Lesson 2. Technical Specification (SOW)

Lesson 2. Technical Specification (SOW)#

Goal: create a detailed technical specification for the project.

SOW Structure#

1. Project description

• agent name
• target audience
• problem and solution

2. Functional requirements

What the agent must be able to do (list of features).

Example (qualification bot):

• accept requests from leads (Telegram, web form)
• ask 6 questions in sequence
• validate answers (e.g., budget must be a number)
• determine lead "temperature" (cold / warm / hot) based on answers
• send data to CRM (Bitrix24, AmoCRM)
• send notification to sales rep (Telegram, email)

3. Use cases

Describe how users will interact with the agent.

Example:

Use Case 1: Lead fills out questionnaire

1. Lead follows link to Telegram bot (from ad, website)
2. Bot greets: "Hello! I'll help you find a car. Answer a few questions."
3. Bot asks questions:
   • "What's your budget? ($K)"
   • "When do you plan to buy? (within a week / month / later)"
   • "Ready to meet with a sales rep? (yes / no)"
   • "What's your name?"
   • "Phone number?"
   • "Comment (if any)?"
4. Lead answers questions
5. Bot determines temperature:
   • Hot: budget > $200K, purchase within a week, ready for meeting
   • Warm: budget $100–200K, purchase within a month
   • Cold: budget < $100K or purchase later
6. Bot sends data to CRM (creates deal, sets temperature in field)
7. Bot sends notification to sales rep (Telegram): "New lead: [Name], budget [X], temperature: [Hot]. Phone: [number]."
8. Bot replies to lead: "Thank you! A sales rep will contact you within an hour."

4. Integrations

List of external services the agent will interact with.

Example:

• Telegram Bot API — for communicating with leads
• Bitrix24 API — for creating deals in CRM
• Google Sheets (optional) — for logging data
• OpenAI API (optional) — if free-text processing is needed (e.g., comments)

5. Logic and rules

Describe how the agent works.

Example (temperature determination):

IF (budget > 200000 AND purchase == "within a week" AND ready_for_meeting == "yes")
  → temperature = "Hot"
ELSE IF (budget >= 100000 AND purchase == "within a month")
  → temperature = "Warm"
ELSE
  → temperature = "Cold"

6. Data

What data the agent collects and stores.

Example:

7. Technical requirements

• platforms and tools
• performance (requests per minute)
• security (HTTPS, limited data access)

Example:

• Platform: Zapier (for chatbot/automation) or n8n (for more complex logic)
• Performance: up to 100 requests per hour (avg 2–3 per minute)
• Security: data transmitted via HTTPS, API keys stored in secret storage

8. Acceptance criteria

How to verify the agent works correctly.

Example:

• Bot responds to greeting (command /start)
• Bot asks all 6 questions in order
• Bot validates answers (budget is number, phone in correct format)
• Bot correctly determines temperature (hot / warm / cold)
• Data is sent to CRM (deal created with correct fields)
• Sales rep receives Telegram notification
• Lead receives confirmation ("A sales rep will contact you")

SOW Template#

# Technical Specification: [Agent Name]

## 1. Project Description
[Target audience, problem, solution]

## 2. Functional Requirements
[List of agent features]

## 3. Use Cases
[Use Case 1, Use Case 2, ...]

## 4. Integrations
[List of external services]

## 5. Logic and Rules
[Agent rules: conditions, formulas]

## 6. Data
[Table: data collected, types, required/optional]

## 7. Technical Requirements
[Platforms, performance, security]

## 8. Acceptance Criteria
[Checklist: how to verify the agent works]