Form Design Workflows using AI

Support
1

1. What is the issue? Please be detailed.
Ideally, I'm wanting to be able to give AI enough context to help quickly develop forms in Google Sheets. I've only seen this with regards to others looking for better workflows for developing XLSforms and the brief summary is that @Brenton_Peterson is looking for:

  1. Plain text format
  2. Ability to edit simultaneously (and Version Control)
  3. Identifying Errors Proactively (Form Validation)
  4. Clean and Efficient Commenting

I've moved to Google Sheets exclusively from Excel because of many of these issues. In addition, Google Apps Scripts allow you to do a lot of things like custom functions which could help with autocomplete. I've also used it to more quickly iterate/upload form definitions and media to my self-hosted ODK Central server and ONA servers.

However, given changes in AI in the last few years, I'm wondering if someone has started using AI when addressing a complex issue. I've found that its difficult to give AI context in a normal chat prompt. So my question is for people who are developers or who are familiar with AI tools that are using CLI in IDEs like VS Code. In these setups, users can use their local files as context for the AI, and I'm wondering if anyone has been able to do this well using Google Sheets or an XLSX file to be able efficiently identify problems, fix things, etc., ideally where you could address specific cells/rows and AI would give suggestions. My attempts with Gemini in the side window of Google Sheets has resulted in really poor quality responses, whereas chat prompts with Kappa AI (on the ODK Docs page), Gemini, and ChatGPT have been better quality.

Has anyone ever seen an interface for Excel or Google Sheets for AI suggesting changes similar to code diff views like this?

vscode-agent-mode-diff-view
vscode-agent-mode-diff-view2846×1726 687 KB

2

Hello @Tyler_Depke

I’ve been experimenting with an agentic workflow for rapid XLSForm development and wanted to share the stack that’s working for me.

Currently, I’m using the Hermes agent via Telegram for on-the-go initial drafting, paired with OpenCode for long-running, complex modifications. To maintain strict validation, I wrap these agents around core ODK XLSForm Skills, the PyXform CLI, and a PyXComparer tool.

The results for structural heavy-lifting have been highly satisfactory. However, while the agents excel at syntax and logic automation, they need lot of steering and the final finesse and field-readiness still fundamentally rely on the human touch.

3

Make sure to check out the MCP server that's now exposed to bring the Kapa docs tool into other AI tools of your choice. I've found that it greatly improves what ChatGPT can do. Unfortunately Gemini within Google Sheets doesn't yet make it possible to connect to add tools but hopefully it's coming.

We've been experimenting with some of these kinds of ideas but don't have anything polished enough to share yet.

4

This is a very relevant discussion. I agree that a general chat prompt is often not enough for XLSForm work — the AI needs structured XLSForm context and validation feedback to be genuinely useful.

We have been exploring this in an open-source project here: https://github.com/ARCED-International/xlsform-ai. It focuses on AI-assisted XLSForm creation, editing, import, validation, and translation workflows. I am also building SurveyLoopr, which grew out of similar ODK/XLSForm workflow problems.