Human-Agent Workflows
REALM Examples
Concrete human-agent workflows showing how REALM turns loose AI tools into a company operating system: direction from the Player, clear Character roles, explicit autonomy, and every result written back to the Codex.
Agents change the bottleneck
The hard part is no longer only doing the work. It is deciding who owns context, which actions need approval, and where the company remembers what happened.
Traditional frameworks assume human teams
Scrum, Kanban, and org charts were built around meetings, human memory, and calendar coordination. Agents work through written context and explicit boundaries.
REALM makes the invisible visible
Every Character has a role, every Quest has a result, every handoff leaves a trace, and the Player keeps the decisions that should stay human.
Example 01
Supplier Onboarding Quest
A company of one evaluates a new supplier, checks margins, contacts the partner, prepares listings, and records the relationship as a Guild.
Scenario
The Player finds an interesting artisan brand at a market event. The opportunity is real, but onboarding it touches research, finance, outreach, product setup, content, and long-term memory.
Outcome
The supplier is either rejected with evidence or onboarded with a Guild file, margin notes, approved outreach, product tasks, and launch content ready for review.
Why it matters
This example shows how REALM keeps the human on strategic decisions while agents handle the operational chain around the decision.
Creates the Quest in the Codex, defines the acceptance criteria, and marks outreach and spending as Tier 3 approval points.
- Codex write
- SAGAS/Active/Supplier-Onboarding-Quest.md
- Boundary
- Human decision
Researches the brand, product range, pricing, reputation, availability, and visible demand signals without making a recommendation.
- Codex write
- INTELLIGENCE/Suppliers/Brand-Research.md
- Boundary
- Tier 1
Turns the Hunter report into margin estimates, catalog fit, risk notes, and a simple recommendation brief for the Player.
- Codex write
- FINANCE/Supplier-Margin-Assessment.md
- Boundary
- Tier 2
Reviews the recommendation and decides whether the Realm should pursue the supplier. If yes, approves the outreach boundary.
- Codex write
- QUESTS/Decisions/Supplier-Go-No-Go.md
- Boundary
- Human approval
Prepares the outreach checklist, tracks reminders, records the negotiation state, and creates the Guild file once the relationship is real.
- Codex write
- GUILDS/New-Supplier.md
- Boundary
- Tier 2, Tier 3 before sending
Drafts product descriptions, launch copy, and the story angle for the new supplier, then leaves it ready for Player review.
- Codex write
- MARKETING/Supplier-Launch-Draft.md
- Boundary
- Tier 2, Tier 3 before publishing
Builds the execution list: product entries, inventory records, pricing templates, image tasks, and marketplace checkboxes.
- Codex write
- QUEST-BOARD.md
- Boundary
- Tier 1
Runs the Save Point, reconciles the board with the Codex, captures open risks, and marks the Quest complete or carried forward.
- Codex write
- SESSIONS/Save-Points/Supplier-Onboarding.md
- Boundary
- Tier 1
What the Codex remembers
- Quest file with goal, scope, acceptance criteria, and Player approval points
- Hunter intelligence report with cited sources and uncertainty labels
- Mage recommendation brief with margin and risk assumptions
- Guild file documenting the supplier relationship
- Quest Board tasks for product setup, content, and follow-up
- Save Point note recording what changed in the Realm
Example 02
Weekly Content Engine
A solo operator turns customer questions and market signals into a weekly article, newsletter, or social post without losing source traceability.
Scenario
The company needs consistent content, but the Player cannot spend every week starting from a blank page. The goal is one useful piece of content that supports the current Saga.
Outcome
One approved content asset is drafted, sourced, checked against the Saga, prepared for publication, and saved back into the Codex for future reuse.
Why it matters
This example shows how REALM separates discovery, judgment, voice, execution, and memory so content production does not become a mess of chats and forgotten prompts.
Sets the content direction for the week: audience, purpose, format, and what the piece must help the current Saga achieve.
- Codex write
- SAGAS/Active/Content-Quest.md
- Boundary
- Human decision
Collects customer questions, competitor angles, community discussions, search prompts, and relevant source material.
- Codex write
- INTELLIGENCE/Audience-Questions.md
- Boundary
- Tier 1
Clusters the findings, identifies the strongest angle, checks whether the topic supports the Saga, and flags weak assumptions.
- Codex write
- MARKETING/Content-Insight-Brief.md
- Boundary
- Tier 2
Writes the draft in the brand voice, using the Mage brief and cited Hunter findings instead of inventing from memory.
- Codex write
- MARKETING/Drafts/Weekly-Content.md
- Boundary
- Tier 2
Reviews the message, edits the human judgment parts, and approves or rejects public publishing.
- Codex write
- QUESTS/Decisions/Content-Approval.md
- Boundary
- Human approval
Formats the approved asset, prepares platform-specific variants, schedules it, and checks the publishing checklist.
- Codex write
- QUEST-BOARD.md
- Boundary
- Tier 1, Tier 3 at publish
Records what shipped, stores the final asset, adds follow-up reminders, and includes results in the next Scroll.
- Codex write
- SESSIONS/Scrolls/Content-Results.md
- Boundary
- Tier 1
What the Codex remembers
- Content Quest tied to the active Saga
- Audience question and source notes from the Hunter
- Mage angle brief with assumptions and recommendation
- Bard draft with source traceability
- Player approval note before public publishing
- Save Point and performance reminder for the next Session