Julia — blind review

For Vince, Owners Best. Prepared by Dave.

Where we are

Julia has been handling incoming and recovery leads since late July, always with a person checking her work before anything goes out. Her code has been audited and fixed — she no longer loses your refinements, long histories are trimmed, one bad AI reply no longer kills a cycle, and she now sees the full buyer-sentiment picture. Her playbook was rewritten too: shorter, with your rules baked in, aimed at being warmer and more helpful.

How this works

14 real leads, picked at random. For each you see the buyer's situation, then two candidate responses, A and B. Both come from the same AI engine — the only difference is the version of Julia's instructions behind them. They are shown in a random order, so judge each purely on what it says.

How to score

Under both candidates, tap a number on each line (1 = poor, 5 = exactly right), tap one send option, and add a comment if you have one. Then pick the one you would actually send — that pick is the most important mark.

Nothing is required to move on, and you can go back and change anything until you submit.

Names, numbers and email addresses are partly masked. Some candidates name a specific negotiator, others just say "a negotiator" — treat that as a detail, not a clue.

…