Guides
A stage-by-stage blueprint for deploying AI voice agents across your inbound lead conversion funnel — from sub-60-second speed-to-lead calls through qualification, nurturing, booking, and warm handoffs — using real Thoughtly product mechanics.
Last updated
Most revenue teams lose leads not at one single point but across every stage of the funnel — slow first-call response, inconsistent qualification, dropped follow-ups, and silent handoffs between channels. AI voiceAI voiceAn artificially generated, natural-sounding voice produced by a TTS model. Thoughtly supports a library of AI voices and brand-specific cloning. agents don't just plug the leak at the top; they can actively work every stage from first contact to booked meeting, if you wire them into your workflowWorkflowAn automated, multi-step process — usually triggered by an event (form fill, new lead) and orchestrating one or more voice / SMS / email actions. deliberately.
This guide walks through how to deploy Thoughtly AI voice agents at each stage of an inbound lead conversionInbound lead conversionThe process of turning opted-in inquiries, form fills, calls, and quote requests into qualified conversations, appointments, or transfers. funnel — speed-to-lead, qualification, nurturing, booking, and handoff — using real product mechanics, not abstract theory. By the end, you'll have a stage-by-stage blueprint for wiring AI agents into your existing CRMCRMThe system of record for leads, contacts, deals, and activity. Thoughtly reads from and writes to your CRM continuously., calendar, and communication stack.
Before implementing the stages below, make sure you have the following in place:
If you're starting from scratch, the Thoughtly Agent Builder quick start walks through creating your first agent in about ten minutes.
The top of your funnel is where most revenue leaks. Industry data consistently shows that contacting a lead within the first minute of form submission dramatically increases conversion odds. Human teams rarely achieve this consistently — especially after hours, on weekends, or during peak volume.
Thoughtly's approach is to triggerTriggerThe event or condition that starts an automated workflow, such as a new lead, missed call, CRM status change, calendar booking, or completed call. an outbound voice call the moment a lead enters your CRM or form system. The goal is sub-60-second speed-to-lead: the agent is dialing before the prospect closes the tab.
A typical webhookWebhookAn event-based integration that sends data from one system to another when something happens, such as a form submission, booked appointment, or completed call. payload from a form submission might look like this:
{
"event_id": "lead-abc-123",
"contact": {
"first_name": "Maria",
"phone": "+18005551234",
"email": "[email protected]"
},
"intent": "insurance_quote",
"source": "google_ads"
}Inside the agent, reference this metadata using variables. For example, if the metadata includes an intent field, the agent's Start Node prompt can branch: a lead with intent "insurance_quote" gets a quote-focused opening, while intent "appointment_request" goes straight to scheduling.
Thoughtly's Revenue Autopilot product takes this a step further — it orchestrates voice, SMS, and email follow-upEmail follow-upEmail follow-up is the process of sending timely, context-aware replies or reminders that keep an inbound lead moving toward qualification, scheduling, or handoff. automatically so that if the first call goes unanswered, the agent pivots to SMS, then email, without any manual intervention.
Once the agent reaches the lead, the next job is qualification. The goal isn't to interrogate — it's to determine whether this lead is worth a human rep's time and, if so, to hand off with full context.
Thoughtly handles qualification through a combination of variables, outcomes, and node-based branching. Here's how to structure it.
In the Agent Builder, create variables for the qualification signals that matter for your vertical. For consumer lead conversion funnels — insurance, mortgage, real estate, healthcare, education, home services — these typically include:
The agent extracts these through natural conversation in Speak Nodes. Each Speak Node can use Prompt mode to ask open-ended questions, and variables capture the extracted values.
After each Speak Node, add rule-based outcomes that route the conversation based on what the lead said. For example:
Rule-based outcomes are recommended over prompt-based outcomes after actions or variableVariableA named value the voice agent stores during a conversation — caller name, intent, qualifying answers — and uses to drive routing and post-call actions. extraction, because they evaluate deterministically. The caller doesn't need to speak again before the branch fires.
If your CRM is connected, add actions to Speak Nodes to look up the lead's existing record mid-call. For example, a HubSpot action can check whether this contact already has an open deal, what their lifecycle stage is, and whether they've been contacted before. This lets the agent say: "I see you requested a quote last month — are you still looking?" instead of starting from scratch.
After the action runs, use rule-based outcomes to branch: if the CRM lookup returns an existing open deal, the agent can take a different path than for a brand-new lead.
Not every lead is ready to book on the first call. Most aren't. The typical funnel loses 70–90% of leads between initial contact and conversion — not because they're uninterested, but because no one follows up consistently.
Thoughtly's omnichannelOmnichannelA coordinated customer journey across channels such as voice, SMS, email, web forms, and CRM tasks, where context carries across each interaction. agent model lets a single agent continue the conversation across voice, SMS, and email — with the same contact context, variables, and outcome history. This means a lead who says "call me next week" on Tuesday gets a text Wednesday, an email Thursday, and a callback Friday, all without a human touching the workflow.
Because Thoughtly agents share the same model across channels, a lead who replies to an SMS two days after the initial call is picked up by the same agent with the same conversation history — not a cold restart.
Dark windows and quiet hoursQuiet hoursTime windows when outbound calls or texts should not be sent, based on legal rules, customer preferences, or business policy. ensure follow-up never goes out at 3am or outside compliance-approved calling windows. Configure these in Settings before scaling outbound nurture sequences.
When a lead is qualified and ready, the agent should book the meeting before the call ends. Every minute between "yes, I want to talk" and a confirmed calendar invite is a minute the lead can go cold.
Thoughtly supports scheduling integrations — Calendly, Cal.com, Acuity, Zoho Bookings, and others — as mid-call actions. The agent can check real-time availability, propose a time, and confirm the booking while the lead is still on the line.
This pattern — action → variable → rule-based outcome — is the same one used for CRM lookups and validations. It keeps the flow deterministic and avoids relying on the AI to infer whether a booking succeeded.
Some leads need a human. The question is whether your agent hands off a warm, contextual lead or a cold transfer that forces the prospect to repeat everything they just said.
Thoughtly's Transfer Node supports warm transfers, and Early Summaries can generate a context summary before the call transfers. The receiving rep hears a brief summary — name, reason for calling, qualification details, what was discussed — and picks up the conversation mid-stream.
For after-hours leads, the agent can book a callback instead of transferring. The lead gets a confirmed time on the calendar; the rep gets a CRM task with full context. Nobody waits.
An AI voice agent that calls once and gives up is just an expensive autodialer. The real conversion lift comes from persistent, multi-channel follow-upMulti-channel follow-upA follow-up motion that uses more than one channel — usually voice, SMS, and email — to improve contact rates after a prospect raises their hand. over days or weeks. Build nurture sequences, not single calls.
After a mid-call action (CRM lookup, booking, validation), the caller doesn't speak before outcomes evaluate. If you use prompt-based outcomes, the AI can't see internal action results unless you surface them verbally. Use rule-based outcomes that check action variables directly — for example, lookup_found == true or booking_status == "confirmed".
If call outcomes, dispositions, and variables aren't written back to your CRM, your reps are flying blind. Set up an On Call Completed automation that updates the contact record with the call result, tags, and next steps — every time, for every agent.
Consent, suppression lists, quiet hours, and DNC scrubbingDNC scrubbingFiltering outbound dialing lists against federal and internal Do-Not-Call registries. Required for compliant outbound — Thoughtly scrubs every call. are not optional add-ons. Configure them before you scale outbound. Thoughtly supports consent mode, suppression lists, and dark windows natively — turn them on before your first campaign, not after your first complaint.
The Start Node should be simple: greet, identify the agent, and route into the conversation. Don't try to simulate voicemail detectionVoicemail detectionVoicemail detection is the ability to identify when a call reaches a voicemail greeting instead of a live person, then trigger the right message, callback, or alternate-channel follow-up., run complex qualification logic, or execute CRM lookups in the first node. Keep it clean and let subsequent nodes do the work.
To know whether your AI-optimized funnel is working, track metrics at each stage — not just overall conversion. Here's what to watch:
| Funnel stage | Key metrics | Where to find them |
|---|---|---|
| Speed-to-lead | Time from lead to first call attempt, call connection rate, % of leads called in < 60s | Automations run log, History filter by agent and date |
| Qualification | Qualified rate (% of calls marked qualified), DNQ rate, avg qualification questions answered | History outcomes, Disposition tags, Analytics |
| Nurture | Response rate on follow-up SMS/email, time to first response, opt-out rate | History filter by channel, Audiences attributes |
| Booking | Booking rate (% of qualified leads with a confirmed appointment), no-show rate | Scheduling integration, CRM deal stage, On Call Completed automation outputs |
| Handoff | Transfer completion rate, rep satisfaction with lead quality, time from transfer to conversation | History call status (Transferred), CRM activity log |
| Overall | Total pipeline booked, cost per booked meeting, agent vs. human lift | Analytics dashboard, CRM revenue attribution |
Thoughtly's Analytics page gives you a high-level view of responses, talk time, usage by agent, and credit consumption. For deeper analysis, filter History by agent, outcome, or date range, and export results as CSV. Pair this with your CRM's revenue attribution to measure the number that matters: pipeline booked from agent-worked leads.
Yes. Thoughtly agents are omnichannel — the same agent can handle inbound calls, make outbound calls, send and receive SMS, and send email. The conversation logic, variables, and outcomes are shared across channels. Just make sure prompts are appropriate for each channel — a voice greeting may need to be shortened for SMS.
Configure dark windows in Settings to block outbound calls, SMS, and email during quiet hours. Thoughtly respects contact time zones when configured. For nurture sequences, use Conditions in your automation to skip contacts whose local time is outside your approved calling window.
Thoughtly supports webhook-based integrations for any CRM. Use the Send Webhook step in Automations to push call outcomes to your CRM's API, and use the Incoming Webhook trigger to start automations from CRM events. If your CRM has a Zapier integration, you can also connect through Thoughtly's Zapier integration.
Thoughtly agents scale concurrently — there's no hard cap on simultaneous calls per agent. The practical limit is your plan's credit allocation and phone number concurrency. For high-volume campaigns, use outbound batch calling with a recurring schedule trigger and monitor usage in Analytics.
Yes. Thoughtly agents identify themselves as AI at the start of every call by default. Transparency builds trust and is required for compliance in many jurisdictions. Most teams report higher NPS from agent-handled conversations than from cold human outreach, partly because the agent is consistent, patient, and never has a bad day.
Thoughtly Revenue Autopilot — how autonomous multi-channel follow-up works.
Thoughtly Speed-to-Lead solution — sub-60-second lead response for high-volume funnels.
Thoughtly Analytics — funnel-by-source attribution and agent vs. human lift measurement.
Thoughtly docs: Automations — triggers, actions, and post-call workflows.
Thoughtly docs: Actions — mid-call CRM lookups, scheduling, and integrations.
Thoughtly docs: Outcomes and branching — rule-based vs. prompt-based routing.
How to Build a Speed-to-Lead AI Agent with Thoughtly — detailed setup guide for the top-of-funnel stage.
How to Qualify Inbound Leads Over Voice and SMS with Thoughtly — qualification patterns and branching logic.
How to Build a Multi-Channel Follow-Up Sequence with Thoughtly — nurture sequence design across voice, SMS, and email.
How to Use Thoughtly's Revenue Autopilot for Automated Pipeline — end-to-end autonomous pipeline generation.
How to Measure AI Voice Agent Performance: A Complete Framework — metrics and KPIs for each funnel stage.