Film a task. Turn it into a how-to.

Show someone how to do something by doing it once and explaining it aloud. AI can help turn that demonstration into a how-to with clear steps, pictures or video clips, and the details another person needs to follow it.

This can be a physical task, a computer task, or a combination. Film yourself or a colleague operating equipment, assembling something, or handling paperwork. For computer work, talk through what you do while recording your screen. You can also film the screen with your phone; a screen recording usually makes small text easier to read.

You don't need to arrive with a recording—or even a task in mind. Start by asking the AI to help you brainstorm things you need other people to be able to do. Think about what you explain repeatedly, what only one person knows, and where people tend to get stuck. Choose something useful, then work out how to show it.

The tools are changing quickly. By the time you try this, their capabilities and prices may be substantially different. The prompt will have the AI search for current options that suit your task and can use both the demonstration and your narration. Gemini and TwelveLabs are starting points for video analysis; Guidde and DeepHow offer tools for creating instructional material. For computer workflows, Scribe and Tango are also worth investigating.

While demonstrating, explain the judgments as well as the actions: why you choose that setting, how you know a document belongs in that folder, what you check before moving on, and what a successful result looks like. You can speak naturally. The AI will organize the explanation and ask about anything it cannot see or understand.

Review the resulting guide with the person who demonstrated the task. If possible, have someone who is learning the task try following it and use their questions to improve it. You will end with instructions you can share, including an honest account of what has been checked.

For the user

Give the prompt below to an AI. It will help you choose something worth teaching, research suitable tools, and plan the demonstration. You or a colleague will show the task and explain it aloud. The AI will draft the how-to, check it against the recording, and help you revise it after review.

Use a complicated machine.

Show one operation: setting it up for a particular job, choosing the right settings, and checking the result. An experienced machinist demonstrating a setup on a CNC mill is one example.

https://www.haascnc.com/video/tipoftheday/cn05fx55pqc.html

Handle the equipment problem everyone asks you about.

Reload the label printer, showing the correct supplies, where they go, and what a good printed label looks like.

https://www.youtube.com/watch?v=qgcdmejldps

Assemble something.

Demonstrate a storage unit, product, or kit. Show which way parts face, the order of assembly, and the checks that tell you it is right.

https://www.ikea.com/es/en/customer-service/product-support/assembly-guides/

File documents so someone can find them later.

Show how you scan a signed document, name it, choose its folder, add any required information, and find the right copy again.

https://support.microsoft.com/en-us/sharepoint/libraries/view-and-edit-information-about-a-file-folder-or-link-in-a-document-library

Complete a computer routine.

Submit an expense claim, create an invoice, set up a customer record, or prepare a recurring report. Explain the choices as you make them and show how you check that the task is finished.

https://scribe.com/tools/documentation-generator

Connect physical and computer work.

Receive a delivery, check what arrived, and then enter the quantities or discrepancies in the inventory system. The demonstration can combine phone footage and a screen recording.

https://help.shopify.com/en/manual/products/inventory/inventory-transfers/barcode-scanner

Done when

You have a shareable how-to grounded in the demonstration, checked by the demonstrator and revised with learner feedback if available. You know what has been checked and what still needs a real-world trial.

Guidance for the AI

These are general instructions, not a description of this user's particular AI system, tools, access, or circumstances. Adapt the method to the user's goal and the capabilities you can actually use, while preserving the activity's essential sequence, review points, and intended result.

Start with the activity and the user's context

Read the complete activity and guide the user through it. Use the activity prompt yourself; the user does not need to paste it again. Keep the responsibilities of the user, the AI, and any colleague clear.

Begin with what the user has already supplied. Establish the intended result and resolve missing context through focused questions when needed. Work within the activity's stated scope, permissions, required confirmations, and stopping points, including any deliberate order for introducing information. Make progress with the available information and reasonable, clearly identified assumptions where the activity allows them.

Check the capabilities needed for this work

Inspect relevant available tools, connectors, and data. Use small, relevant, non-destructive checks to establish whether necessary access or a capability actually works. Naming a product does not establish access to it. Let this activity determine what needs checking; do not impose a universal preliminary questionnaire or a full system audit.

When useful access is missing:

• Explain what you cannot reach and how that material would help.
• Investigate practical setup, connection, export, or other access alternatives before asking the user to gather material or troubleshoot.
• If progress depends on access, request only the specific user action or permission needed to enable it or provide the material another way. Continue useful independent work while awaiting the response.
• Offer to walk the user through an export or upload. Welcome files, links, copied text, rough notes, or an interview as supported by your system and the activity, and use the best available route. If the missing access is optional, continue with what is available.

Ask focused questions and respect the answers

Ask questions only when necessary to complete the work well or satisfy the activity's required review points. Build on answers already supplied. Ask one question at a time, or a small group of closely related questions. Wait on work that depends on the answer, and continue useful work that does not. Explain unfamiliar language with examples where helpful.

Confirm answers or decisions you supply on the user's behalf before treating them as the user's final answers or decisions. This does not require separate approval of every sourced finding; follow the activity's specific review requirements.

Keep evidence and independent contributions distinct

Distinguish evidence, inference, and missing information. Check your work against the source material.

Use independent agents when your system actually supports delegation and independent investigation or review would improve the work:

• Give each agent a bounded brief and the context appropriate to its role.
• Keep discovery independent until synthesis where the activity requires it. Preserve any requirement for the user or colleagues to contribute before seeing other explanations.
• Follow the activity's stated fallback and any required confirmation when independent agents or fresh contexts are unavailable. Preserve reviewer isolation where required; describe shared-context or sequential passes accurately.

Produce, check, and revise the result

Do the substantive investigation and production. Carry the activity through to the most complete, usable result within its scope, rather than ending with a plan or handing the main work back to the user. When missing information does not prevent useful work, proceed with clearly identified assumptions where the activity permits them.

Check the result against the activity's completion criteria. At each substantive delivery, state what is complete and any material unfinished work, assumptions, helpful missing information, and remaining uncertainty. Explain how these affect what the user can use, and help the user revise the result.

Preserve the activity's stopping point. Do not turn a rehearsal, investigation, or test-design activity into execution that the user has not authorized.

Your AI prompt

Help me choose something to teach, then make a how-to

Help me brainstorm things I need other people to be able to do, then help me plan how to show them. These could involve operating a machine, assembling something, filing documents, working entirely on a computer, or moving between physical and computer work. After we choose a task and plan the demonstration, help me turn a recording of it into a usable how-to. Carry the work through capture, drafting, review and a shareable result.

Assume I am comfortable using AI but have no video-production or instructional-design expertise. I may already have a task or recording, or I may need help finding something worth teaching. Build on what I provide and skip work we have already completed.

Brainstorm what others need to learn

Start with what you already know about my work. Ask one focused question at a time where needed. Help me notice tasks I explain repeatedly, tasks only one person knows, tasks people often get wrong, and work I need to hand over before an absence or change of role.

If I identify a role, team or project, try accessible process documents, onboarding material, support questions, related emails and chats, and meeting notes. Look for repeated requests for help, gaps in existing instructions, and work that depends on one person's knowledge. Keep the search focused; if records are unavailable, brainstorm with me rather than asking me to assemble a large packet. Distinguish documented needs from your suggestions.

Offer a few possibilities appropriate to my work. Examples to adapt include:

• Operating or setting up a complicated machine for one job, including the checks that show it is ready. Haas machine-setup demonstration.
• Reloading a label printer: supplies, loading, and checking the printed result. Zebra demonstration.
• Assembling a storage unit, product or kit: part identification, orientation, sequence and completion checks. IKEA assembly examples.
• Filing a document: scanning if needed, naming, choosing a location, adding required information, and finding the right copy again. Document-library capabilities.
• Completing a computer routine such as creating an invoice, submitting an expense claim, setting up a customer record or preparing a recurring report. Software-workflow examples.
• Receiving a delivery and then recording quantities or discrepancies in an inventory system, combining physical footage and a screen recording. Receiving-workflow example.

These links offer inspiration, not my organization's procedure or proof that an AI-created guide has been tested. Help me choose one task before planning its demonstration. If I already know what I want to teach, begin there.

Establish the learner and the result

Work out who needs to learn the task, what they already know, and what they should be able to do afterward. Use relevant role descriptions, onboarding documents, previous questions and feedback when accessible to establish learner needs; label assumptions. Identify the person who can demonstrate the task well, whether that is me or a colleague.

Define a useful beginning and end for the task, the necessary tools or materials, and what a successful result looks like. Find any accessible current procedure, checklist, equipment manual or software instructions that govern it. Use those to check terminology, prerequisites, decision points and required steps. Ask about conflicts between the documented procedure and the demonstrated practice instead of silently choosing one. For equipment tasks, preserve the applicable operator and safety requirements in the resulting instructions rather than guessing them from the video.

Recommend an output suited to the learner: a concise illustrated guide, a checklist, a chaptered video, or a useful combination. Agree on the format and level of detail without making me learn production terminology.

Research current tools and plan the demonstration

This is a fast-moving area. Search current official product documentation, pricing and supported workflows before choosing a tool. Do not rely on remembered rankings or treat these examples as a fixed shortlist:

• Gemini, Google AI Studio and TwelveLabs for audiovisual analysis.
• Guidde and DeepHow for instructional content from demonstrations.
• Scribe and Tango for computer workflows; verify their current recording-import and narration features.

Check the AI tools I already use and search for other suitable current options. Distinguish physical camera footage, screen recordings, and software that captures clicks while I work. Verify whether the actual interface I would use processes both the images and the spoken narration. Accepting a file, transcribing its audio, generating a video, and understanding the demonstrated actions are different capabilities.

Compare two or three useful routes briefly. Check supported devices and formats, upload size and duration limits, editing and export options, required subscription tier, and current cost for my likely use. Distinguish normal monthly prices from annual equivalents and temporary offers. Include simple or already available options where they fit. Give primary-source links and the date checked. If browsing is unavailable, identify what remains unverified and help me provide accessible current documentation.

Let me choose a route before account setup, uploads to a new service or paid work. Help with the exact access steps the selected route needs, using tools my organization permits. If another AI must do the media analysis, prepare a complete handoff containing these instructions, our chosen task, learner, format, source references and agreed capture plan so I can continue without reconstructing the conversation.

Make a short capture plan. Specify what must be visible, which details need close-ups, what should be explained aloud, and how to show the finished result. For computer work, help me record my screen and microphone; filming the screen with a phone is also an option. For mixed work, explain how to label and connect the separate recordings. Plan manageable segments for a long or branching task. A brief outline is enough unless a script would help the demonstrator.

Capture and inspect the source

Guide me or the demonstrator to perform the task at a pace someone learning it can follow. Ask for the names of relevant parts, controls, files and fields; reasons for decisions; checks before proceeding; common mistakes; and an explanation of what changes when an exception occurs. Capture ordinary work rather than requiring a polished performance. Preserve explicit corrections if the demonstrator makes a mistake.

Use an existing recording if it contains what we need. Inspect a short section to establish that you can use both the visual actions and the narration. Identify a detail from each, with timestamps where available. If a channel is unavailable or unreadable, help fix that specific gap or move to the chosen capable tool. Offer a transcript-and-stills reconstruction only if I choose that limited alternative, and carry its missing motion or timing information into the guide's review notes.

Review the complete relevant recording. For multiple clips, track coverage and order so steps are not lost between them. Separate observed actions, the demonstrator's explanations, and additional information from the governing sources. Ask for clarification or a short additional shot when a necessary action is hidden, skipped or ambiguous. Do not invent a click, measurement, machine setting or intermediate action to make the guide appear complete.

Turn the demonstration into instructions

Create the agreed how-to yourself using available tools. Give it a clear purpose and intended reader, prerequisites and materials, numbered actions in the demonstrated order, relevant decision points, and checks that tell the learner when to proceed or when the task is complete. Include common mistakes and exceptions where the source material supports them. Use the learner's vocabulary and explain unfamiliar terms.

Extract relevant frames, screenshots or short clips where supported and place them beside the steps they explain. Crop or annotate them when useful while keeping important context. Use real source images for instructional evidence; do not generate an imagined image of an actual control, part or screen. If extraction or editing is unavailable, use accurate timestamp links or references and explain what would be needed to add the media. For a chaptered video, use the selected tool's editing workflow to align each instruction with the right footage and narration. Produce the artifact rather than just a proposed outline.

Preserve the task's actual method. Keep suggested improvements separate unless I explicitly choose to revise the procedure. Check the guide against the full recording and governing documents for missing actions, wrong names or values, ordering errors, unclear references such as “put it there,” and inconsistencies between the pictures and text.

Review, try and finish

Show the draft to me and have the demonstrator check it if that is someone else. Ask focused questions about remaining gaps, then revise the actual guide. Where useful and available, give an independent reviewer the guide, learner brief and source references to look for unsupported or missing steps. Use a separate review pass if independent review is unavailable; an AI review is not evidence that a person can perform the task.

If possible, arrange with me for an appropriate learner to try following the guide under the task's usual training conditions. Help identify what to observe: where they hesitate, ask a question, miss a step or get an unexpected result. Incorporate their feedback and recheck the affected instructions. If no learner is available, finish with the demonstrator-reviewed guide and clearly identify that it has not yet had a learner trial; do not claim successful use.

Deliver the finished guide in the agreed shareable format with its supporting images or clips, or working links to them. Check that referenced media is included or accessible and label the task, relevant equipment/software and review date so the guide can be updated later. Give me a brief account of what was checked and any unresolved steps that affect its use. Help me save it in the intended location; sending it to other people requires my instruction.

Finish when I have a usable, reviewed how-to grounded in the demonstration and know what, if anything, still needs a real-world check. If missing footage, access or review prevents a usable guide, preserve the work and give the exact next step to resume. A capture plan alone is not the completed activity.

These instructions were written in September 2026 using the frontier models available at the time. You may now have better tools or models, or restrictions that make some tasks harder. Your AI should adapt to the tools and access you actually have and find the best available way to help. Everything here is a starting suggestion for you and your AI, not a proven best practice for every setup. Feel free to rewrite any of it to better meet your needs.