The Automations section in Text is now Playbook. The rename is the visible part. The bigger move is what it sets up: folding workflows and custom skills into one thing called a Skill. One person builds it, the rest of the team runs it. The AI agent runs it too.
Workflows and custom skills did the same job, twice
Until this week, Text had two ways to teach the system how to do something. Workflows for the team, custom skills for the AI agent. The catch is that a custom skill was a workflow underneath. Same outcome, different name, different builder, different mental model. If you wanted your AI agent to do something your team also does, you had to build it twice and learn two tools.
That is what Playbook is here to fix. Going forward there are no workflows and no custom skills. There are Skills.
Shipped this week
This first release is the rename plus a new home screen for accounts that haven't built anything yet. The home screen explains what Playbook is for and what a Skill can do, so a new customer isn't staring at an empty Automations tab without a starting point.
A new list view for the Playbook library lands in the next few days. After that, the bigger piece: a unified builder where any Skill, whether it runs on a person, the AI agent, or both, gets created in one place.
Same outcome, whoever takes the chat
Skills standardize an outcome, not a sentence. Detect the right moment, engage with the right message, present the offer, close the interaction. Whether the conversation runs through a person or the AI agent, the steps are the same and the outcome shows up the same way.
A Skill that re-engages a hesitant checkout visitor with a coupon should not behave differently depending on who picks up the chat. A Skill that qualifies a lead from a pricing-page question should not need two implementations. Playbook is the structural change that makes that possible.
For ecommerce teams running on Text, the practical effect is fewer broken handoffs between human and AI work, and one source of truth for the moves that drive revenue. Build the Skill once, attribute the outcome once, decide later whether you want a person or the agent in the seat.
What's coming
The unified builder is the piece that closes the loop. Once that ships, a team builds a Skill once, decides which conversations it should run on, and assigns it to humans, the AI agent, or both. Same logic, same measurement, same revenue attribution.
Get a summary with