Forms - Extract Data (the AI form builder)

Summary PaT gets data from two places: from the apps you already use, through APIs, and from people and manual processes, through SD Forms. This module explains why that data matters, then shows PaT's AI form builder: seeing a form's JSON in the SD Form Builder, having the Form Assistant draft a form from a description, asking PaT Chat AI what a form should include, and starting a form planning conversation to have the form built. It ends with a more advanced example, a ticketing system built from a multi-stage form.

Steps

Introduction. There are only two ways to get the data you need from your organisation, customers and staff to drive profitable change and gain a competitive advantage. Where the data already sits in an existing app, PaT extracts it for you using APIs. The rest sits in the minds of your stakeholders and in manual processes you haven't yet automated or systemised. Capturing it takes surveys, forms, inspections and audits, which take so long to manage that people don't do them. PaT's SD Forms make it easy to capture this data and turn it into insights you can act on.

Note. The API Endpoint Manager connects any public or private API to PaT, so you can merge your 1st, 2nd and 3rd party data into dashboards; the Endpoint Manager: finding endpoints and building your own UDF module covers it. Combining API data with data captured by PaT SD Forms on a single page is what PaT calls borderless data.

Note. If you can't monitor it, you can't manage it, and the answer is always in the data. Yet much of an organisation's data is too hard to capture: paper forms take ages to build, answers are handwritten and then typed into spreadsheets, and someone has to collate them before they make sense. No one has time for that, so it never gets done. With SD Forms powered by PaT AI, design and analysis work that took hours or days takes minutes.

Note. PaT forms have all the usual elements, such as dropdowns, text entry and file upload. They also go beyond tools like Google Forms: they adapt in real time to earlier answers, so they capture smarter, more relevant data.

The SD Form Builder

Step 1

Every SD includes a built-in SD Form Builder, which builds form elements from a text-based code called JSON. To see a form's JSON, open it in the SD Form Builder and click Schema, at the top right between Settings and Reorder. The Schema Editor opens on the right, showing the form in JSON format.

Note. JSON can look scary if you don't write code, but it is just text, and AI tools like ChatGPT work very well with text. PaT's key parts, such as SDs, Tasks and Forms, are built with JSON. Because AI can write JSON from a normal, conversational request, you can describe what you want and let the AI build it. PaT also plans to let you create tasks, forms and whole projects in SDs by speaking normally, and to find chats, tasks, notes, documents, forms and SDs just by asking: an AI agent searches PaT's data, returns the results as JSON, and PaT shows them to you, for example as a list of chats, in about 3-5 seconds.

Note. PaT's SD Forms Respond Agent uses a kind of AI agent that ChatGPT released in February 2025. You can talk to these agents like a person: they remember what you've said and can suggest what to do next. PaT gave its agent a PaT Form Tool and taught it to understand what a user needs from a form, match that to PaT's technology, write the JSON, and then use the Form Tool to build the form.

Step 2

To build a form by hand, use the Controls tab at the left of the SD Form Builder. Its library has dozens of form elements, such as 5-star ratings, yes/no buttons, data grids and check boxes. Drag the ones you need onto the Canvas on the right.

Step 3

To build a form with AI instead, start from a blank form (its Canvas reads Start Building Your Form) and click the Form Assistant tab, next to Controls. PaT released its AI-powered forms in February 2025. For now, the Form Assistant only works when you create a form from scratch; support for editing existing forms with AI was announced as coming soon.

Step 4

Under Option 1: Describe your form, type what you want in plain words, such as "I want to create a customer survey" or "help me build a job application form with a resume upload", then click Generate Form. PaT does the heavy lifting: it creates, configures and lays out the form, saving anywhere from tens of minutes to hours of work. Then adjust the parts you need, now that you can see a full draft.

Planning a form with PaT Chat AI

Step 5

Before building a form, decide what it needs to capture. For example, an absenteeism form might log who didn't turn up today, track sick days by site, and feed a dashboard in real time so you can see how each site is tracking. PaT Chat AI can suggest what the form should include. To ask it, click Back to SD at the top left of the SD Form Builder.

Step 6

In the SD, click SD Chat in the row of grey buttons above the SD's rows. The SD Chat panel opens.

Step 7

Click AI at the top right of the message box, to the right of GIF. PaT Chat panels have supported Prompt AI since January 2025.

Step 8

The PaT Chat AI panel slides in on the right, in place of the human chat. You can ask a question straight away, or first build a prompt that tells the AI how to behave. This is called Prompt Engineering, and it is all about giving the AI the right setup. To do it, click the cogs icon below the type your question here .... box. The Additional system prompt box opens. Use the framework "You are a [role] that [purpose, task, goal]."

Step 9

In Additional system prompt, type the role, for example "You are a People Operations Specialist designing a form for frontline supervisors to record staff absenteeism".

Step 10

Then give the AI its task in the type your question here .... box, for example "Your goal is to collect clear, actionable data in under 2 minutes to help track patterns, ensure fairness, and protect the business from potential leave abuse".

Step 11

Click Ask, at the right of the question box.

Step 12

PaT Chat AI answers below your prompt with a list of things the form should include. This saves the tens of minutes, if not hours, you would otherwise spend talking to specialists, researching articles and thinking through different scenarios. Adjust the list to suit, for example by leaving out anything you don't need to capture, then take it to the SD Form Builder to have the form built.

Step 13

In the SD Form Builder, on the Form Assistant tab, click Start Form Planning Conversation, under Option 2: Have a conversation to build your form. The AI assistant works through the form with you in a conversation, step by step, and builds it.

Step 14

When someone opens the form's shareable link, they see the finished form. You can add a logo, brand colours and fonts. The data people submit is stored automatically, ready to be pulled into dashboards so you can track it in real time and analyse trends; other guides cover the dashboards.

What SD Forms make possible

Note. The decisions that help a business adapt are often hiding in data that nobody captures because it is too hard to collect. That is the big problem SD Forms solve.

Tip. The Form Assistant is live in PaT, so try it on the next form you need: work that used to take hours now takes a few minutes.

Note. SD Forms can handle much bigger jobs. One more advanced example is an enterprise-level ticketing system built from a multi-stage form, with workflows that change the fields based on earlier answers. When the form is submitted, PaT automatically creates a ticket in the CS queue and triggers connected actions across the task engine, PaT Chat, dashboards and SDs. The form section of a build like this can be completed in a couple of hours, with no coding. With dashboards added, it replaces a standalone ticketing system, saving the client $150K a year in licensing fees.

Note. SD Forms powered by PaT AI suit any organisation that values what its people and customers know, and the data buried in manual processes that haven't yet been systemised or automated. You can quickly deploy smart digital forms to capture that data. Work that used to take hours or days, and often never got done because everyone was too busy, now takes minutes.