The Beginner's Guide to Conversation Starters

TL;DR: Conversation starters are the suggested prompts on your AI agent's welcome screen — up to 5 per agent, each optionally with an emoji. Well-written starters don't just hint at what the agent can do. They pull visitors into the specific interactions the agent handles best, setting up a strong first impression every time.

The welcome screen is the first thing every visitor sees. For most agents, it's also the thing builders spend the least time on.

Conversation starters are configured entry points — up to 5 per agent — that appear on the welcome screen as clickable buttons, reducing the first-message friction that prevents users from engaging with the agent's knowledge base.

That's a mistake. The conversation starters you set on your welcome screen shape every first interaction — and first interactions determine whether someone trusts your agent enough to keep talking to it. A visitor who has a great first exchange becomes a repeat user. A visitor who gets a vague or confused answer on their first try doesn't come back.

Here's how to use conversation starters intentionally.

What Conversation Starters Actually Are

Conversation starters are pre-written prompts displayed on your agent's welcome screen. When a visitor clicks one, it sends that prompt to the agent as if they'd typed it themselves, and the agent responds immediately.

You can add up to 5 starters per agent. Each one can optionally include an emoji — a small touch that makes the welcome screen feel less like a form and more like a person.

The starters appear before any conversation begins. They're the first impression alongside your welcome message. Get them right, and visitors immediately understand what the agent is good at and feel pulled toward a useful interaction.

One thing that surprises most builders: starters appear even if you've set a custom welcome message. The welcome message appears above the starters, not instead of them. This means your opening can do two jobs at once — the welcome message can explain what the agent does (two sentences), and the starters can demonstrate what kinds of questions it answers best (up to five specific examples). The combination of context plus concrete options consistently outperforms either alone.

The Two Jobs a Starter Has to Do

Every good conversation starter serves two purposes simultaneously:

1. It sets expectations. A visitor who sees "What's included in the $299 headshot package?" immediately knows this agent knows something specific and useful about the business — not just "general AI."

2. It leads to a win. The starter should trigger one of your agent's best answers. Not a question the agent handles adequately — a question it genuinely nails, with the right detail and the right tone.

When a visitor's first interaction is a win, the rest of the conversation goes better. They ask harder questions with more confidence. They're more forgiving of edge cases. They recommend the agent to others.

When the first interaction is a miss, the whole session is compromised.

Examples by Industry

The best starters are specific and predictive — specific enough to signal what the agent knows, predictive enough to guarantee a strong answer.

For a photography studio:

• 📸 "What's included in the $299 headshot package?"
• 📅 "How do I book a session?"
• 🌟 "What should I wear to my portrait session?"

For a life coach:

• 🔑 "Walk me through how the coaching process works"
• 📋 "What's in the 90-day program?"
• 💬 "What kinds of clients do you work with?"

For a university course:

• 📚 "When is the final exam and what does it cover?"
• ✏️ "How do I submit Assignment 3?"
• 🗓️ "What are the office hours this week?"

For a consulting practice:

• 🔎 "What's the difference between your two engagement types?"
• 📊 "Walk me through your diagnostic framework"
• ⚡ "What's the fastest way to get started?"

Notice what these starters have in common: they're questions with specific, detailed answers already in the knowledge base. You're not hoping the agent handles them well — you're choosing starters because you know it does.

The pattern across all these examples: the best starters are specific enough to signal scope but open enough to apply to multiple situations. "How do I reset my password?" is too narrow — it has one answer. "How do I get started?" is too broad — it applies to everything. "What should I do in my first week?" hits the sweet spot for an onboarding context: specific enough to be actionable, broad enough to cover the first-week question cluster.

The Emoji Question

You don't have to use emoji. But there's a reason most high-performing agents do: the welcome screen is an experience, and emoji create warmth without words.

A row of plain text questions looks like a FAQ page. The same questions with an emoji each look like something someone put thought into. That perception difference — between "auto-generated" and "crafted" — affects engagement.

Use them purposefully: one per starter, matched to the topic. Don't use random decorative emoji. 📸 for a photography question, 📚 for an educational one, 🔎 for something diagnostic. Aligned, purposeful emoji feel professional. Misaligned ones feel sloppy.

When to Update Your Starters

Starters aren't set-and-forget. After your agent has been live for a few weeks, you'll have real data on what visitors actually ask. Compare those real questions to your current starters.

If you see a question type coming up repeatedly in conversations that isn't represented in your starters, swap one in. If a starter is getting clicked but leading to mediocre answers, retire it and replace it with something the agent handles better.

Your starters should evolve as your knowledge base deepens. An agent that handles pricing questions better after you add a detailed pricing document should have a pricing starter it didn't have before.

Ready to level up your agent's welcome screen? Log in to Alysium and spend 10 minutes improving your conversation starters — it's one of the highest-ROI changes you can make.

For the full picture on making your agent perform well from the first interaction, see How to Make Your AI Agent Actually Useful (Not Just Cool).

The signal that a starter needs updating: users aren't clicking it. If your analytics show one starter getting 80% of clicks and three others getting near zero, the low-performers are either redundant, unclear, or not matching what users actually want. Replace low-performing starters with questions you've seen users ask in conversation history but that aren't currently surfaced on the welcome screen. The goal is a welcome screen where any of the five starters could be the most clicked depending on who the user is.