Google Gemini Complete #04 — How to Give Gemini a Useful Task

AIYING | GEMINI COMPLETE | #04

How to Give Gemini a Useful Task

Understand the platform. Discover possibilities. Accomplish useful work.

By Ravindra Dastikop

“Create an AI workshop.”

Gemini may respond with a polished agenda. But is it for schoolchildren or engineers? Does it last an hour or a day? Can participants use computers? What should they be able to do afterward?

A fluent answer can leave those questions unresolved. A useful task brief makes the important requirements visible before the work begins.

In this episode

Turn a vague request into a clear brief, compare the resulting drafts and create a reusable task card.

1. Give Gemini work to complete

A topic names an area. A task specifies an action and a result.

“AI workshops” is a topic. “Draft a 60-minute agenda for first-year students” is a task. Add the learning goal, constraints and expected structure, and you have a brief you can review.

Google's Workspace guidance highlights role, task, context and format as useful prompt components. Its developer guidance also recommends clear instructions and iterative refinement. [1][3]

You do not need elaborate wording. Include the information that changes what a successful response should contain.

2. A practical task card

The following six-part card is an AIYING teaching framework, not an official Google formula.

Part	Question	Workshop example
Outcome	What should the work achieve?	Students practise framing a task and checking AI output.
Audience	Who will use it?	30 first-year college students.
Inputs	What information should it use?	The approved workshop notes.
Constraints	What boundaries must it respect?	60 minutes; no assumed student devices or internet.
Output	What should it produce?	An agenda table and activity instructions.
Checks	How will you judge it?	Correct timing, practical activity and visible unknowns.

3. Separate facts, suggestions and unknowns

We continue the fictional college workshop used in earlier episodes.

The duration and audience are confirmed within the exercise. The venue, date and device availability are unknown. An activity design is a suggestion.

Keep those categories distinct. A proposed computer exercise must not quietly become a claim that a computer lab has been booked.

Add a simple instruction: “Preserve the supplied facts, label suggestions and list unconfirmed arrangements.”

This instruction establishes a requirement. You still need to check whether Gemini follows it.

4. Compare a vague request with a usable brief

Starting request:

Create an AI workshop.

Clearer request:

Draft a 60-minute introductory AI workshop for 30 first-year college students. Students should practise writing a clear AI task and checking its output. Include one small-group activity that works without student devices. Present an agenda table, activity instructions and unconfirmed arrangements.

The second request gives you concrete criteria for review. It does not guarantee a better result on every run, and this example is not a measured product comparison.

5. Reusable prompt: from brief to reviewed draft

Help me prepare a draft for human review.

OUTCOME
Students practise writing a clear AI task and checking its output.

AUDIENCE
30 first-year college students.

CONFIRMED INPUTS
- Introductory AI workshop.
- Total duration: exactly 60 minutes.
- Include one small-group activity.

CONSTRAINTS AND UNKNOWNS
- Do not assume student devices or internet access.
- Date, venue and budget are unconfirmed.
- Do not invent approvals, bookings or purchases.
- Label proposed arrangements as suggestions.

OUTPUT
1. A one-sentence learning objective.
2. An agenda table: minutes, activity, student output.
3. Instructions for a device-free small-group activity.
4. One observable check of student learning.
5. A list of arrangements still requiring confirmation.

SUCCESS CHECKS
- All agenda durations total exactly 60 minutes.
- Students practise both task framing and output checking.
- The activity is feasible for 30 students.
- Confirmed facts, suggestions and unknowns stay distinct.

If a missing detail prevents a useful draft, ask up to
three focused questions. Otherwise draft with unknowns marked.

Why include a rule for missing information? It gives Gemini a way to proceed without silently filling every gap. You can request questions when necessary while still allowing a useful provisional draft.

6. Review, then request a specific revision

Google recommends natural language, relevant context and breaking complex work into separate prompts. [2] For this exercise, first create the agenda; then inspect and revise it.

A focused follow-up is easier to assess than “Make it better.”

The activity assumes every student has a device.
Revise it for six groups of five using paper only.
Keep the agenda at exactly 60 minutes.
Show the revised activity and any timing changes.
Preserve the confirmed facts and list remaining unknowns.

Use this follow-up only if it fits the problem you observe. For another failure, name that failure and the required correction.

If you ask Gemini to review its own draft, treat the review as assistance. It is not independent evidence that the draft is correct.

7. Try it yourself: a 15-minute experiment

Tool: Gemini web app.
Prerequisites: an eligible account with access to a text conversation and somewhere to record your observations. No API key is required.

1. Write your checklist first. Use the four success checks in the reusable prompt.
2. Run the starting request. In a new conversation, submit “Create an AI workshop.” Keep the response.
3. Run the detailed brief. Use a separate new conversation. Keep the same visible model and settings where possible; record automatic selection if used.
4. Compare against the checklist. Record met, partly met or not met, with a short supporting observation.
5. Revise one weakness. Give a specific follow-up based on what you found.
6. Save the checked draft and task card. Note what you corrected yourself.

This comparison tests two requests with different amounts of task information. It does not isolate wording alone or establish a general model ranking. Even if the vague request succeeds, record that result honestly.

8. Verification checklist

• Add the durations yourself: do they total 60 minutes?
• Does the activity practise both task framing and output checking?
• Can six groups of five complete it with the proposed materials?
• Are devices, internet, venue and budget handled as specified?
• Are suggestions and unknowns clearly labelled?
• Does the learning check show something students can actually demonstrate?
• Did the revision fix the problem without creating another?

9. What have you accomplished?

You should now have two drafts, an evidence-based comparison, one reviewed workshop plan and a reusable task card.

For your next real task, replace the workshop details with your own outcome, audience, inputs, constraints, output and checks.

A useful extension is to write the card for a study guide, proposal summary or meeting follow-up before asking Gemini to produce it.

AIYING takeaway

Define what useful work looks like before asking Gemini to do it.

A clear brief gives the assistant direction and gives you a basis for judgment.

Continue Google Gemini Complete

Previous: #03 — Models, Plans and Limits: Understanding Your Choices.

Next: #05 — Context, Personalization and Conversation History.

Official sources and fact-check note

Documentation checked: 15 November 2026.

1. Google Docs Editors Help: Learn about prompts for Gemini — role, task, context, format and refinement.
2. Google Workspace Learning Center: Tips to write prompts for Gemini — clear language, context and separating complex tasks.
3. Google: Prompt design strategies — explicit instructions and iterative prompt design.

These references address different Gemini environments. This episode applies general prompting principles in a text conversation; it does not assume developer controls or Workspace integrations exist in every account.

The six-part task card, prompts, fictional scenario and exercise are AIYING teaching material. No hands-on product test or measured improvement is claimed.

AIYING | Choose work that Matters.
Google Gemini Complete • Ravindra Dastikop
Independent educational series. Not affiliated with or endorsed by Google.