Can AI Make Online Forms Less Painful? Here's What We're Building.

Online forms haven't changed much in years.

Whether it's customer feedback, student surveys, client onboarding, or recruitment, the process usually looks the same.

Create a form. Add fields. Share a link. Hope people complete it.

But what if people could simply explain what they wanted to say, and software handled the structure?

That's the idea behind Talking Links, a new product under development at Digiwares.

The idea: Let people respond naturally

Imagine a teacher asking students:

"What was the most confusing part of today's lesson?"

Instead of navigating multiple form fields, a student could respond by voice or text:

"I understood the first example, but I got confused when we started working with fractions with different denominators."

That's useful information, but collecting dozens of responses like this creates another problem: organising them.

Talking Links aims to solve both sides.

The respondent gets a simpler way to answer, while the creator receives structured information that is easier to review.

How it works

The workflow is deliberately simple.

1. Create a link

A teacher, business owner, recruiter, or researcher creates a link for collecting responses.

2. Share it

Respondents open the link and answer by voice or text. No account required.

3. Collect structured records

The system processes responses and organises the extracted information in a central dashboard, grouped under the original link.

For example, the student's response might become:

• Topic: Fractions
• Difficulty: Different denominators
• Context: Understood the first example but struggled with the next explanation

The goal isn't to replace what the respondent said. It's to make the useful information easier to identify and work with.

The accompanying GIF illustrates the concept.

The technical challenge

The interesting part isn't creating links or storing responses.

It's turning unpredictable human language into reliable structured data.

People don't communicate like database schemas.

They provide incomplete information, mix different topics, change direction halfway through a sentence, or speak in different languages.

That introduces several engineering challenges:

• Speech recognition: Accurately transcribing responses across different speaking styles and languages.
• Information extraction: Identifying relevant details without inventing missing information.
• Data integrity: Keeping every structured record associated with the correct link and submission.
• Privacy and security: Protecting respondent information and enforcing appropriate access controls.
• Failure handling: Ensuring processing errors don't silently produce misleading records.

These are some of the areas we're focusing on during testing.

Why we're keeping the MVP small

It's tempting to build every feature users might eventually need.

Integrations, automated follow-ups, advanced analytics, templates, and complex workflows are all possibilities.

But we're deliberately keeping the initial version focused on three capabilities:

Create a link → Share it → Collect structured responses.

The priority is making the core experience reliable before expanding the product.

Where things stand

The Talking Links MVP is built and undergoing testing.

We're validating the end-to-end workflow, investigating technical issues, and refining the experience before wider availability.

We've also shared the concept with a small number of people internationally, including in Thailand and Vietnam. Their initial reactions have been encouraging.

But positive reactions aren't the same as real-world product validation.

That's the next challenge.

A question for the DEV community

If you were developing an AI-powered response collection tool, which technical challenge would concern you most?

Speech transcription accuracy? Reliable structured extraction? Data privacy? Or designing a workflow simple enough that people actually use it?

And from a product perspective, could voice and text responses replace traditional forms for certain use cases?

We'd be interested in hearing your thoughts, especially critical feedback.

Talking Links: https://tlk.so

Built by Digiwares. Currently in MVP testing.