Agents and people in one run: a weekly blog post
Chapters
Every week this team publishes one blog post. A research agent finds the search question, a writer agent drafts, an editor reads every draft, an image agent makes the figures, and a person schedules what goes out. Nobody drew that plan as a flowchart first. The agent running the job decided each step, and Fivexer recorded what happened.
Agents join like people do
An agent is connected to the workspace with its own skills, such as Keyword research, Writing or Images. The router gives it work it has the skill for, the same way it gives work to a person, and it only counts as ready while it is connected. Editing stays with people, so drafts can only reach someone with that skill.
The run, in order
Agent runs lists every goal an agent pursued. Open one and you see the goal at the top and each step below it in the order it happened. Every step says who did it, whether that was a person or an agent, why the agent handed it out, and what came back.
A person said no, and the record kept it
The editor read the first draft and sent it back with notes. The agent recorded its decision to rewrite, based on the editor's answer, and the second draft went back to an editor who approved it. Click any step to see why it went to that person and the score that decided it, and to check what came back against the answer the agent asked for.
Save it as a workflow
When a run works, Save as workflow turns it into fixed steps with the same rules. Who does each step is still picked fresh every time. A case the run never saw, such as a second draft that is not approved, stops and asks a person. Next week your agent starts the same steps as a workflow.
Agents keep only the outcome on the record by default. This team chose to keep everything its agents report, so the editor's notes stay readable on the run.
Read more about agentic orchestration with people and agents, or how workflow orchestration fits around it.
Transcriptthe full narration, 7 paragraphs
Every week, this team puts out one blog post. Three agents and one editor did the work, and nobody drew a flowchart first. This is the run they left behind.
Agents join the workspace the way people do, each with the skills it takes work for. This team has a research agent, a writer and an image agent, and all three are ready. Editing stays with a person.
Open the run. The goal sits at the top, and every step below it is in the order it happened. Each step says who did it, a person or an agent, and why it was handed out.
This is the step to watch. The editor read the first draft and sent it back, with notes on what to change. The agent recorded why it chose a rewrite. The second draft went back to an editor, and this time it was approved.
Click any step to see why it went to that person. The panel shows who was chosen and the score that decided it. Below that, it checks what came back against the answer the agent asked for.
When a run works, keep it. Save as workflow turns it into fixed steps with the same rules. The person or agent for each step is still picked fresh every time. If the second edit ever comes back without approval, the workflow stops and asks a person. Save it, and next week your agent runs the same steps as a workflow.
Agents do the work and people keep the final say, with the reason for every step on record. Sign up free at 5xer.com.