Companies are putting machines to do work people used to do. Nobody really knows how often the machines get it wrong.

We believe no one should have to trust automated work based on hope, so we give it thousands of tasks we know the answer to, count how often the machines miss, how to increase quality, and what it costs to make the machines do their work better.

We don’t fix it. We tell the truth about how well they perform, and how you can make them perform better in the future.