How to Create Your First Custom Field with ChatGPT or Claude

Yuvin Kim

12 thg 11, 2026

How to Create Your First Custom Field with ChatGPT or Claude

Yuvin Kim

12 thg 11, 2026

You want respondents to distribute 100 points across three priorities, see the remaining points, and submit the allocation as useful data. Three separate number questions can collect the inputs, but the interactive total and validation need more than ordinary question labels.


With Walla connected to ChatGPT or Claude, you can describe that interaction and ask the assistant to build a Custom Field. This walkthrough takes one small example from a copyable request to a test submission and a check of the saved answers. The point allocations below are illustrative test cases, not results collected from real users.


A Custom Field creates an input experience beyond the built-in field types. Walla’s documented process is to describe it, review the preview, approve the build, and add the resulting field to a form. Read the English Custom Fields guide


1. Connect Walla and define a small first example

Connect Walla to your preferred assistant before asking it to create anything. The English setup guide covers the Walla plugin in ChatGPT and the Walla connector in Claude, including signing in and granting access. Connect Walla to ChatGPT or Claude


Use a workspace you can access and a separate test form. Ask the assistant to identify the intended workspace before creating the field if your account has more than one. You do not need a real customer list for this exercise.


What the example should do

Part

Requirement

Question

How would you allocate 100 points across these priorities?

Inputs

Training, Software, and Support. Whole numbers from 0 to 100, starting at 0.

Interaction

Show each value, the total, and the remaining points. Keep the inputs usable on mobile and with a keyboard.

Validation

Allow a zero allocation to an individual category. Require a combined total of exactly 100 before a valid submission.

Saved answer

Keep three numeric allocations and the numeric total, rather than saving only the displayed total.

This is a design specification for the field you are asking the assistant to generate. The preview and saved response checks below are how you verify that the implementation matches it.


2. Paste this request into ChatGPT or Claude

Replace [workspace name or ID]. The same request can be used with either assistant after connecting Walla.


Use the connected Walla tools to build a reusable Custom Field named “Priority Allocation” in [workspace name or ID]. If the workspace is ambiguous, ask me to select it before creating anything.


The question is “How would you allocate 100 points across these priorities?” Build three labeled number inputs for Training, Software, and Support. Accept whole numbers from 0 to 100. Start all three at 0. Show an updating total and the number of points remaining.


Use these saved answer properties: training, software, support, and total. All four must be numeric. The total is the sum of the three allocations. Keep a legitimate zero as 0, not an empty or missing answer. Do not collect names, email addresses, or other identifying data in this field.


Validate the inputs and require a total of exactly 100 for a valid answer. Show a clear error if the total is below or above 100, or an input is negative, fractional, missing, or above 100. Use Walla’s supported Custom Field validation and value-saving contract so a visual total alone is not treated as a valid saved answer. If any requirement cannot be implemented, tell me what is missing.


Make the labels readable on mobile, provide visible keyboard focus, and include a Reset action that clears the current allocation back to the initial invalid state. Explain what Reset does before a respondent submits.


Build the field and give me its review preview. Summarize the stored answer shape and validation behavior. Stop for my approval before registering the build. Do not attach it to an existing production form or publish any form.


A response containing a specification or code is not the same as a registered field. Check the build status and open the preview the assistant provides. If the build fails, ask it to fix the reported error and produce an updated preview.


3. Test the preview and revise it through conversation

Try the interactions yourself before approving the build. A preview lets you inspect the field’s behavior; it does not prove that a real form submission has been stored.


Test input

Expected behavior

Training 40, Software 35, Support 25

Total 100 and remaining 0; allocation is valid.

Training 100, Software 0, Support 0

Valid allocation; both zeros are preserved.

Training 40, Software 35, Support 20

Total 95; show that 5 more points are needed.

Training 60, Software 35, Support 25

Total 120; explain that the allocation exceeds 100.

Negative value, fraction, or a blank input

Reject or explain the invalid input; do not silently save a valid answer.

Reset after a valid allocation

Return to the initial state; the old valid answer must not remain available for submission.

A follow-up request you can copy

Revise the Priority Allocation build. Keep the saved answer properties training, software, support, and total unchanged. Make the remaining-points message easier to notice, show the validation message beside the inputs, and keep all labels visible on a narrow phone screen.


Check that changing an allocation updates the saved value as well as the visible total. Reset must clear the previous valid answer. Show me the revised preview and explain what changed. Wait for approval again; do not publish a form.


Describe the behavior you observed when requesting a fix. For example, “The total changes on screen, but Reset leaves the old allocation in place” is more useful than “Make it work.”


4. Approve the field and add it to a test form

Once the preview meets your requirements, explicitly approve the field build. Registering the reusable field and publishing a form are separate steps. Keep the form as a draft while arranging the test.


I approve the reviewed Priority Allocation build. Register this version, then create a new draft form named “Custom Field Storage Test” in [workspace name or ID]. Add a required short-text question named “Test case ID” and place the Priority Allocation Custom Field after it as a required question.


Use the registered field we just reviewed rather than building another copy. Show me the form ID, field placement, and current publication status. Do not publish the form yet. If attaching the field cannot be completed with the available tools, tell me the exact remaining step.


Walla also provides an editor route: when adding a field, open the Custom Fields tab and select the one you built. This is a fallback if you need to finish the placement in the editor. You can continue using conversation for the supported creation and editing steps.


Before testing stored responses, check the final form has both questions and the required settings. A successful field build alone does not mean it has been added to this form.


5. Submit test answers through the actual form

To check storage, use a separate published test form with fictional data. Publishing here is a deliberate test step, not something the initial build request authorizes.


Publish only the form “Custom Field Storage Test” with ID [test form ID] so I can make test submissions. Return its respondent link. Do not publish or modify any other form.


  1. Open the respondent link and enter CF-001 as the Test case ID. Allocate Training 40, Software 35, and Support 25. Submit the full form and check the confirmation screen.

  2. Make a second submission labeled CF-002 with Training 100, Software 0, and Support 0.

  3. Try a total of 95 and then a total of 120. Check that the invalid allocation cannot be submitted as a valid answer.

  4. Return to a valid allocation, reset it, and try to submit again. Verify that an old valid value cannot bypass the rules.

  5. Repeat the essential checks on a phone, and test returning to the field and changing an answer before submission.

If the full form accepts an invalid allocation even though the preview showed an error, stop and correct the field or its form placement. A visible warning is not enough if it does not enforce the rule during submission.


6. Check what Walla actually saved

Open the test form’s Response Sheet. Find CF-001 and CF-002, then inspect each response in detail. The sheet supports search and response details, and it can export CSV or XLSX. Read the English Response Sheet guide


Expected values for the two valid tests

Saved property

CF-001

CF-002

training

40

100

software

35

0

support

25

0

total

100

100

These are the values requested by this example. Verify them in the stored response; do not assume the saved property names or types match merely because the preview looked correct. In particular, CF-002 checks that a legitimate zero was not converted into a missing answer.


Check that the stored allocations match the final values entered, rather than the first values shown. Confirm that the answer includes the category values and not just a screenshot, display label, or total. Use the response ID and Test case ID together when reporting a discrepancy.


Ask the assistant to help inspect the test responses

In the form with ID [test form ID], find the test responses labeled CF-001 and CF-002. Inspect the saved Priority Allocation values, not the field’s default properties or its preview state.


Compare them with these expected allocations: CF-001 is training 40, software 35, support 25, total 100. CF-002 is training 100, software 0, support 0, total 100. Report the response identifiers, which properties are present, and any mismatch or type issue visible in the returned data. Do not edit or delete responses.


If the available tools cannot expose the relevant saved values, say so and tell me what to check in the response details or export. Do not report storage as verified from the preview alone.


An export provides another check of how the answer is represented outside Walla. Open the downloaded file and locate the two test records. A structured Custom Field answer is not necessarily flattened into one spreadsheet column per property; inspect the actual representation before building reporting formulas around it.


7. Use this acceptance checklist before collecting real responses

Check

Pass condition

Placement

The intended registered field is present in the correct form.

Interaction

Inputs, total, remaining points, and Reset behave as specified.

Validation

Invalid or reset allocations cannot be submitted as valid answers.

Storage

Both test records match the expected properties and values, including zeros.

Changes

The final edited values are saved, without stale values from earlier interaction.

Export

The team understands how the answer appears in the actual downloaded file.

Usability

The field is readable and operable on the devices and input methods your audience uses.

After changing how a field saves answers, repeat the storage checks before using the new version for live collection. Keep the test records identifiable and separate from the data used for operational reports.


You now have a repeatable way to describe an interaction, refine it in conversation, and verify the response it produces. Start with one field and a small set of explicit tests, then expand the same approach to the next requirement.


For a broader introduction, see Meet the New Walla: Build Smarter Forms with ChatGPT, Claude, and Custom Fields. When you are ready to build, use the English Custom Fields guide and Walla connection instructions alongside your test checklist.


Open Walla