The Invisible Architecture: Why Instructional Design Is the Most Important Skill Nobody Is Teaching Properly

Picture a course. Forty-five minutes of recorded lecture. A PDF of slides. A five-question quiz at the end. Published on an LMS and called online learning.

This scene plays out thousands of times a day across schools, universities, and corporate training departments. The content exists. The platform exists. The learner has technically been exposed to the material. And in most cases, very little actual learning has happened.

This is the most expensive mistake in digital education right now. Not expensive in money, but in something harder to recover: learner trust.

Every time someone sits through a poorly designed digital course and comes out the other side having retained almost nothing, they learn a lesson the designer never intended. They learn that online learning does not work. And that belief, once formed, is remarkably difficult to undo.

Instructional design exists to prevent exactly this. And right now, as digital learning moves from fringe practice to mainstream expectation in classrooms, boardrooms, and universities across the world, it has never mattered more. The problem is that almost nobody is being taught to do it properly. Read on for what that actually looks like in practice.

The Convergence Nobody Planned For

In a recent piece for this magazine, Danielle Thomas made an observation that deserves to be taken seriously: teachers who design competency-based learning environments are, in practice, doing instructional design work. They are mapping outcomes, sequencing content, building feedback loops, and creating differentiated pathways. The craft, she argued, is the same, only the job title differs.

She is right. And the convergence is accelerating.

The shift to digital and hybrid learning has collapsed the distance between teaching and design. A teacher building a Google Classroom unit, a lecturer recording an asynchronous module, a school leader commissioning an e-learning course for staff. All of them are making instructional design decisions, whether they know it or not.

Decisions about sequence. About what gets assessed and when. About how a learner moves from confusion to competence. About what happens when they do not.

The difference is that trained instructional designers make those decisions deliberately, with frameworks and evidence behind them. Everyone else makes them by instinct, habit, or accident. And in a field where the quality of those decisions directly determines whether learning happens or not, that gap is consequential.

Design without design thinking is just content arrangement. And content arrangement is not a course.

The Principle Underneath: Learning Is Not Content

The most fundamental mistake in digital course development is treating learning as a content delivery problem. If the learner has access to the information, the thinking goes, learning should follow. It does not. Not reliably. Not at scale. Not without design.

Learning is a cognitive journey through content, and that journey needs to be architected.

The learner needs to be oriented before they can engage, challenged at the right level of difficulty before they can grow, and given feedback before they can self-correct. None of that happens by accident.

This is where the major ID frameworks come in, and where practitioners often get tangled in framework tribalism that is more confusing than helpful.

ADDIE, SAM, and Bloom’s Taxonomy are not competing religions. They are complementary lenses on the same design problem.

ADDIE — Analyse, Design, Develop, Implement, Evaluate — thinks in phases. It gives a structured workflow for moving from a learning need to a finished course, and it is especially useful for large, complex projects where multiple people and stakeholders are involved. It is sometimes criticised for being linear and slow, but that criticism conflates the framework with poor implementation. ADDIE does not prevent iteration; it just does not mandate it.

SAM — the Successive Approximation Model — thinks in iterations. It is built for speed and collaboration, producing rapid prototypes that are tested and refined before full development begins. Where ADDIE plans comprehensively upfront, SAM learns by building. For digital learning in fast-moving contexts. A new tool rollout is an urgent compliance requirement. SAM’s iterative logic is often the more honest response to uncertainty.

Bloom’s Taxonomy thinks in cognitive depth. It gives designers a language for the level of thinking a learning outcome demands. From basic recall at the bottom to synthesis and evaluation at the top. It is the framework that connects learning objectives to assessment design, because the verb in an objective should determine the kind of evidence that demonstrates mastery. ‘Identify the causes’ demands a different assessment than ‘evaluate the consequences.’

A skilled instructional designer moves fluidly between all three depending on context. The framework is always in service of the learning, never the other way around.

Course Architecture: What Good Design Actually Looks Like

This is where the craft lives. Not in the frameworks, which are maps, but in the decisions that get made while building. The ones the learner never sees but always feels.

Danielle’s three questions from the CBE framework translate directly into design language here.

What should learners know and be able to do? How will mastery be demonstrated? What happens when a learner is not yet competent?

These are not just pedagogical questions; they are the structural specifications of a well-designed course. Every design decision that follows should be traceable back to an honest answer to one of them.

Outcome Mapping

Before a single piece of content is created, the designer should be able to articulate every terminal learning outcome for the course: the specific, observable, assessable things a learner will be able to do at the end. Not ‘understand the importance of data privacy,’ but ‘identify three categories of personally identifiable data and apply the organisation’s handling protocol for each.’

The discipline here is ruthlessness. Every piece of content should exist because it serves a learning outcome. If a designer cannot draw a line from a content element to an outcome, that content should not be in the course.

This is the first and most important act of design, and the one most frequently skipped in favour of starting with what the subject matter expert wants to say.

Content Chunking and Sequencing

Cognitive load theory tells us that working memory is limited and easily overwhelmed. A learner encountering too much new information at once does not learn more; they retain less.

Good course architecture chunks content into units that the working memory can hold, and sequences those units so each one builds on what came before.

In practice, this means thinking carefully about the instructional arc within each module: orient the learner to what they will learn and why it matters, introduce one concept at a time with concrete examples, give the learner a chance to interact with the concept before moving on, then connect the new knowledge explicitly to what came before and what comes next.

This arc should be so consistent that a learner could feel the rhythm of a well-designed course even if they could not articulate why it felt coherent.

Interaction Design

Interaction in digital learning is not a feature; it is the mechanism through which passive exposure becomes active processing. A learner who reads about a concept and then immediately applies it in a scenario retains significantly more than a learner who reads about it and moves on.

The design question is always: what kind of interaction serves this learning outcome?

Branching scenarios work for decision-making skills. Drag-and-drop classification works for conceptual categorisation. Reflective prompts work for attitudinal or professional practice goals.

The mistake is treating interaction as decoration, adding a quiz because the content needs ‘something interactive’ rather than because the interaction directly develops the target competency.

Assessment Alignment and Feedback Loops

Assessment in a well-designed course is not a gate at the end. It is woven throughout, formative checkpoints that give the learner information about where they are relative to where they need to be, and give the designer information about where the course is and is not working.

The alignment principle is non-negotiable: assessment must measure exactly what the learning outcome specifies. If the outcome asks learners to apply a process, the assessment must ask them to apply it, not to describe it, not to recognise the correct answer in a multiple-choice list.

Misaligned assessment is one of the most common design failures, and it produces the worst kind of false confidence: learners who score well and retain nothing.

Feedback, meanwhile, should be corrective, not just evaluative. Telling a learner they got something wrong is the beginning of a design response, not the end of one. The feedback should tell them why it was wrong, what the correct thinking looks like, and ideally route them to a different explanation of the concept, not a repetition of the original.

This is exactly what Danielle’s CBE framework requires when it asks: what happens when a learner is not yet competent? The answer is a designed corrective pathway, not an instruction to try again.

Pause here: take one assessment you have built or currently use. Can you draw a direct line from each question to a specific learning outcome? If not, the assessment is measuring something other than what was designed to be learned.

Tools in Their Proper Place

The tools conversation in instructional design has a tendency to run ahead of the design conversation, which is exactly where it should not be. A tool is a production environment. It enables design decisions that have already been made. It does not make those decisions, and it cannot compensate for the absence of them.

With that framing in place, here is an honest look at the current landscape.

Articulate Storyline remains the industry standard for highly interactive, scenario-based e-learning where custom branching logic, precise animation control, and complex interaction design are required. It has a steep learning curve and rewards designers who understand the principles well enough to know which features to ignore. Storyline is powerful precisely because it does not make decisions for you. Which means it produces exactly what the designer knows how to build, no more and no less.

Articulate Rise works in the opposite direction. It is opinionated, fast, and produces consistently clean output through a library of pre-built blocks. For teams building at volume, or for subject matter experts who need to produce learning content without a dedicated designer, Rise lowers the barrier significantly. The trade-off is that it constrains creative and structural decisions in ways that Storyline does not. That constraint is a feature for some projects and a limitation for others.

AI-powered authoring tools. Synthesia for AI avatar video, Coursebox for AI-assisted course generation, and tools like Curipod for interactive lesson generation are arriving fast and improving quickly. They are genuinely useful for reducing production time on content that is well-designed at the structural level. They are not useful as a substitute for that structural design. An AI-generated course outline built on vague outcomes produces a polished, well-packaged course that does not reliably develop the competencies it claims to address.

Platforms like FayEDU’s Course Studio represent an interesting development in this space. Purpose-built environments that attempt to embed ID principles into the authoring experience itself, so that the tool’s architecture guides the designer toward sound structure rather than simply providing a canvas. This is a meaningful design philosophy, and it reflects a growing understanding that good tools should support good design decisions, not just enable fast production.

The principle that governs all tool selection decisions should be simple: the tool should disappear into the learning. If a learner notices the authoring platform, the design has not yet done its job.

A Signal Toward What Is Coming: Personalised Learning

Competency-based education, as Danielle Thomas argues in her piece, promises that every learner can reach the same standard given the right support and the right amount of time. Personalised learning is what that promise looks like when it meets technology.

Adaptive learning systems: platforms that use learner performance data to adjust content, pacing, and pathways in real time are the technological expression of CBE’s philosophy.

A learner who demonstrates mastery of a prerequisite concept moves forward faster. A learner who struggles is routed to additional explanation, a different modality, or a corrective scenario before progressing. The system responds to the learner rather than processing all learners identically.

This is genuinely powerful. It is also genuinely dependent on the quality of the instructional design underneath it. A personalised pathway through a poorly designed course is not a better learning experience; it is a faster route to confusion. The algorithm can optimise the journey, but only if the destination is clearly defined, the waypoints are well-sequenced, and the assessment is measuring what it claims to measure.

This is the frontier that the next piece in this series will explore in depth: what personalised learning actually demands from instructional designers, which tools are closest to delivering on the promise, and what the evidence says about where adaptive systems work and where they fall short. The architecture has to come first. That is what this piece has been about.

The Profession Arriving at What It Was Always Meant to Be

There is an irony at the centre of the current moment in instructional design. As AI tools take over more of the production layer: generating scripts, producing voiceovers, building slide structures, assembling basic assessments, the human value in the profession is moving upstream. Into learning science. Into systems thinking. Into the empathy required to understand how a real person, sitting in a real context with real constraints, moves from not knowing to knowing.

That is not a threat to instructional design as a discipline. It is the discipline finally being valued for what it was always supposed to be. The craft was never in the production. It was in the thinking that made the production matter.

The invisible architecture is the decisions nobody sees: the outcome mapped before the content was written, the interaction designed before the tool was opened, the feedback loop built before the assessment was scored.

When those decisions are made well, the learner moves through the course and feels nothing except that the learning made sense. That is the standard. And it is worth building toward.

Danielle Thomas explored the pedagogical foundations of this work in her piece on competency-based education. Read it here if you have not already. The next piece in this series will go deeper into personalised learning, adaptive systems, and the tools that are closest to making the promise real. Subscribe to Grounding EdTech and you will not miss it.

References & Further Reading

ERO Mastery Learning Evidence Toolkit — evidence.ero.govt.nz

Bloom’s Taxonomy Overview (Vanderbilt CFT) — cft.vanderbilt.edu

ADDIE Model Overview (ATD) — td.org

SAM Model (Allen Interactions) — alleninteractions.com

Cognitive Load Theory — Sweller (1988), Educational Psychology Review — springer.com

CBE: Theory and Practice (ERIC) — files.eric.ed.gov

Mastery Over Time: The Promise and Reality of Competency-Based Education by Danielle Thomas, Grounding EdTech

John Gitonga writes on instructional design, course architecture, and learning experience design for Grounding EdTech Magazine. groundingedtech.fayedu.com