This lesson is being recorded
Every video of this course is being uploaded to my YouTube channel. Subscribe so you do not miss it.
Bruno Ashford @brunoashfordSubscribeModule 7 · Lesson 7.5 · 12 min
Designing Your Agentic Org Chart
Map human roles and AI roles side by side so everyone knows who does what, who approves, and who is accountable.
Use ← → to move. The marker highlights text and the pencil draws on the slide; navigation locks while you annotate. Undo removes the last mark, Esc exits.
1List the recurring work, not the job titles
Start from tasks, not people: research leads, draft follow-ups, prepare the weekly report, check proposals. Pull them from the Weekly Time Map you built in Module 0.
2Assign each task to human, agent or both
For each task decide: human only (relationships, negotiations, final decisions), agent drafts and human approves (most client-facing work), or agent alone (internal research and data prep). Most small businesses end up with the middle category for 60 to 70 percent of tasks.
3Give every agent a human owner
Every subagent has one named person who owns its instructions, reviews its output quality and decides when it changes. No owner, no agent. This one rule solves most accountability questions.
4Fill the worksheet
Use the Agentic Org Chart Worksheet from the materials. Ask Claude to prefill it from your business context, then correct it yourself.
Using CLAUDE.md, my subagents in .claude/agents/ and my weekly time map, fill in materials/07-agentic-org-chart-worksheet.csv for my business. One row per role. Mark clearly where you are guessing. Then list any recurring task that has no owner.5Share it with the team
Put the finished chart where the team can see it and commit the .claude/agents/ folder with the project so everyone uses the same specialists. When someone asks who writes the follow-ups, the answer is on one page.
Here is a question that makes owners go quiet: if an AI agent sends the wrong number to a client, whose fault is it? If you cannot answer in one sentence, you do not have an AI department. You have a collection of experiments.
The agentic org chart is a simple table. Every subagent has a human owner, a clear job, a list of allowed tools, an approval step and a measure of success. It is the document that turns clever prompts into a department you can manage.
Key points
- Map tasks first, then decide human, agent or both.
- Every agent has one named human owner.
- Most client-facing work is 'agent drafts, human approves'.
- Share the chart and the agents folder so the whole team works the same way.
Avoid
- Designing agents around job titles instead of recurring tasks.
- Leaving an agent with no human owner.
- Moving client-facing tasks to 'agent alone' before months of evidence.
A seven-person accounting practice mapped 23 recurring tasks. Six stayed human only (client meetings, advice calls, sign-offs). Fourteen became agent drafts, human approves, including engagement letters and document chasers. Three became agent alone: data extraction from bank statements, deadline reminders for the team, and the weekly workload report. Each of the four subagents got a named owner, and the managing partner finally had one page showing who did what.