CharlieKey takeaways from your favorite podcasts
← 7 October 2026AI & I

We Were Wrong About AI Agents at Work

Every moved from personal agents for each employee to a shared company agent because personal ones were hard to maintain, and Willie and Dan expect a split between work and personal agents.

Watch on YouTube ↗

Charlie's summary

  1. Dots shine at filtering noise, so you get Slack's signal without being pulled into its rabbit hole

    Dan's dot handles school emails, missed Slack access requests and hands-free video edits by phone. His favorite use is filtering. While testing models over a weekend he posted results in Slack and his dot told him only about replies, so he stayed current without getting “sucked into the whole rabbit hole.” Rough edges remain: permissions are a mystery and need repeated approvals.

    ▶ 0:29
  2. Don't get attached to the cute agent character: Boo will be dead in a year, but persistent agents will stay

    Dan expects a persistent agent to be his default way of using AI in a year, but not as a named character. Drawing on his experience with OpenClaw and other named agents, he says people throw the cute persona away as soon as something better comes along. In a work context he mostly cares whether it does the thing. His own agent, Boo, has a lifespan of maybe six months.

    ▶ 2:39
  3. Seeing colleagues' agents in a shared Slack taught the team how to use AI

    Willie says that when OpenClaw agents moved into Slack, people could finally see how others prompted and used them. Before, they only saw outputs from private command centers on local machines. Showing people AI beats telling them. Willie found he learned something every time he looked in Slack, and this visibility made agents spread through the company.

    ▶ 8:56
  4. Personal agents faded because they were too hard to maintain, which led to one company agent

    Willie and Dan found that non-technical people didn't want to maintain their own claws, and tinkerers got tired once better options appeared. Building a platform on fast-moving, vibe-coded OpenClaw was messy, and their claws died. Meanwhile one agent, Austin's growth agent Montaigne, became the gravitational center. Pooling effort into one shared agent was better than everyone improving their own. Builders want knobs; operators running a business want something solid that just works.

    ▶ 12:41

Give your email to unlock the rest of the key points — for every episode.

Do more with Charlie

Free — sign in with your email, no password.