When Experience Becomes a System: A Small Law Firm's Agent-Built Intake

At a small Australian law firm, four to five lawyers cover a broad mix of practice areas while one receptionist handles every new enquiry.

For years, that receptionist carried the firm’s front door in her head.

She knew what to ask when a message sounded like a family matter. She knew that a criminal enquiry might need immediate attention. She knew which lawyer handled which kind of case, which details were useful, and which missing answer would send the conversation around again.

The firm had a website. It listed phone numbers, practice areas, and a general contact form. Clients could find the firm there.

But the website did not contain the knowledge required to receive them.

That knowledge lived across old Excel sheets, lawyer emails, intake habits, and one person’s memory. For a practice of four to five lawyers covering tax, employment, criminal, family, corporate, data protection, insolvency and restructuring, and other legal work, that was too much invisible infrastructure for one receptionist to carry alone.

This is the story of how the firm used ChatGPT and the FormHug API to turn that experience into a system—without asking the agent to replace the professional judgment at its centre.

The firm asked not to be named. Identifying details and original form titles have been generalized; the AI-generated header image does not depict the customer.

A Website Could Describe the Firm, but Not Receive a Matter

New enquiries arrived through phone calls, WhatsApp, Facebook Messenger, and email.

Whatever the channel, the receptionist had to reconstruct the same basic picture: who was contacting the firm, whether they were an individual or a business, what kind of matter they had, what had already happened, whether anything was urgent, and which lawyer should see it.

The questions changed with the area of law.

A family matter might require information about relationships, children, and property. A data protection matter might involve a possible reporting deadline. A criminal enquiry might first need to establish whether someone was in custody.

This was not simply administrative work. It was routing knowledge shaped by the firm’s experience. Yet because that knowledge had never become a shared workflow, the quality of intake depended on what one person remembered in the moment.

Sometimes the information reaching a lawyer was complete. Sometimes the lawyer had to ask part of the first interview again. The problem was not a lack of care. It was that the process had no memory of its own.

The System Already Existed—Just Not in One Place

The firm’s principal did not ask ChatGPT to invent a legal intake system from a blank prompt.

He brought the material the practice already trusted:

• its service areas;
• its requirements for distinguishing individuals from businesses and identifying urgent matters;
• the information needed to begin a conflict check;
• and Excel sheets accumulated for different kinds of work.

Those spreadsheets mattered because they were not merely old files. They contained traces of how the firm already thought: the questions clients were asked, the facts lawyers expected to see, and the distinctions that changed the next step.

The knowledge was fragmented, but it was not absent.

Using ChatGPT as the agent and connecting it to FormHug through an API workflow, the principal asked the agent to translate that operating context into structured forms.

We think this is an important distinction. The strongest agent workflows often do not begin by asking AI to invent expertise. They begin by giving existing expertise a structure it can move through.

The Agent Became an Experience Translator

The result was not one enormous questionnaire. It was a three-layer intake system that could reveal more detail as the matter became clearer.

A shared front desk

One general registration form works for both individual and business clients. It gathers contact details, client type, whether the person has worked with the firm before, and a short description of what they need.

It creates a common beginning. Every enquiry can become a record before it becomes a specialist conversation.

Fifteen paths into different areas of law

The agent generated 15 specialist forms covering the firm’s practice areas, including tax, employment, criminal, family, corporate, data protection, and insolvency and restructuring matters.

Each form carries questions relevant to its field. A family-law intake can collect relationship, child, and property context. A criminal-law intake can surface whether custody or another urgent event is involved. A data-protection intake can ask about an incident and a known reporting timeline.

The questions do not decide the case. They make important context less likely to remain trapped in a message thread or omitted from a hurried call.

More detail only when it is useful

The firm also created narrower forms for matter screening, engagement preferences, urgency, and fee preferences such as hourly billing or a request for a fixed quote.

Not every client completes every form. The receptionist guides each person toward the relevant path and asks for deeper information only when the matter requires it.

This layered structure is what separates an intake workflow from a long contact form. It moves from who are you? to what kind of help do you need? to what does the firm need in order to decide the next step?

The workflow is a more complete version of the client intake forms people can build in ChatGPT. ChatGPT shapes the structure, FormHug gives that structure a real public surface, and people outside the conversation can enter the process.

The Receptionist No Longer Has to Recreate the Firm Every Time

When an enquiry arrives, the receptionist can send the appropriate form instead of conducting the entire first interview from memory. The client responds in a consistent structure. The lawyer receives a clearer starting record.

The change is practical:

• the same type of matter begins with the same core questions;
• time-sensitive details have a defined place to appear;
• lawyers spend less of the first review reconstructing basic background;
• new staff can learn the firm’s intake expectations from a working system;
• and lawyers can improve the process when they discover a missing or unnecessary question.

The firm has not provided a measured reduction in intake time, information gaps, or response time, so we are not attaching a percentage to the result.

The deeper change does not need one. Calls, messages, spreadsheets, and receptionist notes no longer have to remain separate pieces of institutional memory. They can enter a repeatable workflow the whole team can see and improve.

This is what a form layer gives an AI-agent workflow: somewhere for human information to become structured, and somewhere for the agent’s work to meet the people the workflow exists to serve.

Structure Comes Before Judgment. It Does Not Replace It.

In legal intake, automation needs a clear stopping point.

The forms can collect names and relationships needed for a preliminary conflict check. They do not complete that check by themselves. They can surface a potentially urgent situation. They do not provide legal advice or decide the firm’s response.

A submission does not by itself create a solicitor-client relationship. The firm remains responsible for engagement, confidentiality, data handling, access to sensitive information, and every professional decision that follows.

Those boundaries are not a limitation hidden at the edge of the story. They are part of good workflow design.

The purpose of the agent is not to make the judgment disappear. It is to help human judgment begin with better structure.

From One Person’s Memory to the Firm’s Memory

It would be easy to measure this project by counting what the agent produced: one general registration form, 15 practice-specific forms, and several more focused workflows.

But the most valuable output is not a form count.

Before this work, the firm’s intake knowledge belonged to documents and individuals. Now it can belong to the practice. A receptionist can follow it. A lawyer can revise it. A new colleague can learn from it. The next enquiry does not have to begin from a blank page.

At FormHug, we often think about agents as builders of structure between a conversation and the real world. In this case, the agent played an even more human role.

It became an experience translator.

It took knowledge the firm had earned over years and helped turn it into something the whole team could carry.