Guide: Agents

Share

Agents are AI Agents I have built to help in performing my role. Not overhype and overused, just simple helpful agents that can do things repeatably in a fraction of the time I could.

Agents published on the CTO Operating Systems are used for mainly research / analysis and information gathering. Examples are the Conversation Preparation Agent that prepares me to speak to executives, the CEO and board members of an organisation I am interested in discussing opportunities with, and the Panel Preparation Agent that hones my perspective on an up and coming speaking engagement.

Anything to do with research / analysis that I will repeat will typically be built as an Agent. These are relatively small and bounded agents, in that they have a defined task envelope, a limited set of tools and a predictable output shape.

Agents are published as markdown files which can be run as is, or can be adapted where necessary.

Sometimes I might run an agent multiple times, over a period of days, weeks even months, sometimes with a specific focus I give the agent. Due to this, agents typically produce output files that include a version number, timestamp and the name (organisation, product, person...).