The question behind AI and online courses is fair: if a learner can ask for an explanation, examples and a study plan at little or no cost, what are they paying your training business to provide?
A course can still earn its price through useful structure, credible expertise, practice, feedback and evidence of progress. AI can support some of those things too. The task is to show where your particular programme improves the learner’s outcome, time or confidence, and to deliver that value at a sustainable cost.
Information can still be worth paying for
It would be too simple to declare recorded courses obsolete. A well-organised programme can save time, sequence difficult material, explain assumptions and give a beginner a reliable starting point. A learner may reasonably pay for that convenience and clarity.
The challenge is a vague premium promise attached to material the customer can easily replace. More videos, more modules and a larger resource library do not automatically create more value. They may create more work for the learner.
Review the claim you actually sell. “Understand the basics at your own pace” is different from “perform a task to a defined professional standard”. The second promise requires stronger teaching, practice and evidence. Neither should be confused with a guaranteed job, earnings increase or licence to practise.
Remove the videos from the description
As a design exercise, describe the programme without mentioning video hours, worksheets or platform access. What is left? Perhaps an ordered route through a difficult skill, supervised practice, expert correction, a recognised assessment or access to people solving the same problem.
If very little remains, investigate whether the learning design is weak or the description simply fails to communicate it. Do not assume the existing course is worthless. Speak with learners and inspect the work they produce.
Use three questions for each major module: what does the learner practise, how do they know whether it was good enough, and what changes after feedback? A module that only adds more reading may still be useful, but it should have a clear place in that progression.
Before and after: make one course promise tangible
Illustrative example, not a PureRapid client result. Consider an online course for operations managers who need to prepare a monthly spreadsheet report. The original offer is “30 videos, templates and lifetime access”. Those describe delivery assets; they leave the buyer to infer what the course helps them do.
A stronger draft offer is: “Prepare a monthly report from an untidy practice dataset, reconcile the totals and explain three decisions to a colleague. Work through a sample brief, submit one report for tutor review, then revise it against the feedback.” This is proposed offer copy to test, not an established learner outcome.
| Design choice | Concrete provision | Boundary to make clear |
|---|---|---|
| Practice | An original fictional dataset with missing fields, duplicates and conflicting totals. | Use supplied practice data, not an employer’s confidential records. |
| Feedback | One marked report and one resubmission, within a stated response window. | Describe what is reviewed and what additional support costs, if offered. |
| Evidence | A revised report plus a short explanation of checks and choices. | Completion does not establish a regulated qualification or guarantee workplace performance. |
| AI use | Permit help where it fits the task; ask learners to disclose and verify it. | Do not claim unaided competence if the assessment allowed substantial assistance. |
The buyer can now evaluate the task, support and evidence. The provider can calculate the effort required to deliver them. A concise self-paced version may still suit a learner who does not need feedback; it needs a correspondingly clear promise.
AI and online courses need the same outcome test
AI can explain a concept, generate practice material and offer feedback. The reliability and educational value depend on the system, subject, instructions and the learner’s ability to judge the response. Do not make “AI cannot teach” the foundation of your positioning.
Instead, test the learning task. Does the response use the correct standard? Does it recognise an important mistake? Can the learner explain the reasoning and apply it to a new case? Does the tool provide an answer when the learning objective requires the learner to work it out?
UNESCO’s guidance on generative AI in education takes a human-centred approach and calls for considered educational design and evaluation. It supports examining how a tool is used, rather than treating access to it as proof of educational quality.
For tasks where competence must be demonstrated without assistance, assess under appropriate conditions. For tasks where AI use is part of professional practice, assess whether the learner can use it responsibly, verify the output and recognise when specialist judgement is needed. The conditions should match the claimed outcome.
Show the buyer how learning will be evidenced
Explain the practical task, the assessment criteria and the form of feedback before purchase. A sample exercise and a clear rubric can make the learning promise more concrete than another broad claim about transformation.
Use learner work or results only with permission and adequate context. State the starting point, the assessment conditions and what the evidence does and does not establish. Selected examples are not proof that every learner achieves the same result.
Completion is also a limited measure. Someone can finish every video without being able to apply the skill. Conversely, an experienced learner may achieve their goal without consuming every resource. Choose measures that fit the promise.
Be precise about certificates. A certificate of completion, an assessed award and an externally recognised qualification are different things. Verify any recognition with the relevant awarding or professional body before making the claim.
A small assessment rubric makes the offer testable
For the reporting example, agree three criteria before reviewing submissions: accuracy, with totals reconciled and unresolved discrepancies disclosed; reasoning, with important assumptions explained; and communication, with a colleague able to understand the proposed decisions and their limits.
Use “meets the criterion”, “needs revision” and “not yet evidenced”, supported by a specific observation. Compare the first attempt with the revision, and use a fresh task if the claim requires transfer beyond the taught example. Completing a checklist alone does not prove workplace competence.
Test this design with suitable learners and record where reviewers disagree. That reveals a problem in the rubric or teaching before the sales page promises a dependable outcome. Publish real learner evidence only with permission and the assessment conditions attached.
Add support where it changes the learning
Live sessions, communities and one-to-one feedback are options, not automatic upgrades. A discussion group with no clear purpose may add noise. A short, targeted review of a learner’s work may solve a specific difficulty.
Find the points where learners repeatedly stall or misunderstand. Consider a worked example, a better sequence, an expert feedback checkpoint or a small-group session. Test whether the change improves the relevant outcome before applying it throughout the course. Include learners who disengage or struggle in the review; feedback only from enthusiastic completers can hide the people the course serves least well.
Some independent learners prefer a concise self-paced product at an appropriate price. Others need structured support or an assessment route. Clearly described tiers can serve those needs, provided the lower tier still delivers its stated promise.
Make the support promise sustainable
Illustrative example, not a PureRapid client result. A programme enrols 30 learners at £400 each, generating £12,000 net revenue. Variable platform, payment, material and expected refund costs total £40 per learner, or £1,200. Acquisition costs are £2,400 for the cohort.
If each learner receives 45 minutes of feedback at an assumed fully loaded tutor cost of £50 an hour, feedback costs £1,125. The cohort leaves £7,275 before fixed course production, administration, other overhead and tax.
If feedback actually takes 90 minutes per learner, it costs £2,250 and the remaining contribution falls to £6,150. The extra support may be worthwhile, but the business needs to price, schedule and deliver it deliberately.
These are GBP assumptions for one cohort, excluding VAT where chargeable and recoverable. The example assumes no other variable delivery cost and includes an expected refund allowance within the stated £40. It is not a course-pricing benchmark. Use the provider’s actual VAT treatment and cost structure.
Make the support boundary clear
“Unlimited feedback” can sound attractive while leaving learners unsure what response to expect and tutors unable to plan. Define scope, response times, submission limits and escalation clearly. A dependable promise is more valuable than an expansive one the team cannot fulfil.
For the example offer, the feedback allowance must cover the first review and the promised resubmission together. Time both during the pilot. Use the course profitability guide to add fixed production and delivery commitments before choosing a final price or cohort size.
Test a stronger offer before rebuilding the school
- Audit one learning outcome. Identify the practice, feedback and evidence currently attached to it.
- Talk to recent learners. Ask what they used, where they struggled and what helped them apply the skill.
- Compare the real alternatives. Include free resources, AI assistance, books, another course and learning through work. State where each may be sufficient.
- Pilot one improvement. Measure learner work, support time and willingness to pay, with clear consent and expectations.
Do not force a community, change platforms or add a chatbot simply to signal that the business has responded to AI. Choose the change that addresses the learning gap and can be maintained.
Make the value visible before the sales conversation
Your course page should make the audience, prerequisites, practice, support, assessment and limitations easy to understand. Show why the structure matters and what the learner is expected to contribute. People cannot evaluate value they cannot see.
If that value is sound but enquiries are not becoming bookings, investigate the course buying journey. A sales problem does not, on its own, prove that AI has removed demand for the subject.
The next decision is specific: keep a useful self-paced product, strengthen a learning gap, introduce a properly supported tier or retire an offer that no longer serves its audience. Make that decision from learner evidence and delivery economics, rather than the loudest prediction about the future of courses.
Sources and method
The decision frameworks and labelled examples are PureRapid editorial analysis. Documentation checked on 21 October 2026.
- UNESCO guidance for generative AI in education and research. Human-centred approach and evaluation of educational use; not a forecast of commercial course demand.