Build a Student Progress Coach With Make and Gemini

Build a student progress coach that collects a learner’s updates, asks Gemini to turn them into useful feedback, and sends a supportive next-step message. You’ll connect a form or spreadsheet to an automation scenario in Make, then add Gemini to the workflow; review privacy and output quality before using it with students.

What You’ll Need

• A Make account with permission to create scenarios and connect apps.
• Access to Gemini through a connection or API option available in your Make account. Connection names and available modules can change, so follow the current prompts in Make.
• A source for student updates, such as a form that writes responses to a spreadsheet. Use tools approved by your school or organization.
• A destination for coach messages, such as a draft email, a spreadsheet column, or a learning platform you are authorized to use.
• A small set of fictional test responses. Avoid entering identifiable student information until you have approval and understand your organization’s data rules.

Step-by-Step

1. Decide what the coach should do. Keep its purpose narrow: summarize progress, identify one obstacle, and suggest one achievable next action. Decide who receives the result and whether a teacher must review it. A coach should support instruction, not assign grades or make high-impact decisions automatically.
2. Design a brief progress check-in. Create a form with questions such as “What did you work on?”, “What went well?”, “What is still difficult?”, and “What will you try next?” Include a course or topic field if needed. Ask for only information relevant to the feedback. Explain to students how responses will be used, according to your school’s process.
3. Send responses to a spreadsheet. Connect the form to a spreadsheet and confirm that each submission creates a new row with clearly named columns. Use a test response to check that fields arrive as expected. If your setup uses a different data source, make sure it can trigger or be checked by a Make scenario.
4. Create a scenario in Make. Start a new scenario and add a trigger for the response source. Choose the available “watch” or new-submission event for your form or spreadsheet. Connect the account when prompted, select the correct file or form, and configure the trigger to retrieve new entries. Run the scenario once with a test submission to verify the trigger output.
5. Add a filter for usable responses. Before calling Gemini, check that required fields are present and that the entry is not blank or a test row. In Make, add a filter between modules and set conditions using the fields returned by the trigger. This avoids sending empty or unintended data to the model.
6. Add the Gemini step. Add the Gemini module or app connection available in your Make workspace. If the connection requires an API key, create and store it only through the secure connection setup; do not paste credentials into prompts, spreadsheets, or messages. Select an appropriate text-generation action, then map the student’s relevant response fields into its prompt.
7. Write a constrained prompt. For example: “You are a supportive learning coach. Using only the check-in below, write a brief response with: one specific strength, one question or observation about the obstacle, and one small next step. Be encouraging, concrete, and age-appropriate. Do not invent facts, diagnose, grade, or make claims about the student. If the response lacks detail, ask one clarifying question. Check-in: [mapped fields].” Avoid asking the model to infer sensitive traits or compare students.
8. Set practical output limits. Keep the requested response short and structured, such as three labeled lines. Use available generation settings to limit excessive length and randomness, if those controls are exposed. Do not assume a particular setting is present in every Make connection. The prompt is guidance, not a guarantee, so test the actual output.
9. Send the result somewhere safe. For an initial version, write Gemini’s response to a review column in the spreadsheet or create a draft message rather than sending it directly to a student. Map the response text and a non-sensitive reference, such as a submission ID. Avoid placing unnecessary student details in email subjects or logs.
10. Test the entire scenario. Use fictional examples covering a clear success, a vague response, a missing field, and a frustrated or concerning message. In Make, run the scenario and inspect each module’s input and output. Confirm that the right row is matched, the prompt contains only intended fields, and the returned text reaches the review destination.
11. Review and improve before launch. Have an educator check whether responses are accurate, constructive, and suitable for the students. Revise prompts that produce generic advice, unsupported assumptions, or inappropriate tone. Decide who reviews messages, how quickly, and what happens when a response signals a need for human support. Only enable automatic delivery after the responsible team approves the workflow.
12. Turn on the scenario and monitor it. Set the schedule or instant trigger behavior offered by your source app, enable the scenario, and watch its run history during a limited pilot. Check for failed connections, duplicate processing, unexpected output, and changes to form or spreadsheet columns. Keep a manual way to pause the scenario and correct or delete mistaken results.

Common Mistakes

• Sending too much information: A model usually needs the learning update, not a full student record. Remove names and other identifiers unless they are necessary and explicitly approved.
• Expecting Gemini to know course context: It can only use the supplied prompt and available context. Provide a short, approved description of the task or rubric when needed; do not expect it to access classroom materials automatically.
• Letting unreviewed text go straight to students: Model responses can be inaccurate or poorly phrased. Begin with a human approval step.
• Mapping the wrong spreadsheet fields: A changed column order or renamed header can lead to mismatched feedback. Map by field labels where possible and test after edits.
• Confusing a successful run with a good answer: Make can report that a scenario ran even when the feedback is vague or unsuitable. Inspect content as well as technical status.

Tips

• Ask for a single next action that can be completed in a short study session, rather than a long list of goals.
• Use neutral language: “What part would you like to practice?” is more useful than assuming why a learner struggled.
• Keep a teacher-editable field for notes or corrections, and make it clear which text was generated and which was reviewed.
• Use a stable response ID to prevent duplicate messages if a scenario retries after an error.
• Review the workflow periodically, especially after changing prompts, integrations, or the form.

Troubleshooting

The trigger does not find new submissions: Check that the form is connected to the intended spreadsheet or account, that the scenario is enabled, and that the selected trigger watches the correct file. Submit a fresh test entry; some triggers do not process older rows automatically.

Gemini receives blank or incorrect fields: Open the trigger’s output in the run details and verify the exact field names. Re-map the prompt after changing the form or spreadsheet, then test with a new submission.

The Gemini connection or request fails: Recheck the connection setup and any required access, model selection, or usage limits shown in the current interface. Avoid putting a key in the prompt. If the connection appears valid, inspect Make’s error details and retry with a short, simple test prompt.

The response is generic, too long, or makes things up: Narrow the prompt, ask it to use only the provided response, and specify a short structure. Add a fallback instruction for missing information. Keep human review in place rather than relying on prompt wording alone.

Messages are duplicated or arrive at the wrong destination: Check scenario retries, filters, row identifiers, and recipient mapping. Test with a review-only destination, then add a deduplication check based on the submission ID before enabling delivery.

Frequently Asked Questions

Can Make use Gemini without a spreadsheet?
Yes, if your form or another approved app has a suitable Make trigger and provides the response fields. The exact available integrations depend on your workspace.

Does the coach automatically know a student’s grades or course materials?
No. It can use only data explicitly passed into the scenario or otherwise made available through configured integrations.

Should feedback be sent to students automatically?
Not at first. Use a teacher-review step and assess accuracy, tone, and privacy before deciding whether any messages can be delivered without review.

Can I use this with identifiable student information?
Only if your school or organization has approved the tools, data flow, and purpose. Minimize data and follow applicable institutional policies and privacy requirements.

What should the coach do with a concerning response?
Do not rely on generated feedback to assess or handle a crisis. Define a human review and escalation process with your school before launch, and make sure the workflow does not delay that process.

Will the scenario keep running if I change the form?
It may, but renamed or removed fields can break mappings. After any form, sheet, or prompt change, run test submissions and inspect the full scenario before relying on it.

Related

• Reduce AI Workload Costs Without Changing Your Model
• Auto-Delete OTP Messages: How It Works and How to Set It Up
• Google Nano Banana AI Image Tool on X?