How AI Changes Storyboard Collaboration

This is Part 2 of The L&D Workflow. Part 1 turns the storyboard into a working draft. This article covers what happens when the team reviews it.

The first draft is rarely the slowest part of storyboarding. The review loop is.

A subject matter expert comments in a document. A manager replies in Slack. Legal marks up a PDF. The instructional designer converts everything into tasks, updates the course and sends another version. Someone inevitably reviews the old one.

For review-heavy projects, I use roughly 10–20% of hands-on production effort as a planning range for collecting, interpreting and reconciling feedback. It is an observation from projects, not an industry benchmark, and the calendar cost is often larger than the labor percentage suggests.

A Comment Is Not Yet an Instruction

When the storyboard is working HTML or an editable slide deck, feedback can point to the exact course element that needs attention.

A reviewer might select a scenario and write:

This is too advanced for a new hire. Make it simpler, but keep the approved compliance language.

The comment identifies a problem, but it leaves important decisions unresolved. Does “simpler” mean shorter copy, easier vocabulary, a different example or less difficult practice? The storyboard author answers that question before an agent changes the course.

The author decides whether feedback remains a comment or becomes a bounded instruction.

This interaction model is already familiar in general-purpose tools. ChatGPT Canvas supports inline feedback, Gemini Canvas supports selected-region edits and Anthropic’s recent Claude Code usage research describes a workflow where users retain most planning decisions while Claude handles much of the execution.

The Instructional Designer Keeps Control

Reviewer comments are inputs. The storyboard author chooses what happens next.

Comment type	Instructional designer action	Agent action
Duplicate or out of scope	Discard	None
Ambiguous or strategic	Discuss with the reviewer	Wait
Clear and low risk	Promote with a bounded instruction	Propose the narrow change
Several comments with one root cause	Batch and reconcile them	Apply one coherent revision
Objective, assessment or compliance change	Require explicit approval	Show a proposal before editing

That triage protects the course from literal or contradictory edits. A reviewer may say, “This slide is confusing.” The instructional designer decides whether the right response is shorter copy, a new example, a stronger visual hierarchy or removal of the slide.

For the pricing-objection lesson from Part 1, the author might promote the comment as:

Simplify this example for a new hire. Preserve the learning objective and approved compliance language. Update the dependent feedback, but do not change the assessment rubric. Show the proposed change before applying it.

The agent executes the author’s interpretation, not the raw comment.

One Comment, End to End

An AI-assisted review loop has six steps:

1. A reviewer comments on the exact course element.
2. The instructional designer discards, discusses, batches or promotes the comment.
3. For promoted feedback, the designer defines what may change and what must remain fixed.
4. The agent asks for clarification or proposes a bounded revision.
5. The designer approves, rejects or adjusts the proposal.
6. The course is updated and the decision remains linked to the source comment.

Low-risk fixes such as typos or approved terminology may move quickly. Changes to objectives, assessments, regulated language or scenario difficulty still need explicit review.

Keep the Comment, Change and Version Together

The useful context includes the selected course element, learner, objective, source hierarchy, affected components and approval rules. The workflow also preserves the artifact before and after the change.

Traceability should work in both directions:

• From a comment, the team sees the instruction, proposal, approval and resulting version.
• From a feature or content decision, the team sees the comment or approved batch that prompted it.

That makes the finished course explainable. “Why is this here?” no longer depends on someone remembering a Slack thread from three months ago.

What This Looks Like in AmpUp

This is the collaboration model we are building toward in AmpUp Courses. A reviewer comments on a course element. The storyboard author retains control over promotion, scope and approval. When feedback is promoted, the selected content, approved source and author-written instruction travel together into course chat.

The content, source comment and agent instruction stay connected instead of becoming separate tasks in separate tools.

Applied changes remain linked to their source comments. Earlier versions stay available for comparison or restoration. Once the course is approved, it can be packaged as SCORM for the LMS of record.

Where This Still Breaks

Comment-to-agent workflows do not resolve disagreements or make weak feedback useful by themselves.

• Two reviewers may request incompatible changes.
• A concise comment may hide a larger objective or assessment problem.
• The selected element may have dependencies elsewhere in the course.
• Regulated content may require a named approver outside the training team.
• Automatic fixes can create false confidence when nobody reviews the learner experience.

The answer is a visible decision process with human review gates, an approach consistent with augmentation before automation.

The Takeaway

Storyboard collaboration used to produce comments scattered across tools. The stronger workflow produces controlled, traceable course changes.

The instructional designer remains the author. They decide which comments enter the workflow, translate the feedback into precise instructions and remain responsible for the result.

Frequently Asked Questions

Should every comment be auto-fixed?

No. Clear, low-risk edits are good candidates. Ambiguous, strategic, factual or regulated changes should require discussion or approval.

Can several comments be handled together?

Yes. The instructional designer can batch related feedback and turn it into one coherent instruction instead of applying contradictory edits independently.

How is this different from adding a chatbot beside a document?

The agent receives the selected course element, approved sources, design rules, dependencies and decision history. It works within boundaries defined by the instructional designer.