they can't audit. What actually closes is the boring stuff, knowing when to hand off to a human, pointing at the evidence, same answer every time. Some decision models build abstaining in as a feature, bivia for example which is the shape these clients want. So how do you price an agent whose value is in the calls it declines to make?