Welcome to the TTA Community. TTA Connect is where you can manage and update your profile, search, and view opportunities, manage your work, track payments, and more.
AI in instructional design can accelerate production, but polished output may still be inaccurate, incomplete, or based on a poorly defined performance problem. This blog explains which tasks AI can support, where human judgment remains essential, and the safeguards teams need for accuracy, data protection, review, and accountability. It also shows L&D teams how to measure whether AI is reducing cycle time and rework without lowering learning quality.
As AI becomes part of everyday development, instructional designers face a practical question: Which tasks can it handle safely, and which decisions still need human judgment? AI can speed up storyboards, assessment drafts, narration, and other production work. Designers still need to lead the analysis and review the final content. That balance saves time without lowering the standard.
A field experiment at Boston Consulting Group shows the risk. Published in Organization Science in 2026, the study randomly assigned 758 consultants to three groups: no AI, GPT-4, or GPT-4 with training. On tasks the model handled well, the AI groups worked about 25 percent faster and produced better results. On a task beyond the model’s capabilities, however, they answered correctly only 60 to 71 percent of the time, compared with 84.5 percent for the group working without AI. Reviewers also found the incorrect AI-assisted answers more coherent and persuasive than the correct human answers.
The lesson for learning teams is simple: AI can save real time on tasks it handles well, but it can also produce polished material that’s wrong. Those errors are harder to catch during a busy review.
Use AI for production work after someone has made the key decisions. Once you know which behavior needs to change and who needs to change it, AI can help by:
Work that once took a week may now take an afternoon. That speed only helps if teams use the saved time well. Analysis and evaluation are often the first activities cut when deadlines slip, so put that time back into both. Teams that get meaningful results make this choice deliberately. It’s part of the same trust, transparency, and training approach that effective AI adoption requires across the business.
Most problems start with weak analysis. A model doesn’t know which behavior Operations needs to change, whether a subject matter expert’s guidance reflects the latest process, or what will connect with a night-shift audience that has already heard the same message twice.
Ask a model to build a course on de-escalation, and it can produce scenarios, a knowledge check, and a facilitator guide. It can’t tell whether repeat contacts come from a queue-routing rule or from supervisors who stopped coaching after their span of control doubled. Someone still has to investigate the problem and decide whether training is the right response.
That makes the analysis phase of instructional design more valuable, not less. When drafting took weeks, teams had a strong reason to avoid building the wrong solution. Cheap drafting weakens that check. A poor analysis can now produce polished material in a week, while the real cost appears months later when the program changes nothing.
Weak AI output gets approved because it looks finished. Clean structure, confident language, and consistent formatting can make a draft seem well reasoned even when it isn’t. Reviewers in the Organization Science study showed the same bias: They rated incorrect AI-assisted answers as more coherent and persuasive.
People also have trouble judging whether AI helps them. In a randomized controlled trial, the independent AI evaluation nonprofit METR studied 16 experienced open-source developers working on 246 real issues in repositories they knew well. With AI assistance, the developers took 19 percent longer to finish the work. Afterward, they still thought AI had made them 20 percent faster.
Instructional designers probably won’t judge AI’s impact any more accurately. Feeling faster isn’t evidence. Without clear measures, teams may expand AI use because it feels convenient, not because it improves quality, shortens cycle times, or reduces rework.
Keep the standard short enough to use without looking anything up. Five safeguards cover most risks: approved tools, a named reviewer, clear data rules, provenance, and factual review by a subject matter expert.
Add one more low-cost safeguard: Have a named subject matter expert review every AI-generated scenario and assessment item for accuracy before development begins. Models can invent plausible details, and a convincing error in compliance training may later be treated as fact.
Choose two measures and track them for a quarter. First, measure rework. Count the factual corrections that subject matter experts request for each module, then compare the quarter before AI adoption with the quarter after. Next, track the full cycle time from request to launch. A first draft may arrive faster even when the overall project takes longer. Both measures are inexpensive and show what AI has actually delivered.
This only works when designers can be honest about using AI. If they can label a draft as AI-assisted without being penalized, errors are more likely to surface during review instead of after launch. That openness is central to an AI-friendly culture.
Responsible AI use in instructional design requires clear roles. Let the model help with drafting. Keep the designer responsible for deciding what’s worth creating, and require a named reviewer to approve the content before learners see it. This gives teams a practical way to work faster without lowering the standard.
Yes, especially for production work that follows a clear decision. AI can speed up storyboard drafts, assessment items, narration, translation preparation, and feedback summaries. The designer should still lead the analysis that determines what to build.
Review anything factual that a learner may act on. Before development, a named subject matter expert should confirm the accuracy of compliance content, safety procedures, product specifications, regulatory references, and scenarios presented as real situations.
Review the content against the original analysis, not just the draft. Confirm that it targets the behavior the business identified, that the examples reflect current work, and that a subject matter expert has verified every specific detail. Treat even a polished draft as unreviewed until those checks are complete.
AI changes how instructional designers spend their time. It can handle some drafting and formatting, which makes analysis, measurement, accessibility, and content governance even more important. In many teams, the bottleneck moves rather than disappears. The skills that determine whether a program works—such as challenging a stakeholder’s view of the problem or explaining that training won’t solve it—still require professional judgment.
Setting a standard is easy. Following it across a backlog takes enough capacity to protect analysis, review, and evaluation when deadlines tighten. TTA connects organizations with vetted instructional designers, learning experience designers, and evaluation specialists who can work within your tools and review standards. Its network includes more than 5,000 instructional design professionals across 30 industries.
Find instructional design talent on TTA Connect
