A retrieval-augmented generation explainer for product managers
This excerpt from the project opened in Studio shows a sample refund-policy question and the opening RAG diagram. It comes from the same request shown in the walkthrough above.
Give ChatGPT a concept, an approved passage or a public source and ask ngram to explain it to a specific audience. The explanation profile quotes the format and rate before creation. Watch the finished project in ngram.
This excerpt from the project opened in Studio shows a sample refund-policy question and the opening RAG diagram. It comes from the same request shown in the walkthrough above.
Paste just the approved passage. Leave customer emails and actual invitations out. Open the finished project. Keep the exact policy in writing next to the explanation.
Use approved wording for what changed. Avoid supplying an internal changelog URL. Keep exact player controls in the written instructions. Use a screen recording if those controls are the lesson.
The example starts with a product manager who wants to understand retrieval-augmented generation (RAG). Use the same pattern for a concept your audience needs explained.
On ChatGPT web, an eligible Business, Enterprise or Edu workspace can add a custom MCP server from Plugins. Enter https://mcp.ngram.com/openai and complete ngram sign-in. Ask your admin if you cannot add the connection or run its write actions.
Ask “Explain retrieval-augmented generation to a product manager. Show why retrieving documents changes the answer.” For a source-led explanation, paste the approved passage or provide a public HTTPS source page or PDF.
The explanation profile prepares a detailed, narrated 16:9 explanation with neutral branding and length determined by the source. Review the per-second rate and confirm the same request to start. If the request changes, check the new quote.
Check the job’s status in chat. When it is ready, use the project link to review the video in ngram. Check the explanation against the source before sending it to your audience.
The /openai profile provides three actions for explanations. The general developer endpoint at https://mcp.ngram.com exposes a broader toolset to compatible clients.
The create action first returns a format and quoted rate. It starts the job after confirmation. Name the audience and supply the question or source to explain.
Ask for progress on the explanation you already started. Once it completes, open the result from that job rather than starting another request.
If the source is wrong, cancel the active job before requesting a corrected explanation. Check its status first. Cancellation does not restore work already completed.
Open the resulting ngram project for further work. Use Studio when you need a real screen recording, a branded walkthrough or a creative edit outside the explanation profile.
Use this profile when the audience needs an idea explained and the source contains the facts. Use a recording when the person needs to copy exact clicks or verify a private product screen.
A product manager needs to understand why a system retrieves documents before answering. Ask for a conceptual explanation using an example question. Use the original technical reference to check the result.
See the explainer playbookGive the assistant an approved release passage and ask it to explain what previously confused users. Use a separate recording if the audience needs to see the actual changed controls.
See the changelog playbookProvide the public help text and specify which idea new users misunderstand. Keep exact commands and exceptions in the written article where customers can copy them.
See the help center playbookThese are ngram project features. Open Studio when the next step calls for more creative control than the fixed explanation profile offers.
Use visuals to explain a relationship that the source describes.
Explore featureReview the narration in Studio for important technical terms.
See featureCheck the spoken explanation against the source passage.
See featureEdit the project after inspecting it in ngram.
See featureUse a diagram when the relationship is easier to see than hear.
See featureCheck captions for names, acronyms and technical vocabulary.
See featureUse Studio for a branded variant after the explanation is reviewed.
See featureCreate a separate language version when the audience needs it.
See featureThe ChatGPT explanation profile accepts text or a public HTTPS page or PDF. Use these Studio workflows when the source or desired output needs a different approach.
Use the ngram website tools below for a separate creation or editing job. Their controls are not part of the three-action ChatGPT explanation profile.
Give a specific audience a reason to watch and a source against which you can check the explanation.
Explain a technical concept with a public reference developers can verify.
Give product teammates an example that clarifies an unfamiliar mechanism.
Explain the idea behind a product before asking someone to watch its demo.
Introduce an approved concept before a customer training session.
Explain a recurring question, keeping exact steps in the help article.
Adapt a public release note into a concept explanation for a new audience.
Use the ChatGPT profile for a narrated explanation, or choose a different entry point when your project starts from a repository, event or public page.
Compare the full developer profile with the limited explanation tools in ChatGPT.
Compare MCP profilesRequest a video where the pull request, issue or release supplies the product context.
Explore the GitHub AppStart from the open page when its content belongs in a video brief.
Explore the Chrome extensionStart from a repeatable event when the request needs to run outside chat.
Explore the Zapier integrationChatGPT explanation Custom connection | Developer MCP Broader toolset | ngram Studio Project workspace | Zapier Event automation | |
|---|---|---|---|---|
| Where it starts | ChatGPT web | Compatible MCP client | ngram Studio | Approved trigger record |
| Source | Concept, supplied text, public HTTPS source | Sources supported by the connected profile | Project brief and source files | Mapped trigger fields |
| Creative choices | Fixed narrated 16:9 explanation | Creative controls on the full profile | Project and editor settings | Fields exposed by the action |
| Best fit | Explain a concept to a named audience | Create and edit in a compatible assistant | Demonstrate actual product screens | Route recurring drafts for review |
FAQ
Still curious? Read the ChatGPT walkthrough
Open your eligible ChatGPT workspace, select the ngram connection and ask for a narrated explanation with a source you can check.