Your Course Library Is Technical Debt
Software teams have a name for the shortcuts that let them ship faster today and make every change harder tomorrow: technical debt. The metaphor is precise. You borrow time now and pay interest later, in slower changes, more errors and more effort spent just keeping things running. Left alone, the interest eventually consumes the whole budget, and the team spends its days servicing the debt instead of building anything new.
Learning libraries accumulate the same kind of debt, and most L&D teams have never put a number on it. Every course shipped without an owner, a review date or a reusable structure is a small loan. One loan is harmless. A thousand of them are a liability.
AI is about to print loans faster than any team has ever borrowed. When a plausible course takes an afternoon to generate, the library grows at a pace no maintenance process was built for. This post is about how to see the debt you already carry, how its interest compounds, and how to pay it down before the next policy change exposes it.
What debt looks like in a learning library
Debt hides until something changes, and then it shows up all at once. It usually looks like this:
• The same fact lives in many places. A refund rule, a safety threshold or a product name is typed into fifteen courses, six job aids and a handful of slide decks, each slightly different.
• Content is locked in its container. Material built slide by slide in an authoring tool can't be reused. Changing one sentence means reopening the project, editing, republishing and retesting.
• Nobody owns it. The designer who built the course left two years ago, and the business owner changed roles. No one is accountable for whether it is still right.
• Screenshots and steps are stale. The interface changed twice since the walkthrough was recorded, yet the course is live and learners are still marked complete.
• Versions are a mystery. Files named "final_v3_REAL" sit on shared drives, and nobody can say which one learners actually see.
• Usage is unknown. Every course costs maintenance whether or not anyone takes it.
Here is a quick test. Pick a rule that changed recently and ask five questions. Can you list every asset that mentions it within an hour? Does each have a named owner? When was each last reviewed? Which version is live? How long would it take to fix them all? If the honest answers are "no", "no", "I'm not sure", "I'm not sure" and "weeks", you are carrying debt.
How the interest compounds
Debt is not just untidy. It costs, and the cost grows every month.
Update cost rises with every copy. If a fact is duplicated across twenty courses, changing it means finding all twenty, editing each, republishing and checking. The effort scales with the number of copies, not with the size of the change.
Trust erodes quietly. Learners who meet conflicting answers in two courses stop trusting either. Worse, a stale course teaches the old way with the authority of an official programme. For compliance and safety topics, that is risk, not inconvenience.
Discovery gets harder. When several near-duplicate courses exist, people can't tell which is current. They pick one at random or give up and ask a colleague.
Opportunity cost lands on the team. Every hour spent patching is an hour not spent designing something that changes behavior. Teams drowning in maintenance often conclude they are "too busy" for strategic work, when the real problem is that old loans are eating their capacity.
The pattern is easiest to see in a simple illustration of what one policy change costs as the same fact spreads across more courses.
The numbers are invented to show the shape, not measured. With one course, copying is cheaper, which is exactly why teams drift into it. By the time a fact appears in twenty or forty courses, the same change costs many times more than it would in a library where the fact lives in one place.
AI raises the principal
Generative AI changes the debt equation in two ways, one dangerous and one hopeful.
The danger is volume. Cheap creation means more assets, faster, and most new assets arrive with no more ownership or structure than the old ones. AI-generated content also often lacks provenance: nobody recorded which source document, which version or which prompt produced it. When the source changes, you cannot tell which outputs are affected. And because regenerating is easier than finding and reusing, teams create a fresh near-duplicate instead of updating the original.
The hope is that AI is also a capable debt collector. Given an inventory export, it can cluster near-duplicate courses, extract the facts each one states, suggest tags and topics, flag passages that contradict a current policy and draft updated wording for review. With a human in the loop approving every change, it shrinks the cost of the cleanup that used to make teams give up.
The rule of thumb is simple: use AI to reduce the debt, and never let it add to the debt without structure.
The structural fix: write once, reference everywhere
The deepest cause of the update-cost problem is copying, and the deepest cure is single-source content. Each fact, procedure, definition or policy statement lives in one place and is referenced wherever it is needed.
Anyone who has used a spreadsheet already understands the principle. If ten cells each contain the typed number 5%, changing the rate means editing ten cells and hoping you found them all. If they all point to one cell that holds the rate, you change it once. Applied to learning content, the difference looks like this.
In the copy model, every extra course is another edit and another place to slip. In the single-source model, you update the master once, a named person approves it, and every course that references it picks up the change.
Four things make this work in practice:
• Modular content. Write in chunks that can stand alone: a concept, a step, an example, a policy statement. Avoid slide-shaped blocks tied to a single course.
• Content separated from presentation. The same chunk can appear in a course, a job aid and a quick-reference card without being retyped.
• Metadata. Each chunk carries a topic, audience, owner and review date, so you can search, filter and audit.
• Version history and an audit trail. Every change is recorded, including AI-assisted ones, and can be reversed.
You don't need an expensive platform to begin. A shared "source of truth" page for each critical policy, with courses linking to it instead of pasting from it, already removes much of the duplication. Tools help the habit scale, but the habit comes first.
A debt-reduction plan
You can make real progress in a quarter. The order matters: understand the debt before paying it, and stop adding to it while you do.
1. Inventory. List every asset with its title, format, location, owner, audience, last review date, topics covered and, if you can get it, usage. A spreadsheet is enough. Time-box the effort and start with the riskiest areas: compliance, safety, regulated content and product facts.
2. Triage with four verbs. Keep what is accurate, used and well structured. Refactor what is valuable but duplicated or locked in a poor format. Merge courses that cover the same ground. Retire what is unused, outdated or superseded. Be braver about retiring than feels comfortable, because every asset you keep is an ongoing obligation.
3. Name owners and expiry dates. Every asset gets one accountable owner and a "valid until" date. Add event triggers too: a policy change, a product release or an incident should force a review ahead of schedule.
4. Create masters for your most duplicated facts. Identify the ten facts that appear in the most places, give each a master location and replace the copies with references.
5. Install review gates and a change log. Define who must approve what, require approval before publishing and record every change. Treat AI-assisted edits exactly the same way.
6. Budget for debt service. Reserve a fixed slice of team capacity each quarter for maintenance and protect it. Debt that is only paid "when we have time" never gets paid.
7. Stop new borrowing. Make every new build pass a short check: who owns it, when does it expire, which master content does it reuse, and what is the plan to retire it?
Making the business case
Executives will not be moved by "modular, tagged, single-source architecture". They respond to risk, cost and speed, so translate.
Run a fire drill. Pick a real policy change from the last six months and measure how long it took, or would take, to find and correct every affected asset. Count how many still say the old thing. The result is a risk number that needs no explanation.
Then track a few simple measures: time to update after a change, the share of assets with a named owner, the share past their review date, the number of known duplicates and the maintenance hours per quarter. Report them the way finance reports debt: where it sits, what it costs and whether it is falling.
Frame the investment as protection. Fewer stale answers mean lower compliance exposure, fewer audit findings and less rework, and a team that finally has time to build the practice experiences that change behavior.
Pitfalls to avoid
• Buying a platform to fix a practice problem. A new system loaded with unowned, duplicated content gives you the same debt in nicer packaging.
• Restructuring everything. Start with the highest-risk, highest-reuse material. Over-engineering low-value content wastes the effort you are trying to save.
• Treating the audit as a one-off. Without owners and review dates, the debt returns within a year.
• Retiring silently. Tell stakeholders and learners what is going, why, and where the equivalent now lives.
• Letting AI generate into the void. Every AI-assisted asset still needs an owner, a source and an expiry date.
Start with one fact
Technical debt is not a sign of failure. It is the natural result of moving fast, and every library has some. The mistake is pretending it isn't there. A learning team that knows what it owns, who is accountable for each piece and how a change travels through the library can respond to a new policy in hours rather than weeks, and can use AI to move faster without losing control.
Start small. Pick one critical policy, find every place it appears and give it a single home. You will learn more about your debt in that one exercise than in any audit report, and you will have made your first repayment.
Comments