AI agents need two things to work: eyes to see and understand external data, and hands to act on decisions through code, APIs, or system control. Without both, agents can't reliably decide what to do or execute it.
Give your agents eyes and hands. There's a lot of talk about AI agents, and a lot of this has to do with the fact that, you know, these large language models are able to take all sorts of interesting forms of information, usually turned into tokens, and then turn those into code that can express into the world in fairly generalizable ways. But in both cases, in order for it to be useful, you need to ask how it's going to connect with the information source, i.e., context, and how it's going to connect with actually producing action. I I call that eyes and hands. How are you applying eyes and hands to your AI agent technology? How does it know what to do? How is it able to get the information it needs in order to be able to make a decision? And then what can it do with that decision? What agency does it have? Those are the hands. When you put those two things together, you wind up with technology that can actually make a big difference in the world.
