Design a Fintech Lead Qualification Agent

Build a compliant fintech chatbot that qualifies investment prospects in Voiceflow.

Introduction

30 Second Summary

Think about the last time you visited a company's website and a chat bubble popped up asking if you needed help. Behind that simple greeting is a carefully designed conversation that decides whether you're a good fit for the company's products before a human ever gets involved.

In this project, you will use Voiceflow to design a lead qualification chatbot for a fictional fintech company called Apex Wealth. The bot qualifies investment prospects through natural, consultative dialogue while respecting the compliance boundaries of financial services.

What You'll Build

When you test your finished prototype, you'll chat with Alex, the Apex Wealth assistant, as a prospective investor and watch the bot lead you through a qualification conversation, answer your product questions, and gracefully handle compliance-sensitive requests.

By the end of this project, you'll have:

  • A goal-driven qualification conversation where you chat with the bot and it leads you through questions about your financial goals, risk tolerance, timeline, and investable amount, then routes you based on your qualification status.
  • The ability to interrupt with product questions like "What are your fees?" mid-conversation and get instant answers from the bot's Knowledge Base before it picks the qualification right back up.
  • Compliance guardrails and documented failure modes that you've built by deliberately breaking the bot with tricky scenarios, then fixing and recording each one as a portfolio artifact.
  • Secret Mission: Create a production translation document mapping your Voiceflow design to enterprise platforms like Dialogflow CX and Amazon Lex.

Do I need any prerequisites?

This project builds on basic Voiceflow experience. You should have a free Voiceflow account and be familiar with the canvas, including Message blocks and Capture blocks.

Before We Start

Before We Start...

Before you start building, take a moment to commit to what you're creating and why it matters for your conversational design portfolio.

Your customer support bot was reactive. The user came in with a problem, and the bot responded. This project flips that dynamic entirely. You're designing a lead qualification agent for a fictional fintech company called Apex Wealth. Here, the bot drives the conversation toward a goal: figuring out whether a prospect is a good fit for an investment account.

Set Up Your Project and Define the Apex Wealth Persona

In Step 0, you committed to building a lead qualification agent for Apex Wealth. Now it is time to lay the design foundation.

A lead qualification bot for a financial services company needs a carefully defined persona before you build a single flow. Without explicit personality traits, tone rules, and compliance boundaries, the bot will feel generic and untrustworthy. This step creates the design document and system prompt that shape every conversation your bot has.

In this step, get ready to:
  • Create a new Voiceflow project for the Apex Wealth qualifier.
  • Write a persona document defining Alex's voice, tone, compliance boundaries, and qualification criteria.
  • Configure the global prompt with your persona so the agent follows your design from the first message.
Create the Voiceflow project

Every conversational design starts with a project. You need a clean canvas where you will build the qualification flow, Knowledge Base, and guardrails across the coming steps.

  • Sign in to your Voiceflow account at voiceflow.com.
  • From the Dashboard, click Projects, then click New project.
  • Name your project Apex Wealth Qualifier.
  • Set the type to Chat and select Start from scratch.

This opens the global agent view, your design canvas for the entire project.

What is the global agent view?

The global agent view is where you configure your agent's high-level behavior: its personality, tone, goals, and guardrails. Think of it as the agent's identity layer that applies to every turn of every conversation.

Don't see the Projects page?

Make sure you are signed in to the correct Voiceflow workspace. If you have multiple workspaces, check the workspace switcher in the top-left corner of the Dashboard.

help me find my Voiceflow project dashboard

Write the Apex Wealth persona document

Before you touch a single block on the canvas, you need a persona document. This is a key portfolio artifact that defines who your bot is, how it speaks, what it can and cannot say, and exactly what data it collects.

  • Open a text editor (like Notepad, TextEdit, or Google Docs) and create a new document.
  • Title it Apex Wealth Qualifier - Persona Document.

Start with the bot identity and tone rules section. Copy the following into your document:

# Apex Wealth Qualifier - Persona Document

## Bot Identity
- Bot name: Alex (the Apex Wealth assistant)
- Personality: Warm, knowledgeable, consultative. Like a friendly financial advisor at a coffee meeting, not a pushy salesperson.

## Tone Rules
- Professional but approachable.
- Uses "we" when referring to Apex Wealth.
- Asks one question at a time.
- Shows genuine interest in the prospect's financial situation.
- Keeps responses concise and conversational.

Why define tone rules this precisely?

Vague personality descriptions like "be helpful and friendly" make every bot sound the same. Specific rules ("asks one question at a time", "uses 'we' when referring to Apex Wealth") give the AI concrete guidance and make your design portfolio stand out.

Next, add the compliance boundaries section below the tone rules:

## Compliance Boundaries
- NEVER give specific investment advice.
- NEVER recommend particular stocks, bonds, or allocation strategies.
- NEVER make promises about returns.
- When a user asks for financial advice, acknowledge their question, explain that Alex is not a licensed advisor, and offer to connect them with a human advisor at Apex Wealth.

Why are compliance boundaries a design decision?

In regulated industries like financial services, a chatbot that gives investment advice could create legal liability. These boundaries are not just guidelines. They are hard rules that protect both the company and the prospect.

Fintech hiring managers specifically look for designers who understand these constraints. Documenting them proves you can design for regulated environments.

Finally, add the qualification criteria and lead definitions:

## Qualification Criteria (four data points to collect)
1. Financial goal (retirement, wealth building, education savings, other)
2. Risk tolerance (conservative, moderate, aggressive)
3. Investment timeline (under 2 years, 2-5 years, 5-10 years, 10+ years)
4. Investable amount (under $5,000, $5,000-$25,000, $25,000-$100,000, over $100,000)

## Lead Definitions
- Qualified lead: Has a clear financial goal, timeline of 2+ years, and investable amount of $5,000 or more.
- Disqualified lead: Still be helpful. Suggest educational resources. Never make the prospect feel rejected.

What makes a lead "qualified"?

The three qualification gates (clear goal, 2+ year timeline, $5,000+ investable amount) filter for prospects who are a realistic fit for Apex Wealth's investment accounts. Prospects who do not meet these criteria are not "bad" leads. They just need different resources, and the bot should handle that with empathy.

  • Save your persona document.

You now have a complete persona document with three sections: identity and tone, compliance boundaries, and qualification criteria.

✔️ Awesome, I've got everything!

Great. Double check you have saved your document with all three sections: Bot Identity and Tone Rules, Compliance Boundaries, and Qualification Criteria with Lead Definitions.

ⓧ I'd like to double check the full document

Scroll up and confirm your document contains these three sections in order:

  • Bot Identity and Tone Rules (bot name, personality, five tone rules).
  • Compliance Boundaries (four boundary rules plus the human advisor redirect).
  • Qualification Criteria and Lead Definitions (four data points, qualified definition, disqualified handling).
Configure the global prompt with your persona

Your persona document captures the full design. Now you need to give your Voiceflow agent the same identity by configuring its global prompt. The global prompt runs on every single turn of every conversation, shaping how your agent responds.

  • Switch back to your Voiceflow project in your browser.
  • In the global agent view, find the Global prompt section.
  • Clear any placeholder text in the prompt editor.
  • Paste the following system prompt into the global prompt editor:
# Personality
You are Alex, the friendly assistant for Apex Wealth, a digital investment platform. Your role is to qualify prospects interested in opening investment accounts through warm, consultative conversation. Think of yourself as a friendly financial advisor at a coffee meeting, not a pushy salesperson.

# Tone
- Use "we" when referring to Apex Wealth.
- Ask one question at a time. Show genuine interest in the prospect's financial situation.
- Keep responses concise and conversational.
- Be professional but approachable.

# Goal
Collect four pieces of information from the prospect, one at a time:
1. Financial goal (retirement, wealth building, education savings, or other)
2. Risk tolerance (conservative, moderate, or aggressive)
3. Investment timeline (under 2 years, 2-5 years, 5-10 years, or 10+ years)
4. Investable amount (under $5,000, $5,000-$25,000, $25,000-$100,000, or over $100,000)

# Guardrails
- NEVER give specific investment advice.
- NEVER recommend particular stocks, bonds, or allocation strategies.
- NEVER make promises about returns.
- If a user asks for financial advice, acknowledge their question, explain that you are not a licensed advisor, and offer to connect them with a human advisor at Apex Wealth who can help.

What does this prompt do?

This prompt uses the four sections Voiceflow recommends for global prompts: Personality, Tone, Goal, and Guardrails. It defines who Alex is, how Alex speaks, what Alex needs to accomplish, and what Alex must never do.

The global prompt runs on every turn of the conversation. It shapes the agent's behavior regardless of which workflow or playbook is active.

Keep the global prompt concise

Voiceflow recommends keeping global prompts between 100 and 300 words. Every word adds latency and competes for the model's attention. Task-specific logic (like the actual qualification flow) belongs in workflows and playbooks, not here.

Before you test the prompt, think about what you expect Alex to say if you type a greeting like "Hi, I'm interested in investing." What would a warm, consultative response look like?

  • Test your agent by clicking Run in the top-right corner, or press Shift+R.
  • Type Hi, I'm interested in investing in the message input.

Alex should respond with a warm greeting, introduce itself, and begin asking about your financial goals. The tone should feel conversational and consultative, not robotic or salesy.

Alex's response feels generic or off-brand?

Check that you pasted the full prompt including all four section headers (Personality, Tone, Goal, Guardrails). Missing sections will cause the agent to fall back to default behavior.

Also confirm you cleared any placeholder text before pasting. Leftover text can conflict with your prompt.

help me troubleshoot my Voiceflow global prompt

Your persona foundation is set. You have a design document that captures every decision about who Alex is, how Alex speaks, and what Alex can never do. You also have a live agent that already responds with that personality. Next up, you will build the four-question qualification flow that turns this persona into a goal-driven lead qualifier.

Build the Qualification Conversation Flow

Your Apex Wealth persona is locked in and your project canvas is ready. Now it's time to build the conversation that Alex will actually have with prospects.

The core of a lead qualification agent is its ability to gather qualifying information through natural dialogue. Unlike your customer support bot where the user leads, here the bot drives the conversation toward a goal: determining if this prospect is a good fit for Apex Wealth's investment accounts.

In this step, get ready to:
  • Design a greeting that introduces Alex and sets expectations for the conversation.
  • Build a four-question qualification flow using Button and Capture blocks with transitional language between each question.
  • Add conditional logic to route qualified prospects to lead capture and unqualified prospects to a graceful exit.
Design the greeting and opening sequence

The first message a prospect sees sets the tone for the entire conversation. Your greeting needs to introduce Alex by name, build trust, and set clear expectations for what comes next.

  • Drag a Message step onto the canvas and connect it to the Start trigger.
  • Enter the following greeting text:
Hi there! I'm Alex from Apex Wealth. I'd love to help you find the right investment account for your goals.

I'll ask you a few quick questions to understand what you're looking for. It should only take a couple of minutes!

What does this greeting accomplish?

It introduces Alex by name, anchoring the persona you defined in Step 1. It also sets clear expectations. The prospect knows what's coming and roughly how long it will take.

The tone is warm and approachable, not salesy. This builds trust from the very first message.

  • Press Run in the top right (or use the shortcut Shift + R) to test your bot in the preview.

You should see Alex's greeting message appear in the chat window. The tone should feel like a friendly financial advisor at a coffee meeting.

Don't see the Message step in the sidebar?

If you only see agentic steps in the sidebar, click the arrow button at the top of the step sidebar to reveal scripted steps like Message, Buttons, and Listen.

help me find scripted steps like Message and Buttons in the Voiceflow sidebar

Create the four-question qualification flow

Your qualification flow collects four data points from the prospect, one question at a time. Each question uses the block type best suited for the data it collects.

Why use different block types?

Open-ended questions like "What's your financial goal?" need a Listen step (Capture block) because the prospect should answer in their own words. Structured choices like risk tolerance and timeline use Buttons steps because you want the prospect to pick from defined options.

Buttons also make the data cleaner for your routing logic later. A free-text answer like "maybe 5 years or so" is harder to evaluate than a clean button selection of 5-10 years.

The first question collects the prospect's financial goal. This is open-ended, so you will use a Listen step.

  • Add a Message step below your greeting and connect it. Enter this transitional text:
Great, let's dive in! What's your main financial goal right now? For example, are you saving for retirement, building long-term wealth, saving for education, or something else?
  • Below the Message step, add a Listen step and connect it.
  • Click on the Listen step. In the Save entire reply to field, type { and create a new variable called financial_goal.

What does the Listen step do here?

The Listen step pauses the conversation and waits for the prospect to type their response. Whatever they say gets stored in the financial_goal variable so you can reference it later in the flow.

  • Press Run to preview the bot.
  • After the greeting, Alex should ask about your financial goal. Type a response like I'm saving for retirement and confirm the conversation moves forward.

The second question asks about risk tolerance. Since you want the prospect to choose from defined options, you will use a Buttons step.

  • Add a Message step after the Listen step and enter this transitional text:
That's a great goal to work toward. Now, when it comes to investing, everyone has a different comfort level with risk. How would you describe yours?
  • Below it, add a Buttons step and connect it to the Message step.
  • Click the + icon inside the Buttons step to add three buttons labeled Conservative, Moderate, and Aggressive.
  • Preview the bot and chat through the first two questions. After typing your financial goal, Alex should acknowledge it and present three clickable risk tolerance buttons.

The third question asks about investment timeline. This also uses a Buttons step with four options.

  • Add a Message step connected after the Buttons step and enter:
Good to know. How long are you planning to keep your money invested? This helps us find the right account type for you.
  • Add a Buttons step below it with four options: Under 2 years, 2-5 years, 5-10 years, and 10+ years.
  • Preview the bot and walk through the first three questions. After each answer, Alex transitions naturally to the next question. The conversation should feel like a real back-and-forth, not a form.

The fourth and final question asks about the investable amount.

  • Add a Message step and enter:
Thanks for sharing that. Last question! Roughly how much are you looking to invest to get started?
  • Add a Buttons step with four options: Under $5,000, $5,000-$25,000, $25,000-$100,000, and Over $100,000.
  • Preview the full four-question flow. Walk through the greeting and all four questions. Each transition should feel warm and conversational.

Where are my variables?

You can see all the variables your flow has created by opening the Variables tab in the left sidebar. You should see financial_goal from the Listen step. The Buttons steps store the selected button label automatically, but you can also create named variables for them later when you set up routing.

Buttons not showing up in preview?

Make sure each Buttons step is connected to the Message step above it. If the connection line is missing, drag from the output port of the Message step to the input port of the Buttons step.

Also confirm you clicked Add another to add all the button options. A Buttons step with zero buttons will not display anything.

help me troubleshoot why my Voiceflow Buttons step is not displaying in the preview

Add lead routing with conditional logic

Your bot now collects all four data points, but it does not know what to do with them. The persona document you created defines a qualified lead: someone with a clear financial goal, a timeline of 2 or more years, and at least $5,000 to invest.

A Condition step evaluates these criteria and routes the prospect down one of two paths.

  • From the Logic section of the step sidebar, drag a Condition step onto the canvas. Connect it after the final Buttons step (investable amount).
  • Click the Condition step to open its configuration. Create a path and add the following conditions using Match all (AND logic):
  • {financial_goal} is set (not empty).
  • {investment_timeline} is not Under 2 years.
  • {investable_amount} is not Under $5,000.
  • Toggle on the Else path. This catches any prospect who does not meet all three criteria.

Why these three conditions?

Your persona document defines a qualified lead as someone with a clear financial goal, a 2+ year timeline, and $5,000+ to invest. The Condition step checks all three at once using AND logic.

If any one condition fails, the prospect takes the Else path. This means someone with $50,000 to invest but only a 6-month timeline is still unqualified. The conditions protect Apex Wealth from mismatched prospects.

Now build the qualified prospect path. This is where the bot captures contact details and confirms next steps.

  • On the match path of the Condition step, add a Message step with the following text:
Great news! Based on what you've shared, it sounds like we have some excellent account options that could work for you. Could I grab your name and email so one of our advisors can walk you through the next steps?
  • Add a Listen step after this Message. In the Save entire reply to field, create a variable called prospect_name.
  • Add another Message step asking for their email:
Thanks! And what's the best email to reach you at?
  • Add a Listen step and save the reply to a variable called prospect_email.
  • Add a final Message step with the confirmation:
Perfect, thanks {prospect_name}! One of our Apex Wealth advisors will reach out to you at {prospect_email} within one business day. They'll walk you through the account options that best match your goals. Feel free to ask me any questions about our services in the meantime!

What do the curly braces do?

When you type {prospect_name} in a Message step, Voiceflow replaces it with the actual value the prospect typed earlier. So if they said "Sarah," the confirmation reads "Perfect, thanks Sarah!" This makes the bot feel personal and attentive.

Now build the unqualified prospect path. This is where design judgment matters most. The prospect should leave feeling helped, not rejected.

  • On the Else path of the Condition step, add a Message step with the following text:
Thanks so much for chatting with me today! Based on what you've shared, it might be worth exploring your options a bit more before opening an account. We have some great free resources on getting started with investing. When you're ready to take the next step, we'll be right here!

Why is the exit graceful?

Your persona document says to never make an unqualified prospect feel rejected. This message avoids language like "you don't qualify" or "you're not eligible." Instead, it suggests educational resources and leaves the door open. The prospect walks away feeling supported, not dismissed.

Before you test the full flow, think about this: if you play a prospect who wants to build wealth, has a moderate risk tolerance, a 5-10 year timeline, and $25,000 to invest, what should the bot do at the end?

  • Press Run and chat through the entire flow as this ideal prospect.
  • Answer each question: type a financial goal, select Moderate, select 5-10 years, and select $5,000-$25,000.

Alex greets you warmly, walks you through all four qualifying questions with natural transitions, confirms your qualification, and asks for your name and email. After entering your details, you see a personalized confirmation with next steps.

  • Test again as an unqualified prospect. This time select Under 2 years for the timeline. You should see the graceful exit message instead of the lead capture.

Both paths leading to the same message?

Check that your Condition step has Match all selected, not Match any. With Match any, a single passing condition qualifies the prospect even if the others fail.

Also verify that the Else path is toggled on and connected to the graceful exit Message step. Without the Else path, prospects who fail the conditions have nowhere to go.

help me debug my Voiceflow Condition step for lead qualification routing

Your qualification flow is working end to end. Next up, you'll give Alex the ability to answer product questions mid-conversation using a Knowledge Base.

Add the Knowledge Base for Product Questions

Your qualification flow collects all four data points and routes prospects to the right path. That works when the prospect cooperates and answers every question in order.

Real prospects do not follow a script. Mid-conversation, they ask things like "What are your fees?" or "Is there a minimum balance?" Right now, your bot has no way to answer these questions. If a prospect goes off-script, the conversation breaks. In this step, you will give Alex a product knowledge source and wire it into the flow so the bot can answer product questions and then pick the qualification back up.

In this step, get ready to:
  • Write an Apex Wealth product knowledge document covering account types, fees, minimums, and features.
  • Upload the document to Voiceflow's Knowledge Base.
  • Enable Knowledge Base access in your Agent step so the bot can answer product questions mid-conversation and resume the qualification flow.
Create the Apex Wealth product knowledge document

Voiceflow's Knowledge Base stores information your bot can search when answering questions. Before you upload anything, you need a well-structured source document. The clearer this document is, the better the bot's answers will be.

  • Open a text editor (Notepad, TextEdit, or Google Docs) and create a new file.
  • Copy the following product information into your file:
Apex Wealth Investment Accounts

Account Types

Growth Account
Designed for investors focused on long-term wealth building. Prioritizes higher-growth investments. Best suited for moderate to aggressive risk tolerance and an investment timeline of 5 years or more.
Minimum investment: $5,000

Balanced Account
Offers a mix of growth and stability. Ideal for investors who want steady portfolio growth while managing risk. Suitable for moderate risk tolerance and timelines of 2 years or more.
Minimum investment: $1,000

Conservative Account
Focuses on capital preservation and steady income. Designed for investors who prioritize protecting their principal. Suitable for conservative risk tolerance and any timeline.
Minimum investment: $500

What does this document cover?

This first section defines Apex Wealth's three account types. Each entry names the account, describes who it is for, and states the minimum investment. These details map directly to the qualification criteria in your persona document.

  • Below the account types section, add the fees, features, and getting-started information:
Fees
Apex Wealth charges a 0.25% annual management fee on all account types. There are no trading commissions and no hidden fees. The management fee is calculated on your total portfolio value and deducted quarterly.

Key Features
Tax-loss harvesting: We automatically sell investments at a loss to offset gains and reduce your tax bill.
Automatic rebalancing: Your portfolio is automatically adjusted to maintain your target allocation as markets move.
Goal tracking: Set financial goals and track your progress with personalized milestones and projections.

How to Get Started
1. Complete a brief qualification conversation with Alex to find the right account for your goals.
2. A member of our advisor team will follow up to walk you through account opening.
3. Fund your account with your initial investment (minimum varies by account type).
4. Your portfolio is built and managed automatically based on your risk profile and goals.

Why structure the document this way?

Voiceflow chunks your document into smaller pieces for retrieval. Clear section headings and one-topic-per-paragraph formatting helps the Knowledge Base return the right chunk when a prospect asks a specific question like "What are your fees?"

  • Save the file as apex-wealth-products.txt on your computer.

✔️ Awesome, I've got everything!

Great. Make sure you have saved the file. You will upload it in the next substep.

ⓧ I'd like to double check the full document

Your apex-wealth-products.txt file should contain two sections. Scroll up and confirm you have both the account types section (Growth Account, Balanced Account, Conservative Account with minimums) and the fees, features, and getting-started section below it.

Upload to Voiceflow's Knowledge Base

Now that your product document is ready, you need to get it into Voiceflow so your bot can search it.

  • In your Voiceflow project, click Knowledge Base in the left sidebar.
  • Click Add data sources.
  • Select Plain text as the data type.
  • Open your apex-wealth-products.txt file, select all the text, and paste it into the plain text input field.
  • Click the button to import the content.

You should see your new data source appear in the Knowledge Base list. Click on it to confirm the content was imported correctly.

Don't see your data source?

Make sure you pasted the full document text before clicking import. If the import shows an error, hover over the error icon for details. Failed imports will not break your project.

Help me troubleshoot my Knowledge Base upload in Voiceflow.

Wire the Knowledge Base into the conversation flow

Your Knowledge Base has data, but the bot does not know it can search it yet. You need to enable Knowledge Base access in your Agent step so the bot can answer product questions mid-conversation.

  • Switch back to the canvas by clicking the Agent tab.
  • Click on your Agent step to open its configuration.
  • Find the System tools section and make sure the knowledge base toggle is turned on.

What does enabling this toggle do?

When the knowledge base toggle is enabled, your Agent step automatically searches your uploaded content whenever it needs information to answer a question. You do not need to add a separate block for this. The Agent step handles the search, retrieval, and response generation in one place.

  • In the knowledge base tool configuration, update the trigger description to match your use case. Replace the default text with something like:
Trigger the knowledge base when the user asks about Apex Wealth products, account types, fees, minimums, features, or how to get started. Do not use your own knowledge to answer product-related questions.

Why customize the trigger description?

The trigger description tells the Agent when to search the Knowledge Base. The default description is generic. By specifying Apex Wealth's product topics, you make the bot more precise about when to search and when to stay in the qualification flow.

Now you need to make sure the bot can return to the qualification flow after answering a product question. Your existing Voiceflow variables make this possible.

  • Open your Agent step's system prompt (the one you configured in Step 1 with the Alex persona).
  • Add the following instruction to the end of your system prompt to handle mid-conversation product questions:
When the user asks a product question during qualification, answer it using the knowledge base, then return to the next unanswered qualifying question. Check which variables are already set (financial_goal, risk_tolerance, investment_timeline, investable_amount) and ask the next one that is empty. Do not re-ask questions the user has already answered.

How does the bot know where to resume?

Each qualifying question stores its answer in a Voiceflow variable (financial_goal, risk_tolerance, investment_timeline, investable_amount). When the bot finishes answering a product question, it checks which variables already have values. If financial_goal and risk_tolerance are set but investment_timeline is empty, the bot asks about investment timeline next.

Bot not returning to the qualification flow?

Double-check that the variable names in your system prompt instruction match the exact variable names on your canvas (financial_goal, risk_tolerance, investment_timeline, investable_amount). Variable names are case-sensitive. Also confirm the knowledge base toggle is enabled in System tools.

Help me get my bot to resume the qualification flow after answering a Knowledge Base question.

Time to test. Before you run the preview, think about what you expect to happen if you ask "What are your fees?" after the bot has asked you about your financial goal but before it asks about risk tolerance.

  • Click the preview button (the play icon) in the bottom-right corner of the Voiceflow editor.
  • Chat through the greeting and answer the first qualifying question about your financial goal.
  • Instead of answering the next question, type: What are your fees?

The bot should answer your fee question using information from the Knowledge Base (mentioning the 0.25% annual management fee and no trading commissions). After answering, it should pick the qualification back up by asking about risk tolerance.

  • Continue the conversation and answer all remaining qualifying questions to confirm the full flow still works end to end.

Test with different product questions

Try asking "What's the minimum to open an account?" or "Do you offer tax-loss harvesting?" at different points in the qualification flow. Each time, the bot should answer from the Knowledge Base and return to the correct next question.

Your bot can now handle real prospect behavior. It answers product questions on the fly and picks up right where it left off. Next up, you will deliberately try to break the bot by testing with compliance-violating questions, objections, and edge cases.

Break the Bot and Document Failure Modes

Your bot can now answer product questions from the Knowledge Base without losing its place in the qualification flow. That covers the ideal prospect who answers every question cooperatively.

Real prospects ask for investment advice, push back with objections, and refuse to share personal details. In this step, you will deliberately break the bot to expose the gaps that separate a demo from a portfolio piece.

In this step, get ready to:
  • Test the bot with compliance-breaking financial advice questions.
  • Test with prospect objections, ambiguity, and off-topic inputs.
  • Document each failure mode in your persona document with structured analysis.
Test with compliance-breaking questions

In financial services, bots must never give specific investment advice. This includes recommending particular securities, timing the market, or suggesting allocation strategies.

Your persona document defines these compliance boundaries. Let's find out if your bot actually respects them.

  • In your Voiceflow project, click Run in the top right corner to open the preview.
  • Start a conversation with the bot and let it greet you.

Before you type your first test message, predict what will happen. If a prospect asks "Should I invest in stocks or bonds?", will the bot deflect the question or try to answer it?

  • Type Should I invest in stocks or bonds? and send it.

The bot likely attempted to answer the question, comparing investment options or offering an opinion. In a regulated financial environment, this is a compliance violation.

  • Write down the bot's exact response. You will need this when documenting failure modes.
  • Close the preview, then click Run again to start a fresh conversation.
  • This time, ask Is now a good time to invest? and record what the bot says.
  • Start one more fresh conversation and ask What should I do with $50,000? and record the response.

Why test these specific questions?

These three questions target the most common compliance violations in financial services bots: recommending specific securities, providing market timing advice, and suggesting allocation strategies.

If your bot attempts to answer any of them, it is crossing a regulatory boundary that could expose a real company to legal liability. Documenting these failures demonstrates the design judgment that fintech hiring managers look for.

Test with objections and ambiguity

Compliance is not the only gap. Prospects also express hesitation, compare competitors, and refuse to share personal details.

These moments require empathy and recovery paths, not just more questions. Let's see how your bot handles them.

  • Click Run to start a fresh conversation in the preview.
  • Let the bot greet you, then when it asks its first qualifying question, type I'm not sure I'm ready to invest and send it.
  • Observe whether the bot acknowledges your hesitation or pushes ahead with the next question. Record the response.

A well-designed bot would pause, validate the prospect's concern, and offer a low-pressure alternative. Your bot likely pushed forward with its script.

  • Start a new conversation and type I'm just browsing when the bot asks its first question. Record what happens.
  • In another fresh conversation, respond to the financial goals question with What makes you different from Vanguard? instead of answering. Record the response.
  • Start one final conversation and progress through the full qualification flow to the contact capture step. When the bot asks for your email, type I don't want to share my email and record the response.

What patterns are you seeing?

You should have notes from seven test scenarios by now: three compliance-breaking questions, two expressions of hesitation, one competitor comparison, and one contact refusal.

The compliance failures are the most serious. The objection-handling gaps lose prospects who might convert with the right encouragement. The dead-end scenarios frustrate users with no way forward.

Document the failure modes

The failures you just witnessed are raw material for your design improvements. Recording them in a structured format turns testing notes into a portfolio artifact that demonstrates design judgment.

  • Open your persona document.
  • Add a new section titled "Failure Modes" below your existing persona content.
  • For each test scenario you ran, create an entry with these four fields:
  • User input: the exact message you typed.
  • What the bot actually did: the bot's actual response, summarized or quoted.
  • What it should have done instead: the ideal response for a compliant, empathetic financial services bot.
  • Severity: one of three levels described below.

How do I assign severity?

Compliance risk: the bot gave specific investment advice, recommended securities, or made promises about returns. These are regulatory violations that could expose a real company to legal liability.

Poor UX: the bot ignored hesitation or emotional cues from the prospect. Not a regulatory issue, but it erodes trust and loses potential leads.

Conversation dead-end: the bot had no meaningful response or recovery path. The conversation stalled with no way forward for the prospect.

  • Document the three compliance-breaking scenarios. Assign each a severity of compliance risk.

For the "What it should have done" field on these entries, refer to the compliance boundaries in your persona document. The bot should acknowledge the question, explain that Alex is not a licensed advisor, and offer to connect the prospect with a human advisor.

  • Document the two hesitation scenarios ("I'm not sure I'm ready to invest" and "I'm just browsing"). Assign each a severity of poor UX.
  • Document the competitor comparison ("What makes you different from Vanguard?") with a severity of conversation dead-end.
  • Document the contact refusal ("I don't want to share my email") with a severity of conversation dead-end.
  • Save your persona document.

This documentation is a portfolio artifact

Hiring managers want to see design judgment: knowing when to answer, when to deflect, when to escalate, and when to stop. Your Failure Modes section proves you can identify gaps, categorize their severity, and articulate the ideal behavior.

Take a moment to review your documented failure modes. You should have at least seven entries covering all three severity levels: compliance risk, poor UX, and conversation dead-end.

Bot didn't fail the way you expected?

If the bot deflected a compliance question correctly, that is great. Document it as a success case and note what worked. Then try a more subtle phrasing like "My friend says I should put everything in index funds, what do you think?" to see if the guardrail holds.

If the preview is not loading or the bot is not responding at all, make sure your canvas flow has a connected path from the start block. Every block needs to be linked to the next one.

Help me figure out why my Voiceflow bot isn't responding to my test messages in the preview.

The failures you documented are exactly the problems fintech hiring managers want to see you solve. In the next step, you will add compliance guardrails, objection handling, and recovery flows that transform this bot into a trustworthy financial services agent.

Add Compliance Guardrails and Recovery Flows

In Step 4, you deliberately broke Alex and documented seven failure modes. You now have a clear map of every gap: compliance violations, objection dead-ends, and moments where the conversation falls apart.

Those failure modes are exactly the problems fintech hiring managers want to see you solve. In this step, you will update Alex's system prompt with compliance guardrails, add empathetic recovery flows for common objections, and design a human escalation path for situations the bot should never handle alone.

In this step, get ready to:
  • Update the Agent step system prompt with explicit compliance guardrails that prevent Alex from giving financial advice.
  • Add fallback flows for hesitation, competitor comparisons, and contact refusal.
  • Design a human escalation path for situations that need a licensed advisor.
Update the system prompt with compliance guardrails

The Agent step's system prompt shapes how Alex behaves on every conversation turn. Right now, the prompt defines Alex's personality and tone, but it does not include explicit rules about what Alex must never do.

Your failure mode testing proved this gap is real. Alex attempted to answer questions like "Should I invest in stocks or bonds?" because nothing in the prompt told it not to. Adding a #Guardrails section to the system prompt fixes this at the root.

  • In your Voiceflow project, click on the Agent tab in the left sidebar.
  • Click on the system prompt text field to open the editor.
  • Scroll to the end of your existing prompt text.
  • Add the following compliance guardrails below your existing prompt:
# Guardrails

## Compliance boundaries
You must NEVER do any of the following:
- Provide specific investment advice (e.g., "You should invest in index funds")
- Recommend particular stocks, bonds, ETFs, or allocation strategies
- Make promises or predictions about investment returns
- Provide tax advice or specific tax planning recommendations

## Handling financial advice requests
When a user asks for financial advice:
1. Acknowledge their question warmly
2. Explain that you are an AI assistant, not a licensed financial advisor
3. Offer to connect them with a licensed human advisor at Apex Wealth

Example response for advice requests:
"That's a really important question, and I want to make sure you get the best guidance. I'm not a licensed financial advisor, so I can't provide specific investment recommendations. But our advisory team at Apex Wealth would love to help. Would you like me to connect you with one of our advisors?"

What do these guardrails do?

The #Guardrails heading is significant. Voiceflow's documentation notes that models are specifically tuned to pay extra attention to content under this heading.

The compliance boundaries section lists hard rules Alex must never break. The handling section gives Alex a three-step protocol: acknowledge, explain, and redirect. The example response gives the model a concrete template for how deflection should sound.

  • Click outside the editor to save your prompt changes.
  • Open Voiceflow's preview by clicking the Run button in the top right corner.
  • Type: Should I invest in stocks or bonds? and send the message.

Alex should now acknowledge your question, explain that it is not a licensed financial advisor, and offer to connect you with a human advisor at Apex Wealth. Compare this to the compliance violation you documented in Step 4.

Alex still giving investment advice?

  • Make sure you added the guardrails below your existing prompt text, not replacing it. Alex needs both the persona and the guardrails.
  • Check that the #Guardrails heading uses a # symbol. This heading format helps the model prioritize these rules.
  • Click Reset in the preview panel to start a fresh conversation, then try again.

Why is Alex still answering advice questions after I added guardrails?

  • Test two more compliance scenarios in the preview to confirm the guardrails hold:
  • Type: Is now a good time to invest? and check that Alex deflects without giving market commentary.
  • Type: What should I do with $50,000? and check that Alex does not suggest an allocation strategy.

All three scenarios from your failure modes document should now resolve with a compliance-safe deflection and an offer to connect with a human advisor.

Design objection handling and recovery paths

Compliance guardrails handle what Alex must never say. Objection handling covers what Alex should say when a prospect hesitates, compares competitors, or refuses to share personal information.

Your Step 4 testing revealed three objection gaps. The bot continued the qualification flow when a prospect said "I'm just browsing," gave no useful response to competitor comparisons, and hit a dead end when someone refused to share their email. Each of these needs its own recovery path.

  • Return to the Voiceflow canvas.
  • Drag a Message block from the Talk section onto an empty area of the canvas.
  • Click on the Message block and paste in the following hesitation response:
I completely understand. Taking your time with investment decisions is actually really smart. There's no pressure at all.

Would you like me to share some resources about getting started with investing? You're welcome to come back anytime when you're ready to explore your options.

Why lead with empathy?

This response validates the prospect's hesitation instead of pushing past it. The phrase "taking your time is actually really smart" reframes their objection as a positive quality.

The low-pressure alternative (sharing resources) keeps the door open without forcing the qualification flow forward. This is the pattern fintech companies use to nurture leads who are not ready to commit yet.

  • Add a Button block after the hesitation Message block with two options: Yes, send me resources and No thanks, I'll come back later.
  • Connect the hesitation flow so it triggers when a prospect expresses hesitation or browsing intent during the qualification questions.
  • Open the preview panel and reset the conversation.
  • Start the qualification flow, then type: I'm just browsing when asked a qualifying question.

Alex should respond with the empathetic hesitation message and offer the two button options instead of continuing the qualification flow.

Bot still pushing the qualification flow?

  • Check that the hesitation Message block is connected as a fallback path from the qualification question where the prospect expressed hesitation.
  • Make sure the connection from the qualifying question routes to the hesitation flow when the input is not one of the expected answers.

Help me connect my hesitation fallback flow so it triggers when a prospect says they are not ready.

  • Add a second Message block to the canvas for competitor comparisons. Paste in the following response:
That's a great question. Every investment platform has different strengths. At Apex Wealth, we focus on personalized goal tracking, automatic rebalancing, and a transparent fee structure of 0.25% annually with no trading commissions.

Would you like to learn more about how our accounts work?
  • Connect this Message block so it triggers when a prospect asks about competitors or comparisons during the conversation.
  • Add a third Message block for contact detail refusal. Paste in the following response:
I completely respect that. Your privacy matters to us.

If you'd prefer, I can share a link to our Getting Started guide so you can explore on your own terms. You're always welcome to reach out when you're ready.
  • Connect this Message block so it triggers when a prospect declines to share their name or email in the lead capture sequence.

What do these fallback flows have in common?

Each response follows the same three-beat pattern: acknowledge the prospect's concern, offer a low-pressure alternative, and leave the door open for them to return. No fallback ends in a dead end or makes the prospect feel rejected.

The competitor response stays neutral. It never criticizes another platform. Instead, it redirects to Apex Wealth's specific strengths and invites the prospect to learn more.

  • Reset the preview and test both new flows.
  • Start the qualification flow and type: What makes you different from Vanguard? to test the competitor comparison path.
  • In a fresh conversation, progress through the qualification to the email capture step. Type: I don't want to share my email to test the contact refusal path.

Both responses should feel helpful and respectful. The competitor response highlights Apex Wealth's strengths. The contact refusal offers an alternative path instead of ending the conversation.

Fallback messages not appearing?

  • Double check that each Message block is connected to the correct trigger point in your flow. The competitor comparison block should be reachable from any point in the conversation. The contact refusal block should branch from the email Capture block.
  • If the bot still follows the main qualification path, check that the fallback connection is wired as an alternative route, not overriding the primary path.

Help me wire my fallback Message blocks to the right trigger points in Voiceflow.

Add human escalation paths

Some situations should never be handled by a bot alone. Compliance-sensitive financial questions, frustrated users, and complex financial scenarios all need a human advisor. A clear escalation path is the fail-safe logic that enterprise fintech employers specifically look for in conversational designs.

  • Drag a new Message block onto the canvas for the human escalation response.
  • Click on the Message block and paste in the following escalation message:
I want to make sure you get the best possible guidance on this. Let me connect you with one of our licensed advisors at Apex Wealth who can help with your specific situation.

You can reach our advisory team at advisor@apexwealth.com or call (555) 123-4567. I'll make sure they have the context from our conversation so you won't need to repeat anything.

Why mention conversation context?

The line about sharing conversation context reassures the prospect that switching to a human will not mean starting over. This reduces friction in the handoff. In a production deployment, this is where an API integration would pass the collected qualification data to a CRM before the handoff.

  • Connect the escalation Message block to the compliance guardrail deflection flow. After Alex explains it cannot give financial advice, route to this escalation block as the next step when the prospect says yes to connecting with an advisor.
  • Also connect the escalation block as an option from the hesitation and contact refusal flows, so a prospect can request a human at any point.

Before you test, think about this: you are about to re-run every failure scenario from Step 4. Do you think all seven will resolve now, or do you expect any gaps to remain?

  • Reset the preview and run through each of the seven failure scenarios from your persona document:
  • Type: Should I invest in stocks or bonds? and confirm Alex deflects with a compliance-safe response.
  • Type: Is now a good time to invest? and confirm Alex does not give market commentary.
  • Type: What should I do with $50,000? and confirm Alex does not suggest an allocation.
  • Type: I'm not sure I'm ready to invest and confirm Alex responds with empathy and a low-pressure alternative.
  • Type: I'm just browsing and confirm Alex acknowledges and offers resources.
  • Type: What makes you different from Vanguard? and confirm Alex stays neutral and highlights Apex Wealth's strengths.
  • Reach the email capture step and type: I don't want to share my email and confirm Alex respects the boundary and offers an alternative.

Every scenario should now resolve without a compliance violation, a dead end, or a conversation breakdown. No failure scenario from Step 4 should remain unresolved.

Some scenarios still failing?

  • If compliance questions are still getting answered, check that the #Guardrails section in your system prompt uses the exact heading format and lists specific prohibited behaviors.
  • If objection flows are not triggering, verify each Message block is wired as an alternative path from the correct point in the qualification flow.
  • If the escalation path is not reachable, confirm you connected the escalation Message block after the compliance deflection response and as an option from the hesitation and refusal flows.

Help me fix the remaining failure scenarios in my Voiceflow qualification bot.

Secret mission

Create a Production Translation Document

Voiceflow is where you designed and tested your qualification agent. But fintech companies deploy on enterprise platforms like Dialogflow CX and Amazon Lex. Can you create a production translation guide that maps your entire design to these platforms and demonstrates you understand the full pipeline from prototype to deployment?

Clean Up Your Resources

Clean Up Your Resources

This project runs entirely on Voiceflow's free tier and a local text document. There are no ongoing costs. Decide whether to keep your resources as they are, share your prototype for your portfolio, or delete everything.

Resources you used:

  • Voiceflow project "Apex Wealth Qualifier" (free tier, no ongoing cost).
  • Knowledge Base document uploaded to the Voiceflow project (stored within the project, no separate cost).
  • Persona document with failure modes and production translation guide (local file on your computer).

Keep everything running

No action needed. Choose this option if you plan to continue iterating on your Apex Wealth Qualifier bot or want to keep it available for testing.

  • Your Voiceflow project stays in your workspace at no cost on the free tier.
  • Your persona document stays wherever you saved it on your computer or in Google Docs.
  • You can share a prototype link so hiring managers can interact with your bot directly. Messages sent through shareable prototypes do not consume credits.

Pause - I'll come back to this later

Export your project for safekeeping so you can re-import it later if you remove it from your workspace.

  • Return to the Voiceflow workspace dashboard.
  • Hover over the Apex Wealth Qualifier project card and click the three-dot menu.
  • Select the export option to download a .vf file.
  • Save the .vf file alongside your persona document for easy access later.

Delete - I don't want to use this again

Remove all project resources. This cannot be undone unless you exported a .vf backup first.

  • Return to the Voiceflow workspace dashboard.
  • Hover over the Apex Wealth Qualifier project card and click the three-dot menu.
  • Select Delete and confirm when prompted. This removes the project and all its data, including the Knowledge Base document you uploaded.
  • Delete your local persona document if you no longer need it. Move it to the Trash on your computer.
  • Delete your local Apex Wealth product knowledge document if you saved a copy outside Voiceflow.

Nice Work!

Nice Work!

You just designed a complete lead qualification agent for a fintech company. This is a fundamentally different design challenge from customer support, and your portfolio now proves you can handle both.

What you learned:

  • Built a goal-driven qualification flow that collects financial goals, risk tolerance, timeline, and investable amount through natural conversation, routes qualified prospects to lead capture, and answers product questions mid-flow from a Knowledge Base without losing the qualification thread.
  • Designed a trust-building bot persona with explicit voice, tone, and compliance guardrails that prevent the bot from giving financial advice and route sensitive questions to human advisors.
  • Deliberately broke the bot to document failure modes as portfolio artifacts, then built fallback and recovery flows for objections, ambiguity, and off-topic queries with empathy-first responses.
  • Secret Mission: Created a production translation document mapping your Voiceflow design to Dialogflow CX and Amazon Lex, demonstrating awareness of the enterprise platforms fintech companies actually deploy on.

Ready to quiz yourself?