How to Describe an AI Automation Task: A Founder's BRD Template

Most automation projects stall at the starting line because founders struggle to explain tasks to technical teams, while technical teams lack business context. With AI, this has changed: you can now describe process logic in natural language, and that is often enough to launch.

Why Your Tasks Weren't Automated Before

Previously, any automation required a programmer who spent months writing code for every manager's request. Today, AI either writes the code itself or acts as the operator. Your job as a founder is not to find the "right button," but to provide the right context.

If you simply say, "Make AI respond to customers," you get chaos. If you say, "Take the incoming email, check product availability in this table, and generate a response using this template," you get a working tool. When you understand how to verify the automation result, the process becomes manageable.

Definition: AI automation is the setup of a system where artificial intelligence performs repetitive intellectual actions (analysis, writing, sorting) without direct human intervention.

Task Description Template (BRD in 5 Minutes)

To get results, fill out these four blocks. This will serve as your technical requirement.

1. Input Data

Where is the information that needs processing located? Example: A CRM customer card, a Google Sheet, a PDF invoice, or a call recording.

2. Business Logic (Process)

What exactly should be done with the data? Describe the steps as if explaining them to an intern on their first day. Example: 1. Listen to the request. 2. Identify the top 3 customer pain points. 3. Check if our price list has solutions for these points.

3. Desired Result (Output)

In what format do you want the response? Example: Telegram message text, a new row in a spreadsheet, or a completed contract template.

4. Guardrails (Constraints)

What is strictly forbidden for the AI to do? Example: Do not promise discounts over 10%, do not answer technical questions about the engine, use only a formal tone.

Checklist: Is Your Task Ready for Automation?

Before handing a task over for development, check it against these points:

• [ ] You perform this action more than 5 times a week.
• [ ] The task does not require physical presence (computer only).
• [ ] You have examples of "perfect execution" (e.g., 10 successful proposals).
• [ ] The task data is digitized (not on paper).
• [ ] You can clearly define a "well-done" criterion.

Comparison: Bad Requirements vs. Good Requirements

If you are concerned about data security when describing tasks this way, it is worth preparing a security policy beforehand so the team knows the boundaries.

Definition: Business logic is a set of clear rules and conditions by which a company makes decisions or processes information.

How this works on our side: In our corporate program, we don't teach theory. Before starting, every participant fills out a "5 tasks that eat up the most time" survey, and we select 3 priorities with the company. Over 2 weeks across 4 live sessions, the team builds these automations themselves. No programming is needed: participants describe the logic in words, and AI writes the code. The result is at least 3 working processes on your own data with a money-back guarantee. https://course.aiadvisoryboard.me/corporate?utm_source=blog&utm_medium=article_body&utm_campaign=corporate

FAQ

Do I need to know programming languages to write these requirements?

No. For modern AI tools, knowledge of your business process is enough. You describe the logic in English, and AI agents interpret it into a sequence of actions or code.

What if the process changes constantly?

That is exactly the advantage of AI. If conditions change (e.g., new prices or shipping rules), you simply update the text instruction (the prompt), and the automation starts following the new rules without rewriting code.

How do I know if a task can be automated at all?

A simple test: if you can record a screen video (Loom) of you performing the task while explaining "if I see A, I do B," it can be automated. If the decision is based on "intuition," AI cannot help yet.

Who should describe these tasks: the founder or staff?

The best results come from a partnership: the founder sets the business goal (e.g., faster response times), and the employee doing the manual work describes the granular steps. Training the entire team on AI is usually more effective than isolated automation attempts.

Conclusion

Describing a task for AI is not a technical document; it is a clear instruction for a smart assistant. Stop thinking in terms of "buttons" and start thinking in terms of "inputs" and "outputs."

Tomorrow morning, pick one routine task that annoys you and try breaking it down into the 4 blocks of our template. It is the first step toward freeing your time for strategy instead of moving files. If you want to review your specific task with an expert, I invite you to a free diagnostic call.