I used to spend the first 30 minutes of every workday just figuring out what to do. Calendar in one tab, Slack pinging me, Asana staring back with a wall of tasks. So when I saw a walkthrough of ChatGPT’s new Workspace Agents, I got genuinely excited. The creator behind the video broke down how these agents can take that messy morning routine and turn it into a clean, prioritized plan, and honestly, I was hooked from the first demo.
Most teams build agents the hard way. They wire up Zapier, drag nodes around in n8n, write instructions by hand, and connect every app one by one. Workspace Agents flip that script. The author shows that you just describe what your team does in plain chat, and ChatGPT builds the agent for you, including skills it invents on its own.
The old way vs the new way
Here’s the contrast that stood out to me as the expert walked through it:
- 🛠 Old way: open a builder, write a system prompt, manually connect APIs, define tools, set schedules, debug.
- ⚡ New way: type “plan my day from Calendar, Asana, and Slack,” hit send, and watch the agent assemble itself, pick apps, write its own description, create memory, and add guardrails.
The original poster pointed out something I keep thinking about: these agents even write their own skills. You don’t need to learn the skill format. The agent identifies what it needs, builds the skill file, and stores it. That’s a real shift in who does the configuration work.
What Workspace Agents actually are
The creator explains them as the evolution of Custom GPTs from 2023, but with a lot more muscle. Each agent can:
- Connect to apps like Google Calendar, Slack, Asana, HubSpot, Notion, Adobe, Intercom, Dropbox, and more through MCPs.
- Build and store its own skills as MD files with input, workflow, and boundaries.
- Keep a persistent memory folder for notes, drafts, and outputs across sessions.
- Run on a schedule or on demand, inside ChatGPT or Slack (with more channels coming).
- Be shared across a team, listed in a directory, or sent as a link.
One small but important note from the walkthrough: as of recording, these are only on the Business, Enterprise, and Education plans. The regular paid ChatGPT plan doesn’t have them yet. Business needs two seats at around $25 per person.
The morning planner demo
This was the moment that sold me. The author clicks the morning planner template, which fires off a prompt like “create an agent that looks at my calendar, tasks, and open Slack threads and turns them into a clear plan for my day.”
What happens next is wild. The builder:
- Picks the apps it needs (Calendar, Asana, Slack).
- Sets a schedule for 8 a.m. daily.
- Names the agent and writes its own role and description.
- Creates a to-do list for itself and checks each item off.
- Adds memory, safety rules, and guardrails automatically.
When the expert runs it, the output is a clean morning brief: top three priorities, schedule conflicts, risks and blockers, follow-ups he owes people on Slack, and a recommended plan for the rest of the day. The catch he flags clearly: the better organized your calendar and tasks already are, the better the plan. Garbage in, garbage out still applies.
How to build one yourself
Following what this industry pro showed, here’s the path I’d take:
- Open the agents panel on the left in ChatGPT (paid Business or Enterprise plan).
- Hit the plus sign to create a new agent.
- Skip the blank “agent builder” option. Use the chat description instead, it’s faster.
- Describe a repeatable task your team does, in detail.
- Let ChatGPT propose the apps, schedule, channels, and capabilities.
- Connect any apps it asks for through the Browse Apps section.
- Test it with “Try it in chat,” tweak instructions, then share with your team via the three-dot menu.
Use cases worth stealing
The creator demoed several templates and patterns:
- Design partner agent connected to Adobe Photoshop, Notion, and Asana.
- Marketing strategy agent that built its own content engine, content reviewer, lifecycle, and marketing intelligence skills.
- Customer service agents that handle repetitive support tasks.
- Daily planners that pull from Calendar, tasks, and chat.
Why I think this matters
This savvy professional made a point I keep coming back to: most agent platforms ask you to learn their interface. Workspace Agents ask you to describe your work. The configuration burden moves from human to model. That’s a meaningful change in how teams adopt automation.
Watch the full video for the live builds, the share flow, and the skills deep dive. Worth your time if you’ve been on the fence about agents.