Practical AI tools for everyday life and business. Explore what AI can do
Learning Hub
PRACTICAL AI FIELD GUIDE

Gemini & Google tools

Give Gemini concrete examples, then choose the right route for chat, coding or structured data.

Free guide8 min readReviewed

Gemini Apps, Gemini CLI and the Gemini API are separate surfaces. CLI commands below belong in an installed terminal tool, not in a Google Docs document or an ordinary Gemini chat.

Your first useful task

Start with one small win

  1. Choose a small task and attach or paste only the relevant, permitted source material.
  2. Describe the required output and show one example of the format.
  3. Check the result against the source; refine one requirement before reusing the prompt.

Practical tips that make a difference

Show the shape of a good answer

Google’s prompt guide recommends concrete examples. Demonstrate the fields, length or tone you want, using a consistent format. Include a tricky case as well as an easy one, so the model has a pattern for handling missing information.

Based on: Google · Gemini prompt design ↗

Give the input a clear boundary

Label the source material and the task separately. Specify whether the model should summarise, extract or compare. Keep the requested output format consistent across follow-ups instead of combining several unrelated jobs into one prompt.

Based on: Google · Gemini prompt design ↗

Inspect sources when they appear

Gemini Apps can show related links and sources, but not every response includes them. A source link is a place to investigate, not proof of every sentence in the answer. Open it and find the passage that supports your claim.

Based on: Google · Sources in Gemini Apps ↗

Developer notes

Give the CLI a reusable project context

Gemini CLI reads GEMINI.md context files. Put the project’s real test commands and relevant conventions there. Use /memory show to inspect loaded instructions and /memory reload after editing them. These are CLI commands, not a universal Gemini Apps feature.

Based on: Google · Gemini CLI project context ↗

Use a schema when software consumes the answer

For API extraction jobs, configure structured output with a supported JSON Schema instead of only asking for JSON in prose. Then validate required fields and business rules in your own application. Matching a schema does not prove the extracted facts are correct.

Based on: Google · Structured outputs ↗

Check the sandbox before allowing edits

Gemini CLI supports sandboxing; the -s flag enables it with the available platform setup. Review the configured access and tool actions. Sandbox setup varies by operating system and backend, so follow the official prerequisites for your machine.

Based on: Google · Gemini CLI sandboxing ↗

Copy, adapt, then check

Original Utilising AI examples. Replace bracketed placeholders before use. These buttons copy text only; they do not run code or contact an AI provider.

Extract a useful action list

From the notes below, extract actions as a table: action | owner | due date | source line. Use “not stated” for absent values. Example: “Mina to draft menu by Friday” → “Draft menu | Mina | Friday | [line]”. Do not add actions that are not in the notes.

Notes:
[paste permitted notes]

Original business example. Confirm names and dates before assigning work.

Based on: Google · Gemini prompt design ↗

Gemini CLI context checks

/memory show
/memory reload

Run each separately in an installed Gemini CLI session. The first shows loaded context; the second refreshes it.

Based on: Google · Gemini CLI project context ↗

A focused coding request

Inspect [file or function] and explain the cause of [exact failure]. Use our existing library versions. Propose the smallest change and a test using input [example] with expected output [result]. After the fix, run the relevant test and show its actual outcome.

Original coding brief. Use Gemini CLI with project access, or attach the necessary code in chat.

Based on: Google · Gemini prompt design ↗

Check your result

  • Do extracted values match the original material?
  • Are CLI instructions being used in the correct tool?
  • For API data, do both schema validation and real-world checks pass?

Common mistakes to avoid

  • Assuming every Google account has the same connected Workspace features.
  • Using structured JSON as evidence that the data is true.
  • Running a pasted shell command before understanding its effects.

Sources & review notes

Reviewed on 15 November 2026. Tips are paraphrased from the linked documentation; examples and practical recommendations are our own. This is a documentation review, not a hands-on test of every provider account or feature. Interfaces, plans and commands can change.

Independent learning resource by Utilising AI Pty Ltd. No endorsement by the named providers is claimed. Review your chosen tool’s permissions and data handling before supplying business information.

Keep building your AI skills

Create your free assistant prompt Browse all six guides
Gemini & Google tools | UtilisingAI