Link: To Succeed With AI, Give up Control

Greenlite has created digital compliance workers for regulated banks and fintechs. The agents are trained on the same content that you would use to train a new compliance analyst; a human manager checks the agents’ work. The agents work more quickly, more thoroughly and are more auditable than humans because they automatically log their work. Greenlite’s AI agents are safer because they reduce both the frequency and severity of errors. By taking the employee out of the loop to complete work, AI agents increase return on investment. Today’s AI assistants and copilots measure success in terms of improved productivity of the employee using them. While productivity gains enable the worker to do more work more quickly, the worker still needs to be involved. And any productivity gains are heavily dependent on how the worker uses the new technology. However, productivity is just a means to an end. What we actually care about is delivering work, which is essentially multiple discrete tasks strung together with reasoning and process. Work as an output can be measured by a dollar saved or a dollar earned to the bottom line. For example, a recent MIT study looked at customer support copilots and found that a generative AI assistant improved productivity by an average of 14%. Among novice and low-skilled workers, the effect was as large as 35%. In contrast, AI agents can reproduce 100% of an employee’s output. For example, Klarna recently reported that it has replaced 700 customer service agents with AI. The AI agents are trained on the same frequently asked questions as human agents and have access to all prior customer-agent interactions. The AI agents are more likely to offer correct answers and less likely to prompt follow-up questions from customers. The impact? Klarna says the switch will boost its profit by $40 million this year. #

--

Yoooo, this is a quick note on a link that made me go, WTF? Find all past links here.