What AI Changes in Instructional Design Work

What changes in instructional design work when AI is used?

AI most clearly changes instructional design’s production layer: brainstorming, transforming supplied material, drafting low-stakes text, and streamlining workflow. Usefulness depends on task definition and review. Current practitioner research does not establish that AI replaces teaching or the full design role.

Instructional design shapes learning resources and experiences; teaching facilitates instruction with learners. Drafting an activity is not delivering it or adapting instruction in context.

In the 2024 mixed-methods study “Exploring Instructional Designers' Utilization and Perspectives on Generative AI Tools,” 70 designers completed a survey and 13 joined interviews. Participants reported brainstorming, low-stakes tasks, streamlining design, and collaboration. These self-reports describe participants, not profession-wide adoption, learning gains, or displacement. The 2026 systematic review “The intersection of artificial intelligence and instructional design practice” analyzed 28 studies from 2020 to 2025. It describes an emerging field and notes challenges such as pedagogical alignment and content reliability.

For early- and mid-career designers, inspect recurring tasks with clear inputs and reviewable outputs, and identify work dependent on context and accountable judgment. Exposure is a reason to inspect work, not a forecast about an individual job. U.S. labor evidence is a bounded proxy, not a local prediction.

Sources: Exploring Instructional Designers’ Utilization and Perspectives on Generative AI Tools: A Mixed Methods Study; The Intersection of Artificial Intelligence and Instructional Design Practice: A Systematic Review

Which design tasks change first—and why?

Tasks with clear digital inputs and checkable outputs are plausible places for AI assistance in instructional design: generating options, outlining, rewriting supplied material, creating examples, and preparing drafts. This is a task-level judgment, not a prediction about how much of a designer’s job can be automated. In a 2024 mixed-methods study, 70 instructional designers completed a survey and 13 took part in interviews. Participants described using generative AI for brainstorming, low-stakes tasks, streamlining design, and collaboration. They also raised concerns about quality, privacy, authorship, ownership, and plagiarism, as reported in “Exploring Instructional Designers’ Utilization and Perspectives on Generative AI Tools: A Mixed Methods Study.” These reports do not establish measured time savings, learning gains, or profession-wide adoption. “Instructional Designers’ Integration of Generative Artificial Intelligence into Their Professional Practice” interviewed six designers in corporate and higher-education settings. Participants mostly integrated GenAI during design or instructional planning and development, and saw those phases as having the greatest impact. Their views reflected perceived efficiency alongside concerns about inaccurate outputs; institutional support also shaped integration. Six cases illuminate particular settings, but cannot show how common use is or whether it reduces labor demand. These studies show experimentation around production and development, not a uniform change in every designer’s work. A task is easier to test when its input can be supplied digitally, the requested transformation is specific, the output can be compared with a source or standard, and mistakes have limited consequences. Imagine a designer permitted to use an approved passage who wants a draft at a different reading level. A tool can offer wording; the designer checks whether meaning and terminology were preserved. Or consider requesting quiz distractors based on a lesson. Candidates may suggest alternatives, but a designer or subject expert must judge whether they test the intended concept or introduce ambiguity. These are illustrations, not reported cases. This separates generating candidates from deciding what belongs in instruction. An outline can organize topics without establishing that they meet a learning need. Rewriting can change sentence complexity without showing that the result suits a particular learner group. Examples can broaden options while omitting what matters locally. Assistance is easiest to inspect when source, purpose, and acceptance criteria are explicit. Without them, a polished output may hide a weak task definition. “Between Scaffolds and Shifts: Novice Instructional Designers’ Experiences with Generative AI” used behavioral event interviews with seven novice designers in a South Korean graduate course. It identified AI as an auxiliary tool, feedback provider, or co-creator, and described interactions changing over time. This qualitative account details novice experience in a course project; it does not measure workplace productivity or represent experienced designers generally. It shows why “AI use” is too broad to explain what changed: assistance depends on the task and designer judgment. The 2026 systematic review “The Intersection of Artificial Intelligence and Instructional Design Practice” synthesized 28 studies selected from a Web of Science search. It describes a field moving from tool-focused uses toward collaborative and pedagogically grounded applications, while finding AI used mainly as automation or design assistance in the reviewed literature. The settings vary, so the review maps patterns rather than a typical workflow or causal effect. It cannot turn reported assistance into proof that learning improved or work was eliminated. AACE Review’s survey of possible uses and concerns can prompt a task inventory, but a menu is not evidence that each use is reliable or valuable in a particular workplace. For a practical inventory, list recurring tasks and ask: Are inputs clear and approved? How consequential would an error be? What will checking and repair cost? Clear inputs, modest stakes, and affordable review make a task a reasonable pilot candidate. Context-dependent work may also benefit, but needs stronger controls and accountable review. Record the route from prompt to usable material, including corrections and approval, before calling it faster. This lets a designer explore change without treating possible uses as established practice or a forecast of job loss.

Sources: Exploring Instructional Designers’ Utilization and Perspectives on Generative AI Tools: A Mixed Methods Study; Instructional Designers’ Integration of Generative Artificial Intelligence into Their Professional Practice; Between Scaffolds and Shifts: Novice Instructional Designers’ Experiences with Generative AI; The Intersection of Artificial Intelligence and Instructional Design Practice: A Systematic Review; Generative AI for Instructional Design: Changes, Chances, Challenges

Where does instructional design end and teaching begin?

Designing instruction and teaching it are related responsibilities, but they are not interchangeable jobs. An instructional designer may investigate a learning need, shape objectives, plan materials and practice, and help an instructor deliver a course. A teacher works with learners during instruction: noticing responses, explaining, asking follow-up questions, managing interaction, and adjusting the next step. Some roles combine these activities; others divide them among designers, teachers, coordinators, and subject experts. AI affecting lesson preparation therefore does not, by itself, show that it can take over teaching or the broader design role. A designer might organize source material, sequence topics, or draft a practice question: artifacts that can be reviewed before learners see them. During class, a teacher interprets a learner’s question or hesitation in context and decides how to respond. A tool that helps prepare a draft may alter preparation time without making that decision about what a learner needs next. The U.S. Department of Labor’s O*NET profile, “Instructional Coordinators,” is useful here as a vocabulary for overlapping work, with an important limit: it is a broad occupational category, not a definition of every instructional designer. O*NET lists “Instructional Designer” among reported titles, alongside curriculum directors, education specialists, and learning-development roles. Its profile includes developing instructional materials and interviewing subject-matter experts, as well as observing teaching staff, recommending changes to teaching skills, conducting teacher training, and advising educators on curriculum. The profile thus places content production beside coordination and support for instruction. It does not mean that every person with a design title performs every listed task, or that all designers teach classes. Generating a lesson outline is design support; offering a teacher a possible explanation is preparation support. Whether either helps depends on the material, learners, and purpose. Neither demonstrates that AI has taken responsibility for a class. Even when used during instruction, someone must interpret responses, decide whether an explanation fits, and take responsibility for the next action. The cited evidence does not establish that a system independently assumes the full role. “Generative Artificial Intelligence (GAI) in Teaching and Learning Processes at the K-12 Level: A Systematic Review” shows why teaching cannot be inferred from design studies alone. Using PRISMA, it selected 197 relevant studies published from 2016 through 2024 across five databases and two journals. It covers teacher training and classroom uses, calls for teacher development and evaluation, and describes gaps in practical guidance. This review maps varied K–12 literature; it does not measure teacher replacement or automatically generalize to adult, university, or workplace settings. The qualitative “Between Scaffolds and Shifts: Novice Instructional Designers’ Experiences with Generative AI” is closer to design practice, but narrow. Participants described organizing material and drafting resources, then revising outputs for learner characteristics and context. Novice designers are not representative of all designers, and their accounts do not measure employment effects. They illustrate that producing a candidate resource and deciding whether it fits learners are distinct steps. AI may change teacher preparation and designers’ materials work, and workflows can cross the boundary. The evidence does not establish autonomous replacement of situated teaching or prove teaching is untouched. Designers can ask which tasks end at a reviewable artifact and which continue into coordination, implementation, learner response, and accountable decisions. Preparation assistance is not evidence that live interaction has been automated. Keep conclusions specific to task and setting: the K–12 review cannot settle every classroom, and O*NET’s broad U.S. category cannot describe every design job.

Sources: Instructional Coordinators — O*NET OnLine; Generative Artificial Intelligence (GAI) in Teaching and Learning Processes at the K-12 Level: A Systematic Review; Between Scaffolds and Shifts: Novice Instructional Designers’ Experiences with Generative AI

Why does a faster draft still need a designer?

A plausible draft is only an input to design. Before it becomes a learning resource, someone must check whether it reflects the approved source, fits the learners and setting, supports the objective, and can be used accessibly and safely. A polished quiz item is visible output; evidence that learners can use it to meet an objective is a different result. Review is part of the task, not optional cleanup.

Consider an example: a designer asks for practice questions from an approved policy document. If an explanation changes a condition in the source, the error can travel: a scenario may teach it, a distractor may treat it as correct, and feedback may reinforce it. Checking whether questions sound clear will not catch this chain. The designer must compare claims with the source, decide which distinctions matter to the objective, and involve the subject expert if a disputed interpretation could affect practice. This is judgment about consequences, not proofreading.

Fit is another question a fluent draft cannot settle. Material may assume vocabulary, equipment, time, prior knowledge or authority that learners lack. An activity can align on paper with an objective and still fail where staff cannot pause work, learners use another interface, or examples do not match their decisions. If the real problem is a confusing procedure, missing tool access or a broken handoff, more instructional text may preserve it rather than resolve it.

Practitioner evidence makes review concerns concrete, with limits. In the 2024 mixed-methods study “Exploring Instructional Designers' Utilization and Perspectives on Generative AI Tools: A Mixed Methods Study,” 70 designers completed an online survey and 13 were interviewed. They described using generative tools for brainstorming, low-stakes tasks, streamlining design and collaboration, and raised quality, security and privacy concerns. These self-reports do not estimate added review time or prove a tool fails. In “Between scaffolds and shifts: novice instructional designers' experiences with generative AI,” seven novice designers discussed course projects. One account describes restructuring output that was not organized for direct use. This illustrates repair, not its frequency among experienced designers.

Review extends beyond accuracy. Check whether language, format and interaction work for learners and meet applicable accessibility requirements. Tests depend on format, audience and governing standards; this is not a claim that every generated item is inaccessible. Data handling is a release gate too: use only inputs and tools permitted by the employer or client, and do not expose confidential information without authorization. The 2026 systematic review “The intersection of artificial intelligence and instructional design practice: a systematic review” notes that limited contextual detail constrains fine-grained comparison. It cannot provide a universal checklist or establish learning gains.

For a productivity claim, count the route to an approved resource: preparing permitted sources, generating, checking accuracy and objective alignment, correcting or escalating uncertain content, revising for learner fit, accessibility review and approval. Compare total effort with the usual workflow under the same quality bar. Record corrections and unresolved issues alongside elapsed time. Without outcome measures, call it a production-time comparison, not a learning gain. A faster first draft matters only if the complete work remains accurate and usable. Review effort varies by context.

Sources: Exploring Instructional Designers’ Utilization and Perspectives on Generative AI Tools: A Mixed Methods Study; Between Scaffolds and Shifts: Novice Instructional Designers’ Experiences with Generative AI; The Intersection of Artificial Intelligence and Instructional Design Practice: A Systematic Review

What does the wider research establish—and leave open?

The available evidence supports task change and active experimentation, but it does not establish broad displacement of instructional designers or teachers. These sources measure reported practice, research patterns, modeled exposure, and projected employment. None can be converted into an individual job-loss estimate. In the 2024 mixed-methods study “Exploring Instructional Designers' Utilization and Perspectives on Generative AI Tools,” 70 instructional designers completed an online survey and 13 joined semi-structured interviews. Participants described using tools for brainstorming, low-stakes tasks, streamlining design, and collaboration, alongside concerns about quality, privacy, authorship, and plagiarism. This describes reported experience; it is not representative prevalence, causal proof of time saved, improved learning, or job change. The abstract does not establish employer-wide adoption.

The 2026 systematic review “The intersection of artificial intelligence and instructional design practice” examined 28 articles published from 2020 through 2025. Most addressed perceptions of AI among instructional designers and teachers; others examined effects or proposed AI systems and models. The review describes an emerging research area and notes that differences among roles and educational contexts, plus limited contextual detail in many included studies, restrict fine-grained comparisons. That is useful for identifying recurring questions, but it cannot show how common a workflow is across employers or whether it reduces headcount. The International Labour Organization's “Generative AI and jobs: A 2025 update” takes a different approach: its refined global model covers nearly 30,000 tasks and reports that one in four workers worldwide are in occupations with some degree of potential exposure. The report says most jobs are more likely to be transformed than made redundant because human input remains necessary. This is a global modeled exposure assessment, not observed adoption, a result for instructional designers specifically, or a personal redundancy probability.

For U.S. labor context, the Bureau of Labor Statistics projects 2% employment growth for instructional coordinators from 2025 to 2035 and about 23,100 annual openings, most expected to replace workers who leave or transfer. It describes duties including developing materials, implementing curricula with teachers and principals, and assessing effectiveness. O*NET places “Instructional Designer” among reported titles for the broad instructional-coordinator category, whose tasks also include interviewing subject-matter experts, developing evaluation measures, and recommending curriculum changes. These are useful boundaries, not a clean count of all instructional designers: the category spans different jobs, and BLS does not attribute its projection to AI. It is U.S.-specific and says nothing decisive about a reader's local employer or market.

There is still a credible risk to examine. If an organization standardizes repeatable content production, it could raise output expectations or reduce some contractor hours. That is a plausible organizational response, not an outcome measured by these studies. Nor do duties involving interaction, implementation, evaluation, budgets, or local decisions guarantee that a role will remain funded. To judge your own position, look for sustained changes in project volume, staffing, role design, hiring requirements, and budgets where you work or seek work. Those local signals should change your recommendation sooner than a broad exposure label; if they deteriorate over time, compare an in-role upgrade, a feasible adjacent move, and a larger change against your salary floor, location, training time, health, and family constraints.

Sources: Generative AI and Jobs: A 2025 Update; Instructional Coordinators — Occupational Outlook Handbook; Exploring Instructional Designers’ Utilization and Perspectives on Generative AI Tools: A Mixed Methods Study; The Intersection of Artificial Intelligence and Instructional Design Practice: A Systematic Review; Instructional Coordinators — O*NET OnLine

How should you compare an in-role upgrade, an adjacent move, and a larger change?

Start with the least disruptive option that addresses evidence from your own work. If a recurring task is permitted, measurable, and costly enough to improve, test an in-role upgrade. Consider an adjacent move when local roles exist and your experience matches their requirements. A larger change deserves stronger evidence because it can bring more retraining, income disruption, and uncertainty. An in-role upgrade keeps your field knowledge and adds a bounded capability: using approved tools for a defined production task, setting review criteria, and documenting how source material became a learner-ready resource. A small portfolio case can show the task, input boundaries, revisions, and final checks. That demonstrates what you can do; it does not establish that a credential or tool will secure a job. Before paying for a course or certificate, identify a skill gap in a target role and compare curriculum, feedback, time, and cost with a guided project or self-study. An adjacent move changes the work while preserving some experience. Learning evaluation, instructor enablement, performance consulting, and curriculum coordination are possibilities to investigate, not guaranteed destinations. Compare current local vacancies against your salary floor, schedule, location, and credentials. O*NET’s U.S. Instructional Coordinators profile includes curriculum development, needs assessment, material development, evaluation, and teacher training; it lists “Instructional Designer” among reported titles. The category is broad, so the titles are not interchangeable. The Bureau of Labor Statistics says some public-school instructional coordinator jobs require a master’s degree, teaching or school administration experience, and possibly state licensure. Those requirements apply to the posts described there, not every adjacent job. Check actual employers and local rules. A larger change should clear a higher bar; compare it with staying on the basis of actual openings, not broad predictions. Consider whether the new field uses skills you already have and what evidence would convince a hiring manager to consider your transition. Look for sustained evidence that your current work is shrinking or changing in ways that conflict with your income needs, location, health, schedule, or family responsibilities. Then test whether a specific alternative has real demand and a feasible entry path. Estimate fees, unpaid learning time, possible income interruption, and prerequisites before committing. If facts are missing, use a small project or research step to reduce uncertainty. The comparison is about fit and cost as well as AI. Upgrade when a real task can be improved and reviewed safely. Explore adjacent work when openings and transferable requirements line up. Plan a larger change when local evidence and personal constraints justify its greater burden. Your next move should follow your goal and circumstances, not a general prescription to collect credentials or switch careers.

Sources: Instructional Coordinators — Occupational Outlook Handbook; Instructional Coordinators — O*NET OnLine

What should a useful pilot measure?

A useful pilot measures whether one defined task reaches an acceptable result with less total effort, while meeting the same learning and standards as the usual process. Compare a low-stakes task with its ordinary baseline; a fluent first draft or shorter generation step alone is not proof of productivity. In “Exploring Instructional Designers' Utilization and Perspectives on Generative AI Tools: A Mixed Methods Study,” 70 instructional designers completed an online survey and 13 joined interviews. Participants reported uses such as brainstorming and low-stakes tasks, alongside quality, privacy, authorship, and plagiarism concerns. These self-reports suggest what to measure; they do not establish causal time savings or improved learning. Choose a task that can be tested without exposing restricted information. Check employer policy before selecting a tool or input. Use approved, non-sensitive material, and record the normal process: input, required output, and approver. For example, a designer might test an outline against approved course material. Do not enter student records, confidential client content, or other protected data unless policy permits it. Before comparing workflows, define “usable.” Specify the learning objective, source fidelity, audience, accessibility checks, and errors that would make the output unusable. Set a stop condition: if the tool introduces unsupported claims, misses a critical objective, or requires more repair than the team can review, stop and return to the standard process. Apply the same criteria to both versions; otherwise speed may reflect a lower quality bar. Log the entire path to approval: preparation, generation, source checking, factual corrections, rewriting, alignment review, accessibility work, expert or teacher review, approval, and follow-up repair. Record elapsed and active review time separately when useful. Note what changed and why, including errors caught before release. This makes the designer’s judgment visible instead of treating it as an invisible cost. “The intersection of artificial intelligence and instructional design practice: a systematic review” analyzed 28 studies published from 2020 through 2025. It found varied contexts and aims, many examining perceptions, and identified pedagogical alignment, reliability, and tool constraints as challenges. This supports measuring fit and quality; it does not prescribe a universal score or prove a workflow improves learning. Finally, distinguish production from learning. If the pilot measures only time, corrections, alignment, and approval, report only those findings. A learning claim requires evidence after use, such as whether learners can perform the intended task in context. Define that outcome and how to observe it before interpreting results; do not infer it from a polished resource alone. Keep, modify, or stop the workflow based on net effort and the agreed quality bar, then document the contextual decisions and verification the designer contributed.

Sources: Exploring Instructional Designers’ Utilization and Perspectives on Generative AI Tools: A Mixed Methods Study; The Intersection of Artificial Intelligence and Instructional Design Practice: A Systematic Review

What is the next proportionate step?

For the next week, record recurring design tasks: what each required, how long it took, what judgment or coordination it involved, and what counted as usable. Choose one low-stakes task with approved inputs and a clear review standard. Define checks for source accuracy, learning-objective fit, accessibility, and local privacy or release rules. Record effort through approval, including correction and rewriting. A quicker first draft matters only if the reviewed result remains sound and total work improves; production time alone cannot show whether learners learned more. Use that comparison alongside local evidence: project volume, workload, staffing, role expectations, and relevant openings. The U.S. Bureau of Labor Statistics’ “Instructional Coordinators” outlook is a broad occupational proxy, not a forecast for every designer or an estimate of AI’s effect. The International Labour Organization’s “Generative AI and Jobs: A 2025 Update” models exposure; exposure does not establish employer adoption or individual displacement. If local work is steady, consider an in-role upgrade; investigate adjacent roles against actual requirements and your salary, time, and family constraints. Reassess if sustained local contraction or a verified duty change appears. The free task checker offers change-pressure signals, not a validated job-loss probability. The paid career roadmap compares scenarios and constraints; neither guarantees employment or income.

Sources: Instructional Coordinators — Occupational Outlook Handbook; Generative AI and Jobs: A 2025 Update

Questions readers ask