What Should You Ask an eLearning Vendor About AI? – 12 Questions
Every eLearning vendor now says they use AI, which sounds good in a sales call and might be great for your project. It might also mean your confidential content is going into a consumer chatbot, or that nobody checks the narration script before it ships, and you can't tell from the pitch alone. What should you ask an eLearning vendor about AI before you sign?
The most useful eLearning vendor AI questions cover four things: where AI is used in their process, who reviews what it produces, how your data is protected, and how the savings show up in your price and schedule. A qualified vendor answers each one specifically, task by task, without vague generalities, naming their tools, describing their human review step and showing you how AI changed their estimate. The twelve questions below are grouped so you can use them in a sales call, an RFP or a contract review.
Why Ask Vendors About AI at All?
You should ask vendors about AI because it now touches scripts, voices, images, quizzes and code, which means it touches your content, your learners and your budget. Used well, it makes projects faster and more affordable, but used carelessly, it creates errors and data risks you'll own after launch.
Here's my read: the vendors doing this well are happy to discuss it, because they've thought through the workflow and they're proud of it. Vague answers usually mean that thinking hasn't happened yet.
Questions About How the Vendor Uses AI
1. Where in your process do you use AI, step by step?
You want a specific list, not a slogan. Good answers sound like "first-draft storyboards, quiz item drafts, synthetic narration and alt text drafts," and if the answer is "throughout our process," request the list anyway.
2. Which tools do you use, and on what kind of account?
Expect real product names, such as ChatGPT, Claude, Microsoft Copilot, ElevenLabs or Synthesia, and expect them on business or enterprise plans. Business plans typically offer stronger data controls than free personal accounts, so ask which settings they use.
3. How does AI change your estimate and timeline?
Ask them to show which line items decreased and by approximately how many hours. Savings are real but uneven, since they're large on drafting and narration and small on analysis and design decisions. A flat "20% off because AI" isn't an estimate.
4. Who reviews AI output, and what do they check?
This is the question that separates mature vendors from the rest. You want a named role (instructional designer, editor, developer) and a clear checklist: accuracy against source content, tone, accessibility, alignment with objectives. Human review is simply good production practice, the same way nobody ships a course without quality assurance.
Questions About Your Data and Ownership
5. Does our content train anyone's model?
Ask whether any tool they use can train on your inputs, and how that setting is disabled. The answer should be specific to each tool, and if they don't know, that tells you what you need to know.
6. How do you handle confidential or regulated content?
Some content can't go into outside tools at all: personal data, unreleased product details, safety-critical procedures or anything under export control. Ask how they identify that content and what they do instead, because a responsible vendor has a rule and follows it.
7. Who owns the AI-assisted outputs?
Your contract should say that scripts, images, audio and source files belong to you on payment, regardless of the tools used. Ownership of purely machine-generated material can be legally complicated. The U.S. Copyright Office's copyright and AI initiative publishes its current guidance, and it's worth having your legal team glance at it before you finalize terms.
8. Are synthetic voices and avatars properly licensed?
If the vendor uses a cloned voice or a custom avatar, ask how consent and licensing were handled. Stock voices and avatars from established platforms come with license terms, while a clone of a real person requires that person's written permission.
Questions About Quality and Accuracy
9. How do you catch factual errors?
Language models can produce confident but incorrect statements, so ask how the vendor verifies every fact against your source material and how your subject matter expert (SME) fits into that review. "We trust the tool" is not acceptable. "We trace every claim back to the source document" is.
10. How does AI affect accessibility?
AI can draft alt text, captions and transcripts quickly, which is a genuine benefit for accessibility, but ask who checks those drafts. Automatic captions need correction, and alt text needs someone who understands the image's purpose in the lesson.
11. Can you show a before-and-after sample?
Ask to see an AI-drafted storyboard next to the reviewed, final version. The difference tells you what their review step actually catches and how much expert work goes into the final product.
12. What's your written AI policy?
A short written policy shows they've made decisions on purpose: approved tools, data rules, review steps and disclosure. If they built it on a recognized reference, such as the NIST AI Risk Management Framework, that's a good sign of discipline. Ask how often they update their policy, since tools change quickly.
What Do Good and Weak Answers Sound Like?
Good answers are specific, name a person and connect to your project, while weak answers are generic, name a tool and stop there. This quick comparison helps during a sales call.
| Topic | Strong answer | Weak answer |
|---|---|---|
| Where AI is used | Named tasks with named tools | "We use AI everywhere" |
| Review | A role, a checklist and a sample | "Our team looks it over" |
| Data | Business accounts, training turned off, a rule for sensitive content | "It's all secure" |
| Savings | Hours by line item in the estimate | A flat discount |
When Should You Ask These Questions?
Ask these questions at three points: in the request for proposal (RFP), in the discovery call and again in the contract. Each stage produces a different kind of answer, and you want all three.
In the RFP, request written answers to all twelve, since written answers are easy to compare side by side and difficult to revise later. In the discovery call, select the three or four answers that seemed thin and ask for an example. "Walk me through the last storyboard you drafted with AI" gets you further than "Do you use AI responsibly?"
In the contract, turn the best answers into terms. If the vendor said they use business accounts with training turned off, document that, and if they said every script gets an editor's review, make it part of the acceptance criteria. Promises made in a sales call fade quickly, but contract language doesn't.
One more point in the vendor's favor is that these questions also protect them, since a vendor with clear rules on data and review is less likely to end up in a dispute with you six months later. Most reputable vendors will thank you for asking.
How Do You Score Answers to eLearning Vendor AI Questions?
Score vendor AI answers on specificity, review and data handling, and weight them alongside the usual criteria like portfolio, process and price. AI capability shouldn't decide the contract on its own, because it's only one factor in how well the vendor will deliver.
A simple approach is a 1-to-3 scale per question: 1 for vague, 2 for partly specific, 3 for specific with evidence. Twelve questions give you a score out of 36. Example: a vendor who scores 30 or more is likely running a thoughtful AI workflow. Below 20, you'll probably spend your own time monitoring their output.
If you'd rather have help designing your own AI workflow, not just evaluating a vendor's, that's a different job. An AI eLearning consultant can audit your current process, pick tools and pilot them with your team.
Twelve Questions, One Clear Picture of the Vendor
The short version: ask where they use AI, which tools and accounts they rely on, how it changes the estimate, who reviews the output, how your data and ownership are protected, and how they keep quality and accessibility high.
The best vendors will welcome these questions, because they make their real strengths visible. Copy the twelve into your next RFP or discovery call, score the answers, and you'll know quickly who's using AI to do better work and who's just using the word.