Your Bulk Content Strategy Is Wrong. Here's the Business Case for Something Better
The Economics Have Changed. The Business Case Hasn't Caught Up. Here's How to Fix That.
14-Minute Read
There’s an assumption baked into most L&D budget conversations that nobody stops to question.
It goes something like this. Learning content is a production problem. The answer to a production problem is volume. Volume at the lowest possible unit cost is good management.
That logic produced two decades of click-through content that nobody remembers, a generation of learners who tune out the moment a compliance module opens, and a procurement model where “offshore content development” became shorthand for “we’re serious about L&D.” The executives who signed those contracts weren’t bad at their jobs. They were measuring the wrong things and buying the answer to a question that wasn’t being asked.
We know what the question should have been: what behavior are we trying to change, and what does it take to change it?
Most of the time, nobody in the room can answer that. So they bought modules instead. Because it’s not the easiest question to answer, and the answer is usually complex and uncomfortable to sit with.
TL;DR
The bulk offshore content model was built on assumptions about cost, speed, and quality that are all shifting simultaneously
AI tools don’t make the model better. They amplify whatever the underlying design quality is. Low-quality design produces more low-quality content, faster
The economics of skilled L&D work are changing. An AI-equipped ID can now produce what used to require a team, at a fraction of the timeline
The real measurement problem isn’t that we can’t track learning. It’s that nobody agreed on what to track before the storyboard was written
The authoring tool vendors are charging more for AI that generates content faster, not better, and their models aren’t asking the right questions upfront
The first authoring platform that lets IDs bring their own AI will win. The rest are stalling.
🧠 Let’s Talk About What “Cheap Content” Actually Costs
I’ve worked with overseas teams. Some of the most talented designers I’ve collaborated with are based outside North America. This isn’t an argument about geography, and it’s not about the capabilities of any individual designer or developer.
It’s about what the bulk production model selects for.
When you contract a high-volume offshore content team, you’re not buying design thinking. You’re buying throughput. The economic model rewards moving fast, keeping scope narrow, and minimizing back-and-forth. The designers who thrive in that environment are the ones who execute to spec quickly and don’t push back on ambiguous requirements.
That’s a reasonable thing to want if your goal is module count. I can tell you from experience that module count has never truly been the solution. It’s a structural problem when the goal is actually behavior change.
Here’s what I kept running into in practice. We’d establish quality standards. Clear rubrics. Defined criteria for what made a module work, not just technically complete. Interaction requirements. Scenario depth. Alignment to observable behaviors.
And the feedback cycle would start. Something would come back that clicked through seven slides with no branching, no friction, no scenario requiring the learner to think. We’d flag it. The team would say it met the storyboard. We’d point to the standards. They’d say the standards weren’t communicated clearly, or that adding real interactivity was out of scope, or that the timeline didn’t allow for it.
And the thing I kept noticing was that the conversation was never about whether the learning would work. It was always about who was responsible for it, not what we needed.
That’s the tell. When the quality conversation becomes a blame conversation, you’ve lost. Because nobody in that exchange is asking the real question. Will this help a learner do something differently on the job?
📏 The Measurement Problem Nobody Wants to Own
Before we talk about production models, we need to talk about what we’re trying to measure. Because the bulk content factory model didn’t just produce mediocre learning. It gave everyone permission to avoid the harder conversations.
Here’s how the avoidance works. Leadership approves a training initiative. L&D builds the content by a deadline that they didn’t have input into. Learners complete it. The LMS records the completion. Someone reports the completion rate. Everyone moves on.
Nobody asked, at the start, what observable behavior would tell us this worked. Nobody agreed on what a manager should see three weeks later if the training landed. Nobody built a coaching loop in which managers were accountable for reinforcing those skills with their teams. The module was the finish line, and completion was the medal. None of that moves the business.
That’s not a vendor problem. It’s a leadership problem and an L&D problem in equal measure. We let it happen because the alternative — sitting in a room with a business leader and asking “what does success look like and how will you know” — is uncomfortable. It often reveals that the training request was never about behavior change to begin with. Sometimes it’s about covering liability. Sometimes it’s a reaction to an incident. Sometimes it’s just activity that looks like caring about development, so you can say “we develop people” when you’re recruiting.
A skilled instructional designer close to the organization can have that conversation before the storyboard is written. They can push back on a training request that has no measurable outcome. They can ask how we measure this and how we’ll know it worked. They can say, plainly, if we can’t define what good looks like after this, we’re producing content for the sake of checking a box.
An offshore content team working from a storyboard can’t. They’re downstream of a decision that was already made badly.
And here’s the harder part. Even when L&D does define good outcomes, managers often don’t follow through. The training deploys. Learners complete it. And then nothing happens. No coaching conversation. No observation. No accountability for whether the behavior actually changed. The module did its job. Whether it moved anything is someone else’s problem.
That follow-through is an organizational accountability question, not a training design question. But the ID who was in the room when the initiative was scoped is in a much better position to name it, flag it, and build it into the design as an explicit assumption. This only works if managers are coaching observable behaviors afterward. Put it in the storyboard. Make it someone’s job.
📊 The Numbers Behind “Cost-Effective”
The per-module cost of offshore development looks attractive on a spreadsheet. Until you add the costs that don’t show up there.
Timezone latency. Every question to an SME that needs to route through a content team in a different time zone adds 24 to 48 hours. That’s not just clock time. It’s decision latency. A scope question that could be answered in a 15-minute conversation becomes a two-day thread. At scale, this is where timelines collapse. A project that was supposed to take six weeks takes twelve, and the content still isn’t what you needed.
Rework cycles. The first version comes back wrong. You review it, document the gaps, and send it back. The second version fixes the obvious problems and introduces new ones. We expected some ramp-up when we started — a new process, a new team, and a learning curve on both sides. What we didn’t expect was going through rework at least twice per module consistently, even after standards were documented and examples were shared. And when we’d show a developer exactly how another team had done something, the response was sometimes that it couldn’t be done. The ego became a bottleneck. By the time you’ve cycled through three or four reviews, you’ve spent more internal time managing the work than the per-module rate saved you. The cost arbitrage evaporates.
Organizational context. An offshore team works from a storyboard. A skilled ID embedded in your organization carries context that no storyboard fully captures. They know the SME who always wants to add five more topics. They know that compliance training needs to be built differently for the field team than for headquarters. They know that the leadership team approved this content, but the actual learners think the policy it covers is broken. That context shapes the design. It’s not in the storyboard.
The interaction problem. Truly interactive learning takes longer to build than click-through. It requires instructional judgment at every decision point. Branching scenarios where choices have real consequences, feedback loops that adapt to what the learner does, and simulations that let someone practice a difficult conversation before they have it in real life. All of that requires a designer who’s thinking about how learning happens, not someone optimizing for deliverable count.
What comes back from a bulk content model, again and again, is a series of slides with next buttons. It’s not because the designers can’t do better. It’s because the model doesn’t reward it, and the timeline doesn’t support it.
The learners notice. And they’ll tell you directly if you ask. The feedback we kept hearing wasn’t “the content was wrong” or “the design was bad.” It was “I wanted to actually practice doing the thing, not just click through a screen about it.” They weren’t asking for more content. They were asking for the chance to do something — to make a decision, get it wrong, and try again in a safe environment before they had to do it for real. That’s not a preference. That’s how skill transfer works. And it’s exactly what a click-through module can’t give them.
The managers of those learners notice. The business outcomes don’t move. And eventually, someone in the C-suite reads a report about how employee training isn’t producing results and asks what they’re spending the L&D budget on.
That’s the moment you need to have a different answer ready.
🔧 What Actually Changed in the Last 12 Months
The economics of building learning experiences shifted. Not incrementally. Dramatically.
Twelve months ago, building a real interactive learning simulation required either a developer, a significant budget, or both. The tools that made it possible were expensive, slow to develop, and produced SCORM outputs that worked fine in an LMS but weren’t designed for a modern learner experience.
Here’s where we are in April 2026.
Vibe coding for L&D is real. Jeff Batt has been teaching this for the better part of a year. The core concept is to describe what you want to build in plain language and let the AI build it. No JavaScript. No developer. A peer-reviewed study published in April 2026 confirmed that Gemini 3.1 Pro can generate a fully working HTML/JavaScript learning simulation from a single natural language prompt in under ten minutes. That’s not a vendor demo. That’s published research.
Claude Opus 4.7 builds interactive branching scenarios. Open Claude. Describe a scenario in plain language — a manager handling a performance conversation, a customer service rep navigating a difficult call, a compliance situation with three realistic wrong paths. The AI builds the structure, the branches, the feedback. It’s not a finished product, but it’s a starting point that used to take days to rough out in a storyboard, and now takes 30 to 90 minutes.
Veo 3.1 generates character-consistent training video. Give it reference images of a consistent character or workplace environment, and it generates training video clips with native audio included. The free tier covers ten generations per month with any Google account. A microlearning series with a consistent presenter persona that used to require a production budget now costs nothing for the first ten clips.
Lovable and Bolt.new build interactive web-based tools. Describe an onboarding checklist with progress tracking, a scenario simulator for your specific context, a manager feedback tool. The AI builds a functional web application in plain language, no code required.
xAPI and LRS integration make behavior-level tracking possible. SCORM tells your LMS that someone completed a module and scored 80%. That’s it. xAPI tracks what someone actually did. Which scenario branch they chose, how many attempts it took, where they hesitated, what they got wrong the first time, and right the second. Pipe that into a Learning Record Store, and you have data that connects learning activity to observable behavior in a way completion rates never could.
What this means is that a skilled instructional designer with real design thinking and editorial judgment can now produce what used to require a three-person team in a fraction of the time, at a per-module cost that makes the offshore arbitrage argument significantly thinner.
Here’s the honest caveat. This only works if the ID using these tools has the judgment to use them well. An AI tool handed to someone who produces click-through will produce better-looking click-through faster. The upskilling required is real — 90 days of consistent experimentation, not a one-hour workshop. IDs who’ve been building with Claude Opus 4.7, Gemini 3.1 Pro, and Veo 3.1 for the last six months talk about it the way developers talk about learning to code. The first month you’re fighting the tools. The second month you’re figuring out what they’re good for. The third month you start building things you couldn’t have imagined before.
That investment is worth making. But it’s an investment, not a feature you turn on.
💰 The Authoring Tool Problem Nobody’s Naming
Articulate raised prices. iSpring raised prices. Lectora raised prices. The per-seat cost of traditional authoring tools has been rising, while the tools themselves haven’t materially improved the quality of learning output or the speed of development.
What you’ve been paying more for, in most cases, is a familiar workflow and SCORM compliance. And it’s worth being honest about what SCORM actually gives you, which is not much. Completion status, a score, time spent, and basic bookmarking. That’s it. And depending on which LMS you’re running, you might get even less than the spec allows, because implementation varies significantly across platforms. SCORM tells you they finished. It doesn’t tell you what they learned or whether they’ll use it on Monday.
That’s nothing. But if the authoring tool is primarily being used to produce click-through slide decks with branded headers and completion tracking, you’re paying a premium for a production line that produces content your learners are skipping through as fast as possible.
Our authoring tool vendors know this, so they’ve shifted to selling AI features. AI-generated text. AI-suggested images. AI quiz generation. What they’re not selling is an AI that asks you, before you write a single slide, what business objective this is tied to and what observable behavior would tell you it worked. The AI in these tools is designed to help you produce content faster. Not to make you stop and question whether you should produce it at all.
There’s a more specific problem. The AI models inside these authoring tools aren’t the best models available, and the vendors know that. Teams figured this out quickly. What actually works is using Claude or Gemini to develop the full scenario and storyboard, then using Articulate or Rise as the build environment only, completely ignoring the vendor’s AI. That’s not a workaround. That’s the workflow. The authoring tool becomes a production shell rather than a thinking tool.
And then there’s the voice model situation. IDS is currently navigating a real and frustrating dynamic between ElevenLabs and Articulate over AI voice licensing. The technical capability has been there for a while. The commercial relationship between the platforms hasn’t kept up. L&D teams trying to build narrated content with realistic AI voices are stuck in the middle of a vendor dispute that has nothing to do with learning quality and everything to do with licensing economics.
The prediction worth making is this. The first authoring platform that opens the gates and says, “bring whatever AI you want, we’re the build environment, not the thinking environment,” will take significant market share from everyone else. The teams already doing the work know this. The vendors are a step behind.
🏗️ Building the Case for the Room
Here’s what I’d bring into the room and how I’d sequence it.
Start with accountability. When learning doesn’t move the needle, who’s responsible? In the offshore model, the answer is almost never clear. The team points to the storyboard. The storyboard points to the SME. The SME points to the timeline. The L&D leader is left holding a deliverable nobody owns. A skilled ID who was in the meetings where the business objective was defined can’t point at a handoff when it doesn’t work. They were there. Accountability follows proximity.
From accountability, move to context. You cannot write a storyboard that captures everything a good ID learns by being present in your organization. The compliance team’s private skepticism about a policy they all think is unenforceable. The manager who will quietly torpedo any training that doesn’t acknowledge how short-staffed the team is. The gap between what HR says happens during onboarding and what a new hire experiences on day three. None of that is in the storyboard. All of it shapes whether the learning lands. You cannot offshore context.
Then bring the math. A skilled ID using AI tools today can go from learning need to working prototype in two to three days on a module that would have taken weeks. That’s not because AI did the thinking. AI handled the production while the ID did the thinking. Every step in the offshore model adds time without adding quality. Close proximity and better tools compress it all. The per-module cost advantage of offshore narrows considerably once you account for management overhead, rework cycles, and timeline drag.
Accountability first, context second, economics third. Each one lands harder because the previous one set it up.
The authoring tool budget is a fourth point worth raising separately. If the organization is spending significant money per seat on Articulate 360 and that tooling is primarily producing click-through slide decks, that budget line deserves scrutiny. The executive who’s currently paying for both the tool and the offshore production model should hear that they’re paying twice for an outcome they’re not getting once.
✅ The Honest Thing to Say to Leadership
The hardest part of this conversation isn’t the data. It’s naming what the current model produced.
We spent years building content that checked the compliance box and moved the needle on completion rates. That was the metric. The metric was wrong, or at least incomplete. A lot of organizations bought a lot of modules that taught people to click through training as fast as possible without retaining anything, and we called it learning because the LMS said, “completed.”
Leadership that approved those budgets wasn’t wrong to want efficiency. They were working with the measurement tools they had and the production model they had been sold. But the measurement tools are improving, the production model is changing, and the evidence from other industries is pretty clear. People who take boring click-through training and don’t get a chance to practice don’t change their behavior. We’ve known this for decades. The model persisted anyway because it was cheap and trackable.
What AI tools and skilled designers make possible is a different version of the argument. We can build things that work at a cost competitive with the content factory model if we stop optimizing for module count and start optimizing for behavior change. We can track it better, too. Not just “did they finish” but “what did they do, what did they get wrong, what do their managers need to reinforce.” That data exists now. Most organizations aren’t asking for it because they built their entire L&D infrastructure around completion metrics.
The case is simpler than any of the technology. The old model didn’t work well, the new tools make a better model affordable, and we finally have the measurement infrastructure to prove it.
🚀 Where to Start This Week
If you’re an L&D leader who’s been managing a bulk offshore content relationship and you’re not entirely satisfied with what it’s producing, here’s a concrete first step.
Pull one module from the last six months. Not the best one. A representative one. Sit with a real learner and watch them go through it without coaching them. Notice where they click without reading. Notice where they answer a knowledge check question by process of elimination without knowing the answer. Notice what they remember ten minutes after it closes.
Then ask yourself what that module costs, including your time spent on storyboarding, reviewing, revising, and managing the relationship. Put that number next to the learning outcome it produced. Then ask what observable behavior you agreed on at the start that would tell you it worked. If you didn’t agree on one, that’s the first thing to fix before you change anything else.
If that math doesn’t sit right, you have the beginning of a business case.
The next step is simpler than you think. Open Claude Opus 4.7. Describe the same learning objective in plain language. Ask it to build a branching scenario with three decision points, feedback at each branch, and different outcomes based on choices. Budget 30 to 90 minutes. See what’s possible.
You don’t need a new vendor relationship. You don’t need a new tool stack. You need one experiment that provides a credible comparison point and one honest conversation with a business leader about what you’re both trying to achieve.
That’s where the conversation changes.
This is the argument I wish someone had handed me earlier. Not because the people we worked with weren’t talented. Many of them were. But the model we were using didn’t let them do their best work, and it wasn’t set up to produce what learning requires. The economics have shifted enough that you don’t need to make this case from scratch anymore. You just need to be the person willing to make it.
—Eian
Eian Newland is a Learning Leader and L&D Systems Builder.
Sources
Jeff Batt, “Vibe Coding: The Power & Paradox for L&D” — The Learning Guild, 2026
MDPI Education Sciences, “Leveraging Generative AI Through Vibe Coding: A Case of Simulation-Based Curriculum Redesign” — MDPI, April 2026
Anthropic, Claude Opus 4.7 release — Anthropic, April 15–16, 2026
Google, Veo 3.1 “Ingredients to Video” — Google DeepMind, January 2026
Articulate 360 pricing increases — Articulate, 2025–2026
ElevenLabs/Articulate AI voice licensing dispute — ElevenLabs / Articulate, 2025–2026
Lovable, vibe coding platform — lovable.dev
Bolt.new, web app builder — bolt.new
ADL Initiative, xAPI specification — adlnet.gov