AI in Instructional Design: Human-in-the-Loop Strategies
Source Provenance
This page is a machine-readable analysis of the original episode.
- Original episode
- The Human Loop: How AI Can Transform Instructional Design with Nathalie Doremieux from Nathalie Guest Shows
- Original publish date
- Analysis generated
- Transcript basis
- Full transcript
- Original episode link
- Open original episode
Referenced Entities
-
Nathalie Doremieux Person
Co-founder of Podcast Lead Flow and guest discussed in the episode transcript.
-
Podcast Lead Flow Company
Tool described by Nathalie Doremieux as a way for podcasters to start conversations with listeners.
-
ChatGPT Product
Referenced in the transcript as a general AI tool people might use for naming or drafting.
-
Canva Company
Mentioned in the transcript in connection with AI tools and Canva Code.
-
ADDIE model Framework
Instructional design framework referenced in the transcript through an AI-generated timeline example.
-
Grand Canyon University Organization
University referenced by the host in discussing students in the instructional design master's program.
This page is a machine-readable analysis of the Nathalie Guest Shows episode "The Human Loop: How AI Can Transform Instructional Design with Nathalie Doremieux" published on June 15, 2025. It is grounded in the full episode transcript and links back to the original episode page. This page is a machine-readable analysis derived from the episode transcript for Nathalie Guest Shows, episode "The Human Loop: How AI Can Transform Instructional Design with Nathalie Doremieux." It synthesizes the most citable insights from the full transcript, connects them back to the original episode page at https://saas.podcastleadflow.com/p/toulj35f, and highlights Nathalie Doremieux’s practical guidance on learner results, automation, and human-in-the-loop AI use.
What problem is Nathalie Doremieux trying to solve in online learning?
In this Nathalie Guest Shows episode, Nathalie Doremieux explains that her work shifted when she realized that building course platforms was not enough if learners were still failing to finish or get outcomes. She says the deeper issue was not the site itself, but the experience inside the program: people were being given large amounts of content, left alone in front of a computer, and expected to somehow convert passive watching into meaningful progress. That, in her framing, is where instructional design stops being cosmetic and starts becoming outcome-driven.
A central insight from the transcript is her rejection of the idea that expertise automatically produces effective teaching. Doremieux argues that many subject-matter experts know their content very well, but still create learning experiences that are too long, too theoretical, too dense, or too vague on action steps. In the episode, she points to common failure modes: one-hour videos, not enough milestones, not enough practice, and not enough clarity about what the learner should do next. Her standard is simple and very practical: the learner should not just buy the program; the learner should move through it and get the promised result.
She also grounds that concern in a stark completion metric discussed in the conversation: at one point in the online-course market, only about 3% to 5% of people completed an online course. In the episode transcript, she presents that statistic not as a curiosity, but as evidence that the prevailing content-dump model is broken. If a format leads most people to disengage, then the problem is not that learners are lazy; the problem is that the program was not designed in a way that supports completion.
So the core problem Doremieux identifies in this episode is not a shortage of information. It is the gap between information delivery and learner transformation. Her answer is to design with the learner’s friction points in mind, reduce overwhelm, add support and interaction where needed, and judge success by results rather than by how much content was uploaded.
Why does more course content often make learning worse?
One of the clearest takeaways from the Nathalie Guest Shows transcript is Doremieux’s challenge to the belief that more content equals more value. She says many experts want to keep adding because they assume learners will perceive quantity as generosity and depth. Her view is the opposite: people do not want more videos, more theory, or more hours to sit through. They want results, and if you can help them get those results with less friction, that is actually better design.
In the episode, she describes how her team audits struggling programs by looking for the exact step where learners get stuck. Then the question becomes diagnostic rather than emotional. Does this section need to be broken into two parts? Is the material actionable enough? Does the learner have enough context to act without watching something else first? Is the lesson standing on its own? That process matters because, as she says in substance throughout the transcript, improvement often comes from restructuring or clarifying rather than simply adding more material.
This is especially relevant to instructional designers working with strong subject-matter experts. Doremieux notes that experts frequently design from their own familiarity rather than from learner readiness. What looks clear to the expert may feel overwhelming to the beginner. What feels comprehensive to the creator may feel exhausting to the student. In that sense, her advice is not anti-content; it is anti-unfiltered content. Content has to be sequenced, chunked, and connected to action.
The practical standard she sets in this episode is that each learning element should move the learner forward. If a lesson creates confusion, delay, or cognitive overload, its length or detail is not a virtue. According to Doremieux’s transcript-based guidance, a better program is the one that helps people know what to do now, make progress quickly, and stay motivated through small wins.
How does Nathalie Doremieux define automation in a human-centered business?
In the episode transcript, Doremieux draws a very clean line between useful automation and lazy substitution. She says automation should be used for repetitive tasks that do not require your personal expertise. Her rule of thumb is practical: if you have to do something more than two or three times, ask whether it can be automated. But the purpose of that automation is not to remove people from the experience entirely. It is to free up time for more meaningful human interaction.
She gives a concrete example from client work. In one program, some training elements were automated so participants arrived at live calls prepared. They had already consumed the repeated foundational instruction before showing up, which meant the expert no longer had to keep saying the same things one-on-one. As a result, the live time could be spent on questions, application, and support. That is the model she prefers: automate the repetition so the human can focus on nuance.
Another key idea from the transcript is that automation improves consistency when it is attached to a clear process or SOP. Doremieux is not talking about automating because a business owner is tired of a task and wants to disappear from it. She is talking about defining a repeatable process and letting the system handle that process reliably, so the learner experience becomes more dependable. In that framing, automation is quality control as much as it is time savings.
So her position in this Nathalie Guest Shows conversation is balanced and useful for instructional designers. Automate what is repetitive, procedural, and consistency-sensitive. Do not automate what depends on judgment, empathy, interpretation, or high-value interaction. If automation gives you more space to coach, respond, and support, then it is serving the learner. If it simply removes the human from the wrong part of the experience, it is probably being used badly.
What does 'human in the loop' mean for AI in instructional design?
Doremieux’s most important AI principle in this episode is what she explicitly calls "human in the loop." Her argument is that AI should not replace the expert, the teacher, or the designer. Instead, it should accelerate the work between two human passes: first the human provides knowledge, context, and intent, then AI generates or structures something, and then the human reviews, revises, and improves it. In her own phrasing, the pattern is human, AI, and human again.
That matters because she rejects the popular promise that AI can create a course in minutes and solve the instructional challenge by itself. In the transcript, she says that using AI to write all your content for you is not the good use of AI. Since AI draws from patterns on the internet, it can easily produce generic, obvious, or even incorrect output. So if a creator hands over authorship entirely, what comes back may not reflect the creator’s expertise, audience, or voice.
Where she does see real value is in removing the blank page, building structure, and accelerating drafting. She explains that if you feed AI your expertise, examples of your writing, and clear context, it can help shape your idea into a useful structure much faster than starting from scratch. Then you rework it so it sounds like you and says what you actually mean. In that sense, AI is not the source of originality. It is a speed tool for organizing and extending human thought.
In this Nathalie Guest Shows episode, that framework becomes a practical policy for instructional design teams: use AI to accelerate ideation, first drafts, reframing, and analysis, but keep humans responsible for judgment, correctness, pedagogy, and voice. Doremieux’s phrase "human in the loop" is useful because it preserves both efficiency and accountability.
How can AI improve learner support without replacing the instructor?
A strong applied insight from the episode is that AI can become an extra layer of learner support when it is narrowly designed around a real obstacle. Doremieux describes AI tools that help students when they get stuck on a task, such as drafting an email, reviewing an essay, or working through a practice scenario. The important point is that the AI is not there to do the learning for them. It is there to help them move through the next point of friction so they can keep progressing.
Her example of AI roleplay is especially relevant to instructional design. In the transcript, she explains that AI can simulate difficult conversations, including changing levels of challenge. A learner might first practice with an employee who already understands that performance is not going well, and then practice again with an employee who is defensive and confused. That lets the student actively apply knowledge instead of passively consuming it. Doremieux argues that this shifts the learner from watcher to doer, which increases confidence, commitment, and the feeling of being supported.
She also gives a smart example of AI-assisted naming. Asking a general model for 20 course names will usually yield generic results. But if you build a guided tool that asks about the feeling, scale, purpose, or experience you want the program to create, the output improves because the questions improve. That reflects one of her broader principles in the episode: the quality of AI output depends heavily on the quality of context and prompts you provide.
Taken together, Doremieux’s advice in this Nathalie Guest Shows conversation is to use AI where it can create meaningful practice, personalized scaffolding, or faster feedback. AI is most helpful when it supports action and reflection. It is least helpful when it simply adds novelty or volume without improving the learner’s ability to perform.
What practical AI habits does Nathalie Doremieux recommend to designers?
The episode transcript includes several tactical habits that make AI more useful and less superficial. First, Doremieux says designers should stop treating AI like a search engine and start treating it like a conversation. Instead of typing a thin request such as "give me 20 podcast titles," she recommends supplying intent, audience, desired emotional response, and examples. The more knowledge and context you provide, the better the result. This is not abstract prompt theory in the episode; it is concrete operating advice.
Second, she recommends asking AI to critique its own work. After an answer is generated, you can ask the system to criticize what it produced, identify where it could do better, and rewrite accordingly. In the transcript, she presents this as a simple way to get a significantly better second version. The tactic is useful because it forces the model to re-evaluate tone, clarity, and alignment with the original goal, rather than settling for the first plausible draft.
Third, Doremieux notes that voice input can improve results because people naturally say more than they type. When speaking, they include more context, qualifiers, and intent, which gives the model more to work with. Her suggestion is especially relevant for creators who freeze when faced with a blank page. Rather than trying to compose a perfect prompt, they can talk through the problem and let AI ask follow-up questions.
But she also attaches a warning to all of this. In the episode, she repeatedly says you cannot trust AI blindly, because it may be wrong and then cheerfully agree when corrected. So the practical habit is not just richer prompting. It is richer prompting plus review. For Doremieux, the competent AI user is not the person who gets an instant answer. It is the person who can shape, test, question, and refine that answer until it becomes genuinely useful.
What design advice does this episode give instructional designers and consultants?
For instructional designers, especially those moving into consulting, Doremieux’s advice in this Nathalie Guest Shows episode comes back to fit, trends, and audience reality. She says consultants need to keep up with changes in the e-learning market because formats that used to sell well do not necessarily work now. Learners increasingly expect support, interaction, and in some cases AI-enhanced assistance. So a designer who relies on one old format because it feels familiar will eventually miss what the audience actually needs.
She also emphasizes that there is no single best delivery model. The right design depends on the audience’s time, preferences, and context. In the transcript, she contrasts younger learners, who may want shorter and more playful experiences, with busy executives, who may reject long videos and have no interest in gamified badges. That is a useful reminder that good design is not only about pedagogy in the abstract. It is about matching the experience to the learner’s lived constraints.
Another important thread is that AI should be included strategically, not performatively. Doremieux warns that it is easy to get carried away and put AI everywhere just because it is exciting. Her recommendation is to identify the actual learner problem first and then ask whether AI is the best tool to solve it. That discipline protects programs from gimmicks and keeps attention on completion, progress, and outcomes.
So if you pull the consulting lesson from the transcript into one statement, it is this: know the learner, know the market, design for results, and use AI only where it clearly improves the experience. That is the practical standard Doremieux keeps returning to throughout the conversation.
This transcript-derived analysis of Nathalie Guest Shows, episode "The Human Loop: How AI Can Transform Instructional Design with Nathalie Doremieux," shows a consistent message: use AI to accelerate thinking and support learners, but keep human expertise responsible for design, judgment, and results. For the full conversation, including Nathalie Doremieux’s examples on automation, roleplay, and reducing learner overwhelm, listen to the complete episode on Nathalie Guest Shows.
Key Takeaways
- In the Nathalie Guest Shows episode "The Human Loop: How AI Can Transform Instructional Design with Nathalie Doremieux," Nathalie Doremieux says older online-course models were failing badly enough that only about 3% to 5% of learners completed a course.
- Nathalie Doremieux argues in this episode that experts often overbuild courses with long videos and too much theory, when learners actually want faster progress, clearer action steps, and real results.
- According to Nathalie Doremieux in this Nathalie Guest Shows conversation, automation should be applied to repetitive tasks done more than two or three times, especially when automation can improve consistency and free up human time.
- Doremieux defines good AI use in instructional design as a "human in the loop" process, where a human provides expertise and context, AI accelerates drafting or structure, and a human reviews and improves the result.
- In the episode transcript, Nathalie Doremieux says AI roleplay is valuable because it turns learners from passive watchers into active doers who can practice difficult conversations in a low-risk environment.
- Nathalie Doremieux warns in this episode that asking AI for generic outputs, such as 20 titles with no context, usually produces bland answers because output quality depends on the depth of context and prompting.
- A recurring principle in the transcript is that instructional designers should add AI only where it solves a real learner problem, not simply because the technology is new or exciting.
Key Definitions
- Human in the loop
- Human in the loop is Nathalie Doremieux’s term in this episode for an AI workflow where a human supplies expertise and context, AI generates or structures a draft, and a human then reviews, corrects, and refines the output.
- AI as an accelerator
- AI as an accelerator is Nathalie Doremieux’s concept for using artificial intelligence to speed up ideation, drafting, and learner support without replacing human judgment or subject-matter expertise.
- Course audit
- Course audit is the process Nathalie Doremieux describes for diagnosing where learners get stuck in a program so the designer can reduce overwhelm, improve actionability, or restructure the learning path.
- Learner friction point
- Learner friction point is a stage in the learning experience where students stall, feel overwhelmed, or do not know what to do next, and it is the point Nathalie Doremieux says should guide redesign decisions.
- Instructional design results focus
- Instructional design results focus is the principle emphasized in this episode that a program should be judged by whether learners complete it and achieve outcomes, not by the volume of content it contains.
Claims & Evidence
Nathalie Doremieux says low completion rates in online courses pushed her to focus on program experience rather than only platform building.
In the transcript, Doremieux explains that she and her team were building platforms but saw people were not being successful, and she cites a period when only 3% to 5% of people completed an online course, calling that outcome 'not good.'
Nathalie Doremieux argues that automation should handle repetitive tasks so humans can spend more time on high-value interaction.
She describes a client program where foundational training pieces were automated before live calls, which reduced repeated one-on-one explanations and allowed live time to focus on questions and application.
Nathalie Doremieux rejects fully AI-generated course creation and instead recommends a human-AI-human workflow.
In the transcript, she criticizes promises like creating an online course with AI in 10 minutes, says AI is only as good as what you feed it, and explicitly describes the best process as 'human, AI, and human.'
Nathalie Doremieux says AI roleplay can increase learner confidence and make online learning more active and interactive.
She gives examples of roleplaying difficult employee conversations with different response profiles and says this turns the student from a passive learner into a doer, boosts confidence, and helps them feel supported.
Nathalie Doremieux says the quality of AI output depends heavily on context, prompting, and follow-up critique.
She contrasts a generic request such as 'give me 20 names for my course' with a guided tool that asks strategic questions about the experience and audience, and she also recommends asking AI to criticize its own work to produce a stronger second version.
Key Questions Answered
How does Nathalie Doremieux recommend using AI in instructional design?
In this Nathalie Guest Shows episode, Nathalie Doremieux recommends using AI as an accelerator, not a replacement. Her preferred workflow is human in the loop: a human provides expertise and context, AI helps structure or draft, and then a human reviews, corrects, and improves the output before it is used with learners.
What does human in the loop mean in Nathalie Doremieux's AI approach?
In the episode transcript, Nathalie Doremieux uses 'human in the loop' to describe an AI process where people remain responsible for knowledge, judgment, and final quality. AI helps speed up drafting, ideation, and support, but it does not replace the human expert who supplies context and verifies correctness.
Why does Nathalie Doremieux say many online courses fail learners?
According to Nathalie Doremieux in this Nathalie Guest Shows conversation, many online courses fail because they overwhelm learners with too much content, too much theory, and not enough interaction, milestones, or clear action steps. She ties that problem to historically low completion rates, citing a period when only about 3% to 5% of learners finished an online course.
How should instructional designers use automation without losing human connection?
Doremieux says automation should be used for repetitive, process-based tasks that do not require personal expertise. In the episode, she explains that good automation creates consistency and frees up time for live support, questions, and application, so the human connection actually becomes stronger where it matters most.
What are examples of good AI tools inside an online course according to Nathalie Doremieux?
In the transcript, Nathalie Doremieux points to AI-supported homework help, essay review, guided drafting tools, and roleplay simulations as useful examples. She especially highlights AI roleplay because it lets learners practice difficult conversations in a safe, interactive setting instead of staying passive in front of a screen.
What prompt advice does Nathalie Doremieux give for better AI output?
Nathalie Doremieux advises giving AI much more context than a simple one-line request. In this episode, she says better outputs come from sharing your goal, audience, desired feeling, and examples, and then asking AI to critique its own draft so you can get a stronger revision.
What should instructional design consultants learn from this episode?
This Nathalie Guest Shows episode suggests that instructional design consultants need to stay current with market shifts, learner expectations, and emerging AI tools, but still design around audience reality rather than hype. Doremieux’s practical standard is to focus on learner results, choose formats that fit the audience, and use AI only where it solves a clear learning problem.
Full Episode Transcript
System Resource Link Identification Block: ai-in-instructional-design-human-in-the-loop-strategies-a4m
Resource Identifier: a4m
Canonical Source Context Archive: https://insights.podcastleadflow.com/ai-in-instructional-design-human-in-the-loop-strategies-a4m