Inbound Lead Qualification Agent
Turn website visitors into qualified pipeline, at the moment they're most interested.
A buyer on your pricing page is at peak intent. They have a question, a deadline and a shortlist. A contact form asks them to wait. A BoldAgent qualification agent answers them on the spot, learns whether they're a fit, puts a meeting in a rep's calendar and leaves a complete record in your CRM, all inside one chat.
Peak intent
Why the pricing page is where deals are won or lost
Buyers rarely fill in a form on their first visit. They read the pricing table, check whether one feature is included, look for a security page, and leave to compare alternatives. The questions that decide whether they come back are small and specific:
Answer those well, while the buyer is still on the page, and the conversation naturally moves to qualification and a meeting. That is the job this agent is designed for.
- Is SSO on the Business plan, or only Enterprise?
- Do you integrate with the tool we already use?
- Can we start with 20 seats and grow to 200?
One agent, four capabilities
A single, focused agent from first question to booked meeting
Unlike multi-agent setups, this use case works best as a single, focused agent. Every turn of a sales conversation needs the same context, and a single agent keeps the tone consistent from first question to booked meeting.
-
Knowledge library
Product, pricing, integration and security answers, kept grounded in your published content
-
Qualification skill
Your framework for what to learn about a buyer and how to ask for it conversationally
-
Lead Fit decision model
A consistent score and routing decision for every lead
-
API Connections
Your CRM and booking system: find accounts, create leads, book meetings
Inside a conversation
What the visitor sees, and what the agent is doing
What the visitor sees on the left; what the agent is doing on the right.
- Visitor Does the Business plan include SSO? We're about 200 people.
The page URL tells the agent the visitor is on /pricing. It searches the library and finds SSO in the Business plan feature list.
- Agent Yes, SAML SSO is included in Business. For 200 people, is this for a security review, or a wider rollout?
The answer is grounded. The follow-up question comes from the qualification skill, which prefers learning the buyer's need over asking for a job title first.
- Visitor Security review this quarter. I'm the IT manager. We're on spreadsheets today.
Need, timeline, role and current tool are now known. The agent calls Lead Fit.
-
Lead Fit returns
segment: mid_market,fit: 4,ready_for_sales: true. -
Find account in the CRM returns no match, so this is a new business lead for the mid-market team.
- Agent Our mid-market team works with companies your size on security reviews. Priya has time Thursday at 2:00 or Friday at 10:00. Would either work?
Get availability returns open slots for the assigned rep.
- Visitor Thursday works.
Book meeting, then Create lead with the structured record below.
- Agent Done. Priya will have your SSO requirement and timeline, so you won't have to repeat yourself.
The conversation is saved with the agent version that handled it.
The qualification model
Every lead judged by the same questions
Lead Fit is a decision model, so every lead is judged by the same questions, in the same order, with the same scales. Your reps stop arguing about what "qualified" means.
| Key | Type | Scale and guidance |
|---|---|---|
segment | Choice | self_serve · smb · mid_market · enterprise, based on company size and requirements |
fit | Score | 1 = outside your ideal customer profile · 2 = possible fit, key need unclear · 3 = good fit, no urgency · 4 = strong fit with a clear need and timeline |
ready_for_sales | Bool | True when the fit is 3 or higher and the buyer has a defined need and timeline |
requires_approval | Bool | True when the buyer asks for discounts, custom contract terms or legal commitments |
Use the decision model's test panel to replay real (anonymised) chats and check that the scores match what your best rep would have decided. Adjust the guidance text, not the agent's instructions, when they don't.
Page-aware conversations
The same agent, tuned to where the buyer is
The current page URL is passed into every run, so the same agent behaves differently depending on where the buyer is.
- /pricing
Pricing page
Plan differences, seat counts and what's included. Moves toward qualification once a question is answered.
- /security
Security or trust page
Certifications, data handling and SSO. Offers the security documentation before asking anything about the buyer.
- /integrations
Integration pages
Confirms the specific integration and how it's set up. Asks what else the buyer needs to connect.
- /blog
Blog and guides
Helps with the topic first. Does not push for a meeting unless the visitor asks about buying.
- via the Developer API
Inside your product
Knows the user's current plan and usage from context. Focuses on what an upgrade would unlock.
From conversation to CRM
The same fields, in the same format, every time
The agent uses structured output to produce the lead record, so the CRM always receives the same fields in the same format, regardless of how the chat went.
{
"company_size": 200,
"segment": "mid_market",
"fit": 4,
"role": "IT Manager",
"need": "SSO for security review",
"timeline": "This quarter",
"current_tool": "Spreadsheets",
"meeting_booked": true,
"summary": "Evaluating Business plan; SSO is required for a security review this quarter."
}Routing follows from the score:
- Fit 4, ready for sales Meeting booked with the segment's team; a webhook can alert the rep straight away
- Fit 3 Meeting offered; if declined, the lead is created for follow-up
- Fit 1–2 Pointed to a free trial or self-serve resources; lead created for marketing nurture
- Existing customer Routed to their account manager instead of new business
- Requires approval Conversation paused for a sales manager
Where the agent stops
Lines it won't cross on its own
Sales conversations carry commercial and legal risk. These are the lines this agent is configured not to cross on its own.
-
No discounts or custom terms.
When
requires_approvalis true, human approval pauses the chat. A sales manager can reply directly, approve a specific offer, or decline. -
No claims beyond your content.
Grounded responses mean the agent won't promise a feature, certification or roadmap date that isn't in your library. Unanswerable questions get a fallback offering a call with the team.
-
No disparaging competitors.
The qualification skill states how to handle comparison questions: describe your own strengths, and offer a rep for detailed comparisons.
-
No being argued off-script.
A Responsible AI policy with prompt injection detection stops visitors from instructing the agent to "ignore your pricing rules".
Tuning it for conversion
Small changes, measurable gains
-
Mine the questions buyers can't get answered.
Knowledge Gaps groups the pre-sales questions your content didn't cover. A cluster of questions about data residency, for example, is a signal to write a page about it, which helps your SEO as well as your agent.
-
Test with buyer personas, not just facts.
Build evaluation datasets for each persona: an enterprise security lead, a small-business owner, a student who will never buy. Check the answer quality and the routing decision for each.
-
Keep the widget fast.
Visitors abandon slow chats. Use Analytics to watch P95 latency for this agent, and lower reasoning effort or retrieved knowledge if responses start to lag.
-
Compare versions.
Every conversation records the agent version that handled it, so when you change the qualification skill you can compare booked meetings before and after.
A sensible rollout
Four phases, each one reversible
- Phase 1
Answer only
Launch on the pricing page with the library and grounded responses, but no qualification or booking. Watch Knowledge Gaps and fill the holes.
- Phase 2
Qualify
Add the qualification skill and Lead Fit model. Create leads in the CRM, but let reps book meetings manually while you check the scores.
- Phase 3
Book
Turn on the booking action once scores match your reps' judgement.
- Phase 4
Expand
Add the security, integration and in-product placements, each with its own page-aware behaviour.
Each phase is a new published version, so you can step back to the previous one if results dip.
Get started
Five steps to your first qualified lead
- Create a library from your pricing, product and security pages.
- Build the Lead Fit model and test it with real chats.
- Connect your CRM and booking system.
- Return structured output for your lead records.
- Embed the chat widget on your pricing page.