Customer story: Envisioning
How Envisioning runs its client work on Starlings.
Envisioning is a research institute. Its work is calls with clients and partners, and the research, documents and follow-ups around them, with AI agents doing a growing share. All of it now happens in Starlings.
Before.
Client calls ran on Zoom and Google Meet. A notetaker wrote the notes, chat lived in Slack, and tasks and documents lived in Notion and a task tool. Each tool worked. The history of a client was spread across all of them, and every agent needed a connection to each one and still had to guess which client a note was about.
What the team retired.
- Zoom and Google Meet for client calls
- A meeting notetaker
- Slack
- Notion and a separate task tool
After.
Clients join calls from a link in their browser. The host’s Mac transcribes the call, and when it ends, Starlings files the transcript, a summary and suggested tasks on the client. The chat, the tasks, the documents and every past call sit on the same record, so anyone who opens a client sees where things stand.
The jobs agents do.
Each one works from the same client record as the team, and Starlings logs every action under the name of the person who asked.
Call briefs
Before a client call, a member asks Claude for a brief. Claude reads the last calls, the open tasks and the decisions on the client through MCP, and files the brief there.
Tasks
Shared agents running on Vercel Eve sit in the team list. A member hands one a task in Work, and the agent reports back on the task.
Documents
Agents draft proposals, briefs and follow-ups from the client’s history and file them in Docs, where the team reviews them.
Code
Repositories are mapped to projects. Agents pick up GitHub issues on them, and every change shows up in the project’s thread.