Pinned post
The information age is over. This is the age of GTM intelligence.
1:46
Pinned post
Claygents are way behind... Zoominfo is significantly better out of the box.View quoted post on X
More connectors, skills that sync with GitHub, and observability
Claygents or always on agents should be able to use functions
Agent observability is incredibly bad in most GTM agents for users now. If clay makes it easy for users to headlessly debug agents they can overtake many other agent platforms
Think about how to use OTEL from Claude cowork and chatGPT
Agents need a dedicated HITL layer or UI to be broadly useful
Agents, workflows, and always on agents should have a business goal and metrics that they optimize towards
Clay should have a semantic and outcome attribution layer that sits on top of their data warehouse similar to databricks and snowflake
Clay should embrace a zero copy pointer model to sit on top of client data warehouses like palantir does to win big dogs
There should be a way to build simple claygents on top of functions inside the mcp by reps and have the reps debug them, and promote them into always on agents
Extracting context from a website is ok at best. You need call transcripts to get good at this.
The context layer should really sit on top of the data warehouse and have dedicated subtypes. IE ICP isn't just a thing, it's a semantic extrusion of the account object with a relationship to products.
Need a way to build predictive models in platform
Workflows shouldn't exist in platform by end of next year, something like @bot should babysit a version of current workflows instead and learn on top of them
Clay should unlock a push to deploy
Clay needs agents workflow and function as code parity with deepline
Clay really needs some form of backend git for agents, skills, workflows, functions, context and possibly memory to win
Make clay easy to use and debug in grokbot, duet, and dot