AI4LDWe IlluminateExplore Bundle B
Bundle B AI Design Agent · Agency

Set the Boundary.
Then Supervise the Work.

Delegate a real feedback-loop design task to this agent at an autonomy level you set. Watch the human checkpoints fire, change the dial mid-task, and walk away with both the finished design and a reusable record of how you configured the delegation.

Companion tool to the ebook Agency as an AI Affordance: A First Look at Agentic Design.

Design Journey (optional)

Save this session so other AI4LD tools can see it, or start it fresh with no prior context. Entirely optional; nothing here is required to use the agent.

Set up your delegation

One task type: draft a feedback-loop structure for a course you're actually building.

What's being taught, to whom, at what stakes?

Short menu, or describe it yourself: this is the construct-fit judgment step 1 will ask you to confirm.

Cadence limits, tooling, reviewer availability.

Before you deploy

Hardman's three decision tests (~30 seconds), framed as the decision the ebook says to make before delegating anything.

Your autonomy dial

Defaults to level 2. H-points fire at every level, without exception.

1. Propose each step
Nothing executes until you confirm it. Highest interaction cost, lowest surprise.
2. Execute and checkpoint
Routine steps run on their own; the agent pauses at your gates and at both H-points.
3. Execute fully, then review
One H-point confirmation, then the whole task runs — you review the full decision log before export.
Calls used this session0 / 20

The published step list

Visible before the first call. Toggle an extra gate on steps 2–5 if you want to review them yourself; steps 1 and 6 are immutable H-points.

1
Restate the design goal and surface assumptionsH-point
2
Identify the feedback moments across the course arc
3
For each moment, choose feedback type, timing, and source — with learning-science rationale
4
Draft the full loop structure (who / what / when / how-corrected)
5
Self-check the draft against your stated constraints; flag uncertainty
6
Present for review at your configured gatesH-point

Nothing you enter is stored on our servers. Your text is processed by Anthropic's Claude to generate this tool's response. See Privacy for details.

Course context and outcome are required to begin.

Your execution stream and decision log will appear here once you begin.