Best Practices for Getting Started

You have set up your chat widget and connected your team platform - congratulations. But setup is only half the equation. The difference between a live chat experience that delights visitors and one that frustrates them comes down to preparation, training, and thoughtful configuration. Here are ten best practices for launching Social Intents successfully, along with common mistakes to avoid and answers to frequently asked questions.

This guide assumes you have already completed setup. If you have not yet created your widget and connected your team platform, start with Set Up Your First Chat Widget first, then return here before going live with customers.

1. Write a Clear Welcome Message

Your welcome message is the first thing visitors see when they open the chat widget. It sets the tone for the entire conversation and determines whether visitors engage or close the window. Research shows that a well-written welcome message can increase chat engagement rates by 30 to 40 percent compared to a generic greeting.

A strong welcome message should accomplish three things in one or two sentences:

Be short. One or two sentences is ideal. Visitors scan, they do not read paragraphs. Keep it under 200 characters for maximum impact.
Set expectations. Tell the visitor whether they are talking to an AI chatbot, a human agent, or both. Transparency builds trust from the first interaction.
Be actionable. Invite them to ask a question or describe what they need. Give them a clear next step so they know how to start the conversation.

Welcome Message Examples That Work

Hi! 👋 How can we help you today? Our AI assistant can answer most questions instantly, or connect you with our team.

Welcome to [Company Name]! Ask me anything about our products, pricing, or support - I'm here to help.

Hey there! I'm an AI assistant trained on [Company Name] docs. Ask me a question, or type "agent" to chat with a person.

Welcome Messages to Avoid

❌ Don't	✅ Do Instead	Why
"Welcome to our website. We are here to help you with any questions you may have about our products and services. Feel free to ask us anything."	"Hi! 👋 How can we help you today?"	Too long. Visitors stop reading after the first sentence.
"Chat with us"	"Ask me anything about pricing, features, or getting started."	Too vague. Give visitors a specific prompt to respond to.
(No welcome message at all)	Always set a welcome message - it invites engagement.	A blank chat window is intimidating. Visitors need a prompt to start talking.
"Please hold, an agent will be with you shortly."	"Our AI can help right away! Ask a question, or I'll connect you with our team."	Asking visitors to wait before they have even asked a question drives them away.

2. Name Your Widgets Descriptively

If you plan to use more than one widget - for example, one per website, one per department, or one per landing page - give each widget a clear, descriptive internal name. Good naming helps you find the right widget in your dashboard and makes your analytics reports meaningful.

Good names: "Main Website Chat," "Pricing Page Widget," "Support Portal Chat," "Blog Sidebar Widget"
Bad names: "Widget 1," "Test," "New Widget," "asdf"

The internal widget name is not visible to visitors - it is only for your reference in the dashboard and reports. Take 10 seconds to name it properly and save yourself confusion later. Learn more in When to Use Multiple Chat Widgets.

3. Set Up Business Hours

Configure your online schedule so Social Intents knows when your team is available to respond to live chats. When your team is offline, the platform can behave in three different ways depending on your configuration:

Keep the AI chatbot running independently - Visitors still get instant answers from your AI around the clock. The chatbot handles conversations on its own and only shows a "leave a message" option when visitors explicitly request a human agent. This is the recommended configuration for most teams.
Show an offline form - Visitors see a simple form asking for their name, email, and message. Your team receives these messages and can follow up by email when they return online.
Hide the widget entirely - The chat button disappears from your website outside of business hours. This ensures visitors never start a conversation that cannot be answered.

Recommended: Keep the AI chatbot running 24 hours a day, 7 days a week, even when your human agents are offline. Data shows that 40 to 60 percent of common customer questions can be answered by a well-trained chatbot without human involvement. Visitors who need a human can leave a message for follow-up during business hours. Learn more in What Happens When Everyone Is Offline.

4. Train Your AI Chatbot Before Launch

If you are using an AI chatbot, invest time in training it properly before going live with real customers. The quality of your chatbot's answers is directly proportional to the quality and breadth of the training content you provide. A well-trained chatbot can handle 50 to 75 percent of visitor questions without human intervention, dramatically reducing your team's workload and providing instant answers that visitors love.

Phase 1: Quick Setup (Day 1)

Add your main website pages. Paste your homepage, pricing page, features page, and "About" page URLs into the chatbot training. The AI will crawl these pages and learn from their content. Start with the five pages that receive the most traffic.
Write 5 to 10 Q&A pairs for your most common questions - things like "How much does it cost?", "Do you offer a free trial?", "What payment methods do you accept?", "How do I contact support?", and "What are your business hours?"
Set a clear system instruction that tells the chatbot who it is, what company it represents, what tone to use (friendly, professional, casual), and what it should not discuss. See Writing Effective System Instructions.

Phase 2: Expand Content (Week 1)

Upload product documentation - PDFs, help docs, user guides, and knowledge base articles give the chatbot deep product knowledge
Add your FAQ pages - If you have an existing FAQ section, paste the URL or create Q&A pairs from it. FAQs are among the highest-value training sources because they directly address real customer questions.
Review your first 20 to 30 conversations - Check chat transcripts in the Chat Transcripts section to see what visitors are actually asking. Identify the questions your chatbot struggles with and add training content to address them.

Phase 3: Ongoing Refinement

Monitor chatbot accuracy weekly - Review transcripts regularly and flag conversations where the chatbot gave incorrect or unhelpful answers
Add content for unanswered questions - When the chatbot cannot answer something, create a Q&A pair or add a URL that covers that topic
Refine system instructions - Adjust tone guidance, add rules for sensitive topics, and clarify escalation boundaries based on what you observe in real conversations

Detailed guide: Training Your Chatbot on Your Content

5. Add Your Team Members

Do not go live as a solo agent. Invite your team members so multiple people can respond to chats simultaneously. This ensures faster response times, coverage across time zones, and backup when individual agents are busy or away.

On the Basic plan ($69/month) and above, you get unlimited agent seats - so there is no cost to add everyone who might need to respond to a customer conversation. Even team members who only handle chats occasionally should be added so they can step in during busy periods.

Recommended Team Roles

Role	Who Should Have This Role	What They Can Do
Admin	Account owner or IT manager	Full access - manage billing, settings, integrations, agents, and all configuration
Manager	Team leads and supervisors	View reports, manage agents, configure widgets and chatbot settings
Agente	Everyone who responds to customer chats - support reps, sales reps, success managers	Respond to chats, view their own transcripts, update their profile

Learn about permissions in detail in Understanding Roles: Agent vs Manager vs Admin.

6. Configure Notifications Properly

Notifications are the lifeline of live chat. If your team does not receive timely notifications when a visitor starts a chat, the visitor waits - and studies show that visitors who wait more than 30 seconds without a response are significantly more likely to abandon the conversation.

Verify these notification settings before going live:

Channel notifications in your team platform - Make sure the channel where chats arrive has notifications enabled. In Slack, right-click the channel and select "Change notifications." In Teams, click the channel name and go to "Channel notifications."
Desktop notifications - Enable desktop push notifications for the chat channel in your team platform. Test by sending a test message and confirming a pop-up alert appears.
Mobile notifications - Install the Teams, Slack, or Google Chat mobile app and enable push notifications. This ensures agents receive alerts when they are away from their desk.
Email notifications as backup - Configure email notification rules in your Social Intents settings as a fallback for missed chats. Email will not produce real-time responses, but it prevents chats from being lost entirely.

Detailed guide: Notifications and Mobile Alerts

Missed chats are the number one cause of poor customer experience with live chat. If your team does not see a notification, the visitor waits in silence, becomes frustrated, and leaves. Test notifications with a real chat before going live and verify they work on both desktop and mobile devices for every team member.

7. Customize the Widget to Match Your Brand

Visitors trust a chat widget that looks like a natural part of your website. A widget that clashes with your design or looks like a generic third-party tool reduces credibility and engagement. Before launch, spend a few minutes on visual customization:

Match your brand color. Set the widget's primary color to your brand's primary color. This makes the chat button and header blend with your site design. Enter your hex code (e.g., #4F46E5) in the widget appearance settings.
Add agent photos. Upload real profile photos for your team members. Visitors are measurably more likely to engage when they see a human face rather than a generic icon. If you use an AI chatbot, consider using your company logo as the bot's avatar.
Choose the right position. The bottom-right corner is the industry standard and where visitors instinctively look for chat. Only choose bottom-left if your site has a conflicting element (like a scroll-to-top button or another widget) on the right side.
Use a clear button label. Instead of "Chat," use something actionable like "Chat with us," "Need help?," or "Ask a question." The label should tell visitors what will happen when they click.

Detailed guide: Customizing Chat Widget Appearance

8. Add a Pre-Chat Form

A pre-chat form collects the visitor's name and email before the conversation starts. This gives your agents context to personalize their responses and ensures you have contact information for follow-up if the visitor disconnects mid-chat.

Recommended field configuration:

Name (required) - Agents can address visitors by name, creating a more personal and professional interaction
Email (required) - Enables email follow-up if the chat is disconnected or if the visitor needs to be contacted later with additional information
Department or topic (optional) - A dropdown selector like "Sales / Support / Billing" helps route the conversation to the right team and gives agents immediate context about the visitor's needs

Balance data collection with friction. Every required field you add to the pre-chat form reduces the number of visitors who start a chat. For most use cases, name and email are sufficient. You can always collect additional information during the conversation itself. If maximizing chat volume is your priority, consider making the pre-chat form optional or disabling it entirely and letting the AI chatbot collect information conversationally.

Setup guide: Pre-Chat Forms and Custom Fields

9. Plan Your Escalation Strategy

Before going live, define a clear plan for how conversations should be escalated from your AI chatbot to human agents. A well-designed escalation strategy ensures that visitors who need human help get it quickly, while the chatbot handles routine questions independently.

Answer these three questions:

What phrases trigger escalation?

Configure explicit escalation triggers - the phrases visitors type when they want to talk to a human. Common triggers include "talk to a person," "human agent," "speak to someone," "transfer me," and "live support." Your chatbot should also be configured to escalate automatically when it detects visitor frustration or when it cannot answer a question after two or three attempts. See Escalation to Live Agents.

Where should escalated chats go?

For simple setups, route all escalated chats to one general support channel. For larger teams, use AI-powered escalation routing to direct chats to specific channels based on topic - billing questions to the billing team, technical issues to engineering, sales inquiries to the sales pipeline channel.

What happens if no agent is available?

Configure your fallback behavior for the scenario where a visitor requests a human agent but no one is online. Options include keeping the chatbot engaged while noting that a human will follow up, collecting the visitor's email and creating a support ticket, or displaying estimated wait times and queue position.

10. Run a Full Team Test Before Going Live

Before opening live chat to customers, run a coordinated test with your entire team. Solo testing does not reveal team workflow issues like notification gaps, response coordination, and channel confusion.

Have each team member join the correct channel in Teams, Slack, or Google Chat. Verify that every agent receives notifications when a test chat arrives - check both desktop and mobile.

Have someone act as a visitor and start multiple test chats - some that the AI chatbot can handle, some that should escalate to an agent, and some that test escalation routing if configured.

Have agents respond from both desktop and mobile devices to verify the experience on both. Check that responses appear in the visitor's widget within a few seconds.

Test concurrent conversations. Start three or four chats simultaneously to verify that agents can manage multiple conversations without confusion and that each conversation remains in its own thread.

Review the test chats in your Reports Dashboard to confirm that data is being captured correctly - chat count, response times, and transcript content should all be accurate.

Debrief with your team. Discuss what worked well and what needs adjustment. Fix any chatbot responses that were inaccurate, adjust escalation triggers that did not fire correctly, and ensure every agent's notifications are properly configured.

Full testing guide: Previewing and Testing Your Chat

Common Mistakes to Avoid

These are the most frequent mistakes teams make when launching live chat for the first time. Avoid them and you will be ahead of 90 percent of new deployments:

Mistake	Why It Hurts	What to Do Instead
Launching without training the chatbot	The chatbot gives wrong answers, visitors lose trust, and your team gets overwhelmed with escalations that the bot should have handled	Invest 30 minutes on Day 1 to add your top 5 website URLs and 10 FAQ pairs. It dramatically improves first-day performance.
Having only one agent monitoring chats	One person cannot cover meals, meetings, and bathroom breaks. Chats get missed.	Add at least two to three agents before launch. Unlimited agents are included on Basic plans and above.
Not testing notifications on mobile	Agents miss chats when away from their desk, leading to long wait times	Every agent should install their platform's mobile app and verify push notifications work with a test chat.
Setting a long, corporate welcome message	Visitors do not read it. Engagement starts lower because you appear impersonal.	Keep it to one or two sentences. Be warm, clear, and actionable.
Requiring too many pre-chat form fields	Every additional required field reduces chat starts. Four required fields can cut engagement by 50 percent.	Stick to name and email. Collect everything else during the conversation.
Forgetting to configure offline behavior	Visitors see "Chat with us" at midnight, start a conversation, and get no response. They leave frustrated.	Set up business hours and configure an offline fallback - either an AI chatbot running 24/7 or an offline message form.
Never reviewing chat transcripts	You miss insights about what visitors actually need, chatbot errors go unnoticed, and improvement stalls	Review 10 to 20 transcripts per week for the first month. Use insights to add chatbot training content and improve system instructions.

Launch Day Checklist

Use this checklist to confirm everything is ready before going live:

☐ Widget created and connected to your team platform
☐ Welcome message written (short, clear, actionable)
☐ AI chatbot enabled and trained on your most important content
☐ System instructions configured with tone, scope, and escalation rules
☐ Team members invited and roles assigned (Admin, Manager, Agent)
☐ Notifications verified for every agent on both desktop and mobile
☐ Business hours and online schedule configured
☐ Offline behavior configured (AI chatbot running, offline form, or hidden widget)
☐ Widget appearance customized to match your brand
☐ Pre-chat form enabled with name and email fields (if desired)
☐ Escalation strategy defined - triggers, routing, and fallback behavior
☐ Full end-to-end team test completed successfully
☐ Widget embed code installed on your website
☐ Widget verified on desktop and mobile browsers

Perguntas frequentes

How many chats should we expect on the first day?

Chat volume depends heavily on your website traffic and the visibility of your widget. Most teams see 5 to 20 chats on day one from a moderately trafficked website. Volume typically grows as visitors become accustomed to the chat option. To increase engagement, consider enabling proactive chat messages that invite visitors to start a conversation.

Should we enable the AI chatbot on day one or wait?

Enable it from day one. Even a minimally trained chatbot with your homepage URL and 5 to 10 Q&A pairs will handle basic questions and provide 24/7 coverage. You can refine it over the first week as you see what visitors ask. Waiting to enable the chatbot means missing the opportunity for instant responses during your launch period.

How quickly should agents respond to live chats?

Industry best practice is to respond within 30 seconds for the first agent reply. Visitors expect chat to be faster than email. If your AI chatbot handles the initial response, you have more time for human follow-up - but aim for under two minutes when an agent needs to take over. Use analytics to track your average response time and set team goals.

What if we get more chats than we can handle?

This is a good problem. First, ensure your AI chatbot is well-trained - it should handle routine questions without escalation. Second, add more agents. Third, use canned responses to speed up replies to common questions. Fourth, consider upgrading your plan to support higher chat volume. The Business plan includes 10,000 chats per month.

Can we turn off the widget temporarily?

Yes. You can disable a widget at any time from the dashboard by toggling it off. This immediately removes the chat button from your website without changing any code. Toggle it back on when you are ready. This is useful for maintenance windows, company events, or situations where your team is unavailable.

How do we measure success after launching live chat?

Track these key metrics in your analytics dashboard during the first month: total chat volume (trending up means visitors are engaging), average first response time (aim for under 30 seconds), chatbot resolution rate (percentage of chats handled without human intervention), customer satisfaction ratings (if you have post-chat surveys enabled), and escalation rate (the percentage of chatbot conversations that require human help - a decreasing trend means your chatbot is improving).