AI Consultant: Stop Clicking Through Apps—Your AI Agents Should Work for You

Sam Gaddis built an AI tool to help him and his wife fight better. It transcribes conflict in real time, tracks who is dominating the conversation, looks for destructive patterns, and can recommend a pause when things get too heated.

That deeply personal experiment is also the clearest window into how Sam builds with AI everywhere else: focused helpers, explicit access limits, and a person responsible for every consequential decision. In this episode, Michael sees how Sam uses Slack, email, meeting notes, repositories, nightly error checks, and voice capture to keep work moving without handing the system unlimited authority.

In this episode:

• How Sam’s relationship tool scores the way he and his wife handle conflict

• What the system learned when it showed Sam he was talking 60% of the time

• How the AI recognizes when a conversation is too heated and recommends a pause

• How Sam asks an assistant in Slack to find meeting notes, locate a client email, and prepare a reply in Gmail

• Why his default rule is “draft, don’t send”

• How Pointman gives his team a simple doorway into focused AI helpers

• How nightly checks identify failures and repair routine problems

• How Sam captures an idea by voice, researches it, and deliberately refines it before building

• Why his assistants can access his own software projects but not protected client work

Sam’s system is not one agent doing everything. It is a set of boundaries: draft before sending, personal repositories but not client repositories, fast capture before deliberate building, and a human still responsible for the relationship.

About Already Here: Michael meets people already living with personal AI in everyday life. We look at the tools, but the real story is what changes in how people work, make decisions, and direct their attention.

Subscribe for more conversations with people already living with personal AI. Then comment with the first safe experiment you would try.

Links mentioned in this episode:

https://runpoint.ai/

https://github.com/hermes-agent-org/hermes

https://openai.com/codex/

https://docs.anthropic.com/en/docs/claude-code/overview

https://www.granola.ai/