Configure Your AI Agent

Configure Your AI Agent

Your Synaptiq AI agent is the virtual sales rep that greets website visitors, answers their questions, and qualifies them as leads. This guide covers how to configure the agent's behavior, personality, qualification logic, and conversation flow so it represents your brand and captures the right information.

Access the Agent Settings

1. Log into the Synaptiq dashboard at /admin
2. Click AI Agent in the left sidebar
3. You will see four tabs: General, Qualification, Personality, and Preview

All changes save automatically. There is no publish step — updates take effect on your live widget within 60 seconds.

Choose an Industry Template

If you selected an industry during onboarding, your agent already has a template loaded. To change it or start from a different template:

1. Go to AI Agent > General
2. Click Change Template at the top of the page
3. Choose from the available templates:

| Template | Best For | Pre-configured Questions | |----------|----------|------------------------| | SaaS | Software companies selling B2B | Company size, current tools, budget, timeline, decision-maker role | | E-commerce | Online stores and DTC brands | Product interest, order size, wholesale vs. retail, shipping requirements | | Professional Services | Agencies, consultants, law firms | Project scope, timeline, budget range, previous provider experience | | Real Estate | Brokerages and property managers | Property type, location, budget, pre-approval status, move-in timeline | | Financial Services | Banks, fintech, insurance | Service interest, account type, investment range, compliance requirements | | Healthcare | Clinics, health tech, wellness | Patient type, insurance status, service needed, urgency | | Education | EdTech, courses, training programs | Learning goals, group size, budget, timeline, format preference | | Custom | Anything else | Starts blank — you define every question |

Selecting a template pre-fills your qualification criteria and greeting message. You can edit everything after applying it.

Set the Greeting Message

The greeting is the first message your agent sends when a visitor opens the chat widget. It sets the tone for the entire conversation.

Go to AI Agent > General > Greeting Message and write your greeting.

Guidelines for Effective Greetings

• Be specific about what you offer. Visitors should immediately understand what they can get help with.
• Give them options. Listing 2-3 things the agent can help with reduces the "blank page" problem.
• Keep it under 3 sentences. Long greetings feel like walls of text in a chat bubble.
• Match your brand voice. If your website is casual, the greeting should be casual.

Examples by Industry

SaaS:

"Hey! I'm the [Product] assistant. I can walk you through pricing, help you figure out which plan fits your team, or set up a demo with our sales team. What are you looking for?"

E-commerce:

"Hi there! Looking for something specific, or want help choosing the right product? I'm here to help."

Professional Services:

"Welcome! I can help you understand our services, get a ballpark estimate for your project, or connect you with a specialist. What brings you here today?"

Real Estate:

"Hi! Are you looking to buy, sell, or rent? I can help you find properties, understand pricing in your area, or schedule a showing."

Configure Qualification Criteria

Qualification criteria are the data points your AI agent collects during a conversation. These determine when a lead is considered "qualified" and what score it receives.

Add and Edit Criteria

1. Go to AI Agent > Qualification

2. You will see a list of qualification fields. Each field has:

   • Field name — The label shown in your leads dashboard (e.g., "Company Size")
   • Field key — The machine-readable identifier used in API exports and webhooks (e.g., company_size)
   • Type — Text, number, single-select, multi-select, email, phone, or URL
   • Required — Whether the agent must collect this field before marking a lead as qualified
   • Weight — How much this field contributes to the qualification score (1-10 scale)

3. Click Add Field to create a new criterion, or click the edit icon on an existing one

Recommended Fields for B2B SaaS

Here is a proven set of qualification fields for B2B software sales:

| Field Name | Type | Required | Weight | Why It Matters | |------------|------|----------|--------|---------------| | Full Name | Text | Yes | 2 | Basic identity | | Work Email | Email | Yes | 3 | Validates business interest (filters out personal emails) | | Company Name | Text | Yes | 2 | CRM matching, account research | | Job Title | Text | No | 4 | Identifies decision-maker vs. researcher | | Company Size | Single-select | Yes | 5 | Segments by ICP tier | | Current Solution | Text | No | 3 | Shows intent and switching cost | | Budget Range | Single-select | No | 6 | Strongest buying signal | | Timeline | Single-select | Yes | 7 | Urgency indicator | | Primary Use Case | Text | Yes | 4 | Product fit validation |

How Qualification Scoring Works

Synaptiq calculates a lead score from 0-100 based on the fields collected and their weights:

• Each collected field contributes points proportional to its weight
• Required fields that are missing cap the score at 40 (a lead cannot be "qualified" if required data is missing)
• Certain high-value signals add bonus points: work email domain matches a known company (+5), budget exceeds your minimum threshold (+10), timeline is under 30 days (+10)

You can customize these score thresholds under AI Agent > Qualification > Scoring Rules:

• Hot Lead: Score 80-100 (immediate notification, auto-route to calendar booking)
• Warm Lead: Score 50-79 (daily digest notification, added to nurture sequence)
• Cold Lead: Score 0-49 (logged but no notification, available for review)

Conversation Flow Control

The AI agent does not ask all qualification questions upfront like a form. Instead, it weaves them into a natural conversation. You can influence this flow:

• Question Order — Drag fields to reorder them. The agent prioritizes fields higher on the list but adapts based on the conversation.
• Minimum Questions Before Qualification — Set the minimum number of required fields that must be answered before the agent considers the conversation "complete." Default is all required fields.
• Follow-Up Depth — Choose how aggressively the agent asks follow-up questions: Light (accepts surface-level answers), Moderate (asks one follow-up), or Deep (probes for detail).

Customize the Agent's Personality

Under AI Agent > Personality, you control how the agent communicates.

Tone Presets

Select a baseline tone:

• Professional — Formal language, full sentences, no slang. Best for enterprise, legal, and financial services.
• Friendly — Warm and approachable with a conversational style. Best for SaaS, e-commerce, and consumer brands.
• Casual — Relaxed, uses contractions, sometimes starts sentences with "So" or "Hey." Best for startups and DTC brands.
• Technical — Precise language, comfortable with jargon, provides detailed explanations. Best for developer tools and engineering audiences.

Custom Instructions

Below the tone preset, there is a free-text field for Custom Instructions. This is where you give the agent specific behavioral guidance. These instructions are injected into the agent's system prompt.

Effective custom instructions include:

- Always mention our 14-day free trial when discussing pricing
- If someone asks about enterprise pricing, offer to connect them with our sales team
- Never badmouth competitors by name — focus on our strengths instead
- If the visitor seems frustrated or confused, offer to schedule a call with a human
- When asked about integrations, reference our integrations page at /integrations
- Do not discuss features we haven't launched yet (roadmap items)
- Our target customer is a VP of Sales or Revenue Operations leader at a company with 50-500 employees

Response Length

Control how verbose the agent is:

• Concise — 1-2 sentences per response. Gets to the point quickly.
• Balanced — 2-4 sentences. Provides context without rambling. (Recommended for most use cases)
• Detailed — 4-6 sentences. Thorough explanations, good for complex products.

Fallback Behavior

Configure what happens when the agent does not know the answer to a question:

1. Acknowledge and redirect — "I don't have specifics on that, but I can connect you with someone who does. Want me to schedule a quick call?" (Recommended)
2. Knowledge base only — The agent only answers questions it can source from your uploaded documents. Anything else gets the redirect.
3. Best effort — The agent uses its general knowledge to provide a helpful response, with a disclaimer that details may vary.

Test in Preview Mode

Before going live with changes, use the built-in Preview to test your agent.

1. Go to AI Agent > Preview
2. A simulated chat widget appears on the right side of the page
3. Have a conversation as if you were a visitor
4. Watch the Debug Panel on the left, which shows:
   • Which qualification fields were extracted and when
   • The agent's confidence level for each extracted field
   • The running qualification score
   • Which knowledge base documents were referenced (if any)
   • The full system prompt being used

Testing Checklist

Run through these scenarios before going live:

• [ ] Happy path — Answer all questions naturally and verify the agent collects every required field
• [ ] Reluctant visitor — Refuse to give your email or company name and see how the agent handles objections
• [ ] Off-topic question — Ask something unrelated to your product and verify the fallback behavior
• [ ] Competitor comparison — Ask "How do you compare to [competitor]?" and check that the response follows your custom instructions
• [ ] Pricing inquiry — Ask about pricing and verify the agent provides accurate information or escalates appropriately
• [ ] Technical question — Ask a product-specific question and verify the agent references your knowledge base (if uploaded)

Next Steps

• Upload Your Knowledge Base — Give the agent real product knowledge so it can answer specific questions
• Test Your First Conversation — Validate the complete lead capture flow
• Connect Integrations — Route qualified leads to your CRM or calendar