How do you script a software how-to answer video?

A software how-to answer video script has five spoken parts: the answer, the prerequisites, numbered steps, the common failure, and one next step. Say each menu label out loud. Storyflow keeps most of a creator video for the body and the ask for the end. A how-to still opens with the answer. You record it yourself.

How to set up SSO with OktaLong-form AI Tools

[COLD OPEN]

Most teams set up SSO in the wrong order, and it locks out half the company on launch day. [pause]

## Why the order matters

Okta only syncs assignments that exist after the identity provider is connected. [slide: 2]

## Step 1: Add Okta as the identity provider

[screen] In Acme Admin, open Security, then Single sign-on, and add Okta before you touch a single user.

## Step 2: Assign the app to a group

Assign the app to a group, not to people, so new teammates get access when they join the group. Keep one backup admin login outside SSO while you test.

## Step 3: Test, then require SSO

Log in once as a test user. When it works, turn on Require SSO for everyone.

## Recap

Provider first, groups second, test before you require it.

## CTA

The full checklist is in the description. Subscribe for the next admin guide.

Speaking time: ~9:04 · 1,360 words
Sample data, fictional brand

What are the five parts?

Five, in this order. The numbered steps are the video. The other four parts stay short. One task, one product.

PartWhat you say
Hook and answerName the task and the product, then give the answer in one sentence. Name the one failure you will show later.
ContextThe plan, role or version the viewer needs before step one.
Numbered stepsOne action per step. Speak the menu and button labels that are on screen.
Gotcha and checkThe usual failure, then the result the viewer should see if it worked.
Recap and one next stepRepeat the one-sentence answer, then one place to go next.

One task, not a product tour. The page around the video is in what an answer video is.

How long should each part run?

Storyflow's how-to template opens on an unexpected result in the first 8 seconds, and puts the strongest insight at 60–70% of the runtime.1 The page tells creators not to open with their best point.1

60–80%
of the video is core content in Storyflow's script framework. The closing ask is the final 10%.Storyflow, 7-part YouTube script framework, 11 Jan 2026

A software how-to opens on the answer. The viewer searched for a fix, so the first lines are that fix. The steps still take most of the runtime. Treat 60–80% as Storyflow's range for them. Google publishes no script clock.

What do the spoken lines sound like?

Write the lines as they will be said. Replace every brace.

  • How do you {task} in {product}? You {one-sentence answer}. I'll walk through the {number} steps, plus the one setting that trips people up.
  • Before you start, you need {plan or role} and {version}. If you don't see {menu}, that's why.
  • Step one: open {menu}. Step two: select {option}. Step three: enter {value} in {field}.
  • The next step is where that setting shows up.
  • If {symptom}, check {setting}. You'll know it worked when you see {result}.
  • So, to {task} in {product}, you {repeat the answer}. For the next task, watch {next video}.

Read it aloud before you record. Rewrite any line that only works if you point.

What prompt drafts this from your docs?

Paste this into ChatGPT or Claude with the help article attached.

  1. 01Write spoken lines only for "how to {task} in {product}". No scene labels. Use only the attached source.
  2. 02Open with the one-sentence answer, name one mistake, then the plan, role or version.
  3. 03Numbered steps from the source labels. Past the midpoint, point back to the mistake, then the failure, a check, the answer again, and one next step.
  4. 04Name the passage after each factual line. If a step is missing, say so.
The prompt

You still read every step against the product you ship today. A model will fill a gap with a step that sounds right.

Poppy AI builds a script from videos, transcripts and a saved voice.3 Subscribr writes one from an idea, in a channel voice.4 vidIQ writes an intro, talking points, transitions and a close.5 Restream takes a topic, an audience and a tone, then a script you edit, with no account.6 Check the draft against the five parts, with your docs attached.

What has to be said, not only shown?

Google asks for important content as text, with no extra requirements and no special optimization for AI Overviews or AI Mode.2 Say "Open Settings, then Integrations." An unspoken click is not in that text.

The rest of the upload has its own pages. The title is in how to write YouTube titles. The description and chapter timestamps are in the description and chapters template. What a chapter changes about a citation is in do chapters affect citations. Transcripts are in do YouTube transcripts get used. Why a video gets cited is in what makes Google cite a video.

Who should be on camera?

You. A software how-to shows the person who knows the product, and B2B buyers are looking at that person. An AI avatar can speak the lines. Subscribr will render an avatar or a faceless video, and it will edit a recording if you make one.4 Make the recording.

FoxCue is the video OS for SaaS. It reads your site and Search Console, suggests the buyer question, and writes the script from sources you approve. It shows the passages it used. A grounding check flags a claim those passages do not support. You record on a teleprompter, on camera.

Studio recorderHow to set up SSO with Okta · 16:9
REC 0:09

Most teams set up SSO in the wrong order, and it locks out half the company. Here's the order that works. First, add Okta as the identity provider. Then assign the app to a group, not to people. New hires get access on day one, and nobody files a ticket.

Stop0:09Camera and microphone · the script is not recorded on screen
Sample data, fictional brand
Read the five parts on the teleprompter, with the camera on you and the product on screen.

After you review, the edit can cut silences, filler and retakes, and add captions. You upload the file. FoxCue does not upload it, and it will not get the video cited.2

Sources

  1. Storyflow, 7-part YouTube script framework, 11 Jan 2026.
  2. Google Search Central, AI features and your website, checked 25 Oct 2026.
  3. Poppy AI, checked 25 Oct 2026.
  4. Subscribr, checked 25 Oct 2026.
  5. vidIQ, AI Script Generator, checked 25 Oct 2026.
  6. Restream, AI YouTube Script Generator, checked 25 Oct 2026.
FAQ

Questions

Can I use ChatGPT to write the script?

Yes. Attach the help article and the five-part prompt. Then check every step against the live product. A step that sounds plausible can still be wrong.

How do I turn a help article into this script?

Paste the article in with the prompt. Map the result onto the five parts, one action per step, labels spoken. Drop any line the article does not support.

Should an AI avatar read a software how-to?

Put the person who knows the product on camera, with the product on screen. An avatar can voice a script. Some creator tools will generate that video. This template is for a person recording it.

Will this structure get the video cited?

No one can promise a citation. Google sets no special optimization for AI Overviews. The five parts make the answer easy to quote. Google decides whether to show it.

Keep reading

All in YouTube & AI Overviews

Make one video for every launch

Claim a founding spot. We'll email you to book your recording.

Join the waitlist

Free. You record on camera with our teleprompter, FoxCue writes the script from your own sources and edits it. Every claim is cited to your sources.