01What the Hermes AI Kanban Update Does
The Hermes AI Kanban update is a new feature in the Hermes Agent Kanban that lets you give one prompt and the AI does the rest. You drop a prompt into the triage, and the orchestrator agent reads it.
The orchestrator breaks your prompt down into every subtask needed to get the job done. Then it assigns the right agent profile to each task based on what that agent specializes in. You used to do this manually, but now the system handles it for you.
This update also lets you write a description for each agent profile. That description tells the orchestrator what the agent is good at, like research or copywriting. The result is a tighter workflow with fewer errors and less manual work.
02Key Features of This Update
This update brings two big features that change how Hermes Kanban works.
- Automatic task breakdown: One prompt becomes a full list of subtasks without you typing each one.
- Agent assignment: The orchestrator assigns each subtask to the best agent based on the agent's profile.
- Agent descriptions: You write a short brief for each agent, like "this agent is great at research" or "this one handles outreach." The orchestrator reads these and uses them to route tasks.
- Sequencing: The orchestrator thinks about dependencies. It knows research happens before writing, so it orders tasks correctly.
Together, these features remove the manual handoff that most AI tools still need.
03How to Set Up Agent Profiles and Triage
Setting up the new Hermes Kanban update takes a few steps. Julian recommends a practical three-step path to get started.
- Audit your weekly workflows. Write down five to ten recurring workflows, like campaign launches or content updates. These become your triage prompt templates.
- Set up your agent profiles with real descriptions. Spend five minutes per agent. Be specific. For example, "writes engaging short form content for online communities" beats a generic "content agent."
- Test your triage prompts. Drop a real workflow into triage and check how the orchestrator breaks it down. If routing looks off, sharpen your agent descriptions.
This setup takes a couple of hours once. You just drop prompts and review outputs.
04Using the Update in Real Workflows
Julian shows a real example in the episode. You run a community and want a new member onboarding campaign. You drop one prompt into triage: "Plan and execute a new member onboarding campaign."
The orchestrator breaks that into subtasks: write the welcome sequence, design the onboarding checklist, draft announcement posts, set up follow-up messages. Then it reads your agent descriptions and assigns tasks. The copywriting agent gets the emails.
The strategy agent gets the campaign plan.
Another workflow uses three agents: research, content, and engagement. You prompt: "Create a weekly content update about the latest AI tools." The orchestrator sends a subtask to research to pull updates, to content to turn that into a post, and to engagement to draft replies to expected questions. All from one prompt.
05Results and Comparisons from the Episode
Julian compares the new update to older AI tools. Most tools still need a human in the middle. You prompt, get output, then move it somewhere and assign the next step.
Hermes removed that handoff.
The orchestrator acts like a project manager who never sleeps. You tell it the goal, and it figures out who does what. You just check results.
The agent descriptions do more work than they look like. Before this update, the orchestrator guessed based on names or generic setup. Now it knows from your brief.
The more specific your descriptions, the smarter the routing gets.
No formal test numbers appear in the episode, but the practical example shows a full workflow running without manual steps.
06Tips, Limits, and Mistakes to Avoid
Julian warns that vague triage prompts get vague results. The orchestrator is smart, but it needs something to work with. A good prompt says: "Research the top five AI productivity tools released this week, write a short summary of each, and create three social posts." A bad prompt says: "Do something with AI news."
Spend two minutes writing a proper triage prompt. The whole downstream workflow depends on that first input.
Also, do not be lazy with agent descriptions. The better your descriptions, the better the routing gets. Skipping this means you will be stuck reassigning tasks manually forever.
- Do not leave prompts vague.
- Do not skip agent descriptions.
- Do not expect perfect routing without testing.
- Do not ignore task sequencing; it matters for dependencies.
07Who This Update Is For and Why It Matters
This update is for anyone running a content operation, community, agency, or service business. If you spend hours every week breaking projects into tasks and figuring out who does what, Hermes just automated that.
Julian explains that the orchestrator handles the coordination layer. You focus on the actual work, not the routing. The agent descriptions make the system smarter every time you add detail.
The logical next step is predictive triage, where the system suggests tasks before you ask. Julian says that is not science fiction; it is the next version of what Hermes already does. People who set up their agent profiles now will have a serious advantage as the system gets more powerful.
08Setup Path and Future of Hermes Kanban
Julian gives a clear setup path: audit your workflows, write agent descriptions, and test triage prompts. This takes a couple of hours once, just dropping prompts and reviewing outputs.
Every triage prompt you run teaches you to write better prompts. Every agent description you refine makes routing smarter. You are compounding your efficiency.
People who wait will be three months behind and scrambling to catch up. Hermes is improving fast, and the agent descriptions you write now determine how good your automation gets over the next few months.
The update makes Hermes Kanban close to a real autonomous workflow system. One prompt, full breakdown, agents assigned, correct sequence. You just review the output.
