Case Study Sefa Aydın 6 min read

Case Study: A Form That Talks With AI Instead of a Survey Before Training

Case Study: A Form That Talks With AI Instead of a Survey Before Training

Short answer

In this case a form that talks with Claude replaced the survey sent before a training. The participant enters only a name, email address and department. The AI asks 10 to 12 questions and offers answer buttons for each. When the chat ends, the answers become structured fields in the trainer's panel. Anyone can switch to a classic form.

Before I walk into a corporate training, I want to know the participants. Who uses which tool, who has never tried one, and what the team expects from the day? To find out, I replaced the classic survey with a form that talks with Claude. In November 2026 I used it in the trainings of a foundation and a textile manufacturer. In this article I explain how the form works, how it is protected and what the first usage numbers show.

What was the problem?

A classic survey before a training has two weak spots. First, everyone gets the same questions. Asking someone who uses AI every day and someone who has never opened it the same question often means getting half an answer from both.

Second, open questions stay empty. In the box that asks "What do you expect from this training?" most people write a word or two, or nothing at all. Yet the information that shapes the training sits exactly in that box. Participants see it the same way: a long survey looks like a chore to finish between tasks.

The third problem is on my side. Reading a survey full of free text and turning dozens of answers into something comparable takes real time out of training preparation.

What I needed was a form that takes the writing load off participants, moves like a conversation and leaves me with clean data at the end.

Why do I want this information in advance? A training day is limited. Spending half an hour in the room to gauge the level means taking that half hour away from the content. If I know the level beforehand, I start in the right place from the first minute.

What system did I build?

The form works in three stages:

  1. Short intro: The participant enters only a name, an email address and a department. Nothing else is asked.
  2. Conversation: The AI then asks 10 to 12 questions. It offers answer buttons for each question, so the participant does not have to type. Any question can be skipped. The flow is not a fixed list. It follows the course of the conversation.
  3. Structuring: When the chat ends, the AI turns the whole conversation into structured fields such as tool use, experience and expectations. These fields land in the trainer's panel. The chat transcript is saved too, so I can read the full conversation when I need to.

The answer buttons do two jobs. Participants can answer with one tap on a phone. I can compare answers across people. If the buttons are not enough, the participant can type a sentence of their own.

To me this last step is the most valuable part of the form. Participants talk freely, and at the end I see a tidy record under the same headings for every person. The warmth of a conversation and the order of a survey meet in one place. Since every participant shows up under the same headings in the panel, I can see the overall picture of the group without reading each one: who uses which tool, who has never tried one, and where expectations cluster.

There is also a way out. A participant who does not want to talk with AI can switch to a classic form with four steps on the same page. I do not push anyone into a chat. For some people a short form with buttons is simply more comfortable, and that is a fair choice.

I also tested the form carefully on mobile. I measured it at 360 and 390 pixels wide: no horizontal overflow, and touch targets of at least 44 pixels. That size matches the target size recommended in the accessibility guidelines. An answer button that is hard to tap on a phone makes people drop the conversation halfway.

How did I keep it secure and under control?

Running AI on a public page carries two risks: misuse and costs that get out of hand. I set limits for both from the start.

  • Length and turn limits: Message length and the number of turns in a conversation are capped. The chat cannot drift beyond its purpose.
  • Daily request limit: The form cannot go above a set number of requests per day. Even if someone tries to use it heavily, the cost stays within the limit.
  • Input cleaning: The name and department text a participant enters is cleaned before it reaches the AI. This keeps anyone from slipping instructions to the AI through those fields.
  • Hidden from search engines: The form page is not indexed. Only participants who receive the link can reach it.

There is also the question of what not to ask. Participants are employees of an organization, and their answers are about their own work. That is why the intro step asks for nothing beyond a name, an email address and a department.

The shared idea behind these limits is simple: the form exists for one job, getting to know the participant. However free the conversation looks, its frame is narrow and clear.

What was the result?

On the form page for the foundation's training there were 45 sessions in the last 28 days. The average session lasted about 5.5 minutes. In that period 13 forms were completed. These numbers belong to the foundation's training only. I am not sharing a separate measurement for the textile manufacturer's training.

The numbers are small because the form is not open to everyone, only to that training's participants. So I read them not as a conversion rate but as proof that the form is actually being used. I will track the same two numbers in future trainings: completed forms and session length are the first signs of whether the form feels like a burden.

The real purpose of the form is on the trainer's side: preparing the training around participants' real level and the tools they actually use. A group that has never used AI and a team that uses it every day should not get the same program. Examples, exercises and tool choices should follow from this information.

How can this work in your company?

This setup is not only for training. It works anywhere you need to collect information and a classic form gives you half answers. Screening interviews for hiring, customer needs analysis, internal satisfaction surveys or project kickoffs are good examples. When I build one, I follow four rules:

  1. Write the questions as a conversation, but keep the answer buttons. Buttons take away the writing load and make answers comparable.
  2. Define the fields the chat should produce before you start. Your goal is comparable data, not a pile of free text.
  3. Keep the classic form option. Talking with AI does not suit everyone, and the choice belongs to the participant.
  4. Set limits from day one. Length, turn and daily request limits, input cleaning and a page hidden from search are not things to add later.

To me these four rules matter more than the technical setup. What makes the form good is not the code but well written questions and clearly defined fields.

Where to start? Try it on a single use first: one training, one hiring round or one customer group. Read the fields and transcripts from that first run, fix the questions, then roll it out.

I explain the wider frame of needs analysis before a training in AI training needs analysis.

If you want a training planned around your team's real level, take a look at my corporate AI training page.

Frequently Asked Questions

How does a form that talks with AI work before a training?

The participant enters a name, an email address and a department. The AI then asks 10 to 12 questions and offers answer buttons for each. When the chat ends, the answers become structured fields in the trainer's panel.

What if a participant does not want to talk with AI?

They can switch to a classic form with four steps on the same page. They can also skip any question they do not want to answer in the chat.

How is the form protected against misuse?

Message length, the number of turns and daily requests are capped. The name and department text is cleaned before it reaches the AI, and the form page is hidden from search engines.

How much was the form used?

On the form page for the foundation's training there were 45 sessions in the last 28 days, the average session lasted about 5.5 minutes and 13 forms were completed.

Can this setup be used outside training?

Yes. It can be adapted to any information gathering task where a classic form gives half answers, such as hiring screening, customer needs analysis or internal satisfaction surveys.

Want to learn to do this with AI?

In individual training sessions, live on your own project, and in corporate training for teams, I teach step by step how to do everything covered in this article with AI. Write in and I will get back to you the same day.

Message on WhatsApp See individual training Corporate AI training for your team
Sefa Aydın

Sefa Aydın · AI Trainer & Consultant

An AI trainer and consultant who has worked on the Turkey projects of world-famous luxury brands. He teaches, hands-on, how every kind of work is done with AI: design, video, branding and vibe coding.

About Sefa Aydın →   LinkedIn ↗

Related articles

Want to apply this to your own work?

Tell me what you are working on. I reply the same day with next steps.

Your details are used only to reply to you. See the privacy notice.

Message on WhatsApp