AI lead scoring and routing: send every lead to the right rep

Turn voice agent qualification answers into clear categories and route hot leads to reps, warm leads to sales, and support requests to care teams.

## How AI lead scoring directs contacts to the right destination AI lead scoring turns spoken conversations into structured data points so you can automatically route leads to sales reps without manual triage. When an inbound or outbound call finishes, the voice agent checks the responses against your qualification criteria and assigns a category or status field. Your business rules then send the contact to the correct person, calendar, or automated sequence immediately. Depending on the collected fields, each contact lands in one of four places: * Live phone transfer: Immediate handoff when configured, transferring to a person or number the company chooses when a prospect meets your criteria for urgent demand. * Direct calendar booking: An appointment placed on a connected calendar according to geography, product match, or availability. * Sales queue: A CRM record assigned to an inside sales rep when the contact is interested but requires research or manual follow-up. * Nurture sequence or closure: Automated marketing emails for early-stage inquiries, customer care routing for support requests, or archive status for unqualified contacts. Every voice call on DeepAgent starts with a statement confirming that the caller is speaking with an AI, as EU AI Act Art. 50 requires. What data you store and for how long is something to agree with your DPO (see our notes on GDPR requirements). ## From answers to a score The voice agent asks your qualification questions, records each answer into a CRM property, and categorizes the buyer using explicit rules. You decide which criteria matter, what weights to apply, and which answers count as a passing grade. | Evaluation Criterion | Agent Question Focus | Recorded Field Name | Field Value Meaning | | :--- | :--- | :--- | :--- | | Need | Primary problem or feature requested | `lead_primary_need` | Specific product line, pricing request, or technical support | | Budget range | Approved spending level | `budget_tier` | Under minimum threshold, standard tier, enterprise tier | | Timing | Expected implementation date | `purchase_timeline` | Immediate (0-30 days), this quarter, exploratory (6+ months) | | Decision role | Buyer's authority in the purchase | `buyer_authority` | Sole decision maker, evaluation committee, end user | | Fit | Current infrastructure or company size | `technical_fit` | Compatible, needs custom integration, unsupported stack | Using BANT qualification before routing (Budget, Authority, Need, Timing) keeps sales teams focused on prospects who have the budget and timeline to buy. The agent writes these answers into the contact profile as soon as the call ends. ## Routing rules for sales and operations Once fields are recorded, lead qualification and routing rules determine the path of the record. You can combine several criteria to build precise workflows: * Routing by score or category: Prospects matching high-priority criteria are flagged as hot for instant attention. Inquiries with lower scores go to an SDR queue, while unqualified contacts close out automatically. * Routing by territory or location: The agent asks for the prospect's city, postal code, or branch preference. For example, Bludental manages patient volume across 70+ locations. * Routing by product line: B2B organizations use the identified need field to assign leads to specialized product reps rather than generalists. * Routing by contact type: Inbound channels handle mixed traffic. For example, Amilon used a DeepAgent web widget to classify 156 inbound interactions across 30 days as commercial, support, or general requests, routing each contact directly to either sales or customer care. * Routing by transfer rules: When configured, the agent can initiate a live transfer to a person or number the company chooses. In a deployment at Agenzia Unipol Cerignola, the AI assistant transferred 200 calls out of 207 attempts directly to agency staff, reaching a 96.6% transfer completion rate. * Warm contact handling: An Italian manufacturing company used the voice agent to screen 2,000 inbound industrial leads, handing only warm contacts to sales reps while updating HubSpot records automatically. ## Where the rules live Lead routing logic operates in two separate locations, depending on whether the action happens during the conversation or after the call hangs up. ### 1. In the CRM or automation platform Post-call rules live in your core database. DeepAgent sends the collected answers, transcripts, audio recordings, and call statuses to your CRM via API or webhook. In our HubSpot CRM integration guide, you configure HubSpot lead routing through standard workflows. When the field `lead_score_tier` updates to `Hot`, HubSpot assigns an owner using round-robin distribution, creates a deal, and alerts the salesperson via Slack or email. Teams managing mixed stacks can read our guide on voice agent CRM integration for HubSpot and Salesforce to align their custom objects. ### 2. In the voice agent instructions Mid-call rules live inside the agent's system prompt. When a prospect provides answers that meet your threshold for immediate handoff, the agent executes a transfer command to a target phone number or SIP address. Our guide to AI voice agent human handoff and call transfers details how to pass context to the human rep before the line connects. Keep both environments consistent. If your CRM treats a 60-day timeline as warm, the voice agent prompt must not treat that same 60-day answer as an emergency live transfer. ## Real cases of automated scoring and routing The following implementations show how organizations configure qualification questions and lead routing rules. These numbers reflect the exact outcomes achieved in each project. | Organization | What the Agent Collects | Where the Lead Goes | Measured Results | | :--- | :--- | :--- | :--- | | Bludental | Budget and procedure urgency across 70+ locations | WhatsApp reminder, dental clinic booking | 10,000 monthly leads; 65% answer rate vs 50% human; 28% end-to-end conversion vs 22% human; 59 of 100 scored, 28 book an appointment | | Italian Manufacturing Company | Project interest, technical requirements, company fit | Warm contacts handed to sales; data written to HubSpot | 2,000 inbound leads worked; 70%+ reduction in handling time; up to 5 attempts per lead; outcomes written 100% automatically | | VEGA (DeepAgent Inbound) | Voice context, call volumes, channel needs, urgency | Profiled records passed to sales reps via CRM | 7,290 attempts in 3 months; 3,030 completed conversations; 1,510 positive outcomes passed to sales; useful outcome every 3.5 minutes | | Amilon | Contact type (commercial, support, or general inquiry) | Sales team for commercial; customer care for support | 156 inbound interactions in 30 days; 137 text chats (87.82%) and 19 voice calls (12.18%); 637.35 total interaction minutes | | Agenzia Unipol Cerignola | Reason for call (information, quote, claim, appointment, callback) | Live agency staff transfer, calendar, or info close | 725 calls in 2 months; 200 of 207 transfers successful (96.6%); 90 commercial opportunities; requests handled correctly rose from 63.4% to 93.2% | ## How to check that routing works Verify regularly that your logic operates correctly across your software stack. Our framework to measure pipeline from voice agent calls in RevOps provides methods for tracking conversions across the customer journey. Check these four points in your DeepAgent dashboard and CRM logs weekly: 1. **Outcome distribution**: Look at the breakdown of call outcomes. If almost all calls land in the unqualified category, your scoring criteria may be too restrictive. If almost all are marked hot, reps will face qualification fatigue. 2. **Failed transfers**: Review call logs for transfer attempts that went unanswered or dropped. Identify whether internal lines were busy or if team schedules were configured incorrectly. 3. **Unassigned CRM records**: Filter contacts in HubSpot or your database for records created by the integration that have no designated owner. Any unassigned record indicates a gap in your workflow criteria. 4. **Time to first sales contact**: Measure the duration between when the agent logs a warm contact and when a sales rep makes contact. Tracking this metric helps ensure that sales reps reach leads while interest is fresh. ## Common routing mistakes to avoid Review our guide to automatic lead qualification with voice agents to understand call pacing. Avoid these common operational pitfalls: * Asking too many questions: Forcing a caller through ten screening questions lowers completion rates. Stick to three or four qualifying criteria. * Collecting unused criteria: If your sales team does not segment prospects by tech stack, do not spend call time collecting tech stack details. Score only what dictates a routing decision. * Generating scores without an action: A tag marked `Tier 2` is useless if it sits in a database without triggering a task, notification, or owner assignment. * Failing to handle callbacks: When prospects ask for a callback later, marking them as closed loses pipeline. Build rules that schedule a recall attempt as outlined in our speed-to-lead voice strategy. Teams capturing paid traffic can consult our guide on calling Facebook lead ads with voice agents to trigger calls immediately. You can also explore WhatsApp lead qualification with voice follow-ups, connect an inbound chat and voice widget on your website, or review our guide to marketing automation workflows for voice agents for multi-channel setup. ## An illustrative qualification and routing call This fictional example shows how an agent identifies buyer criteria, discloses its automated identity, and sets expectations for routing. > **Agent**: Hello, this is the virtual assistant for Apex Cloud Solutions. I am an automated AI assistant. I am following up on your request regarding our security software. Do you have two minutes to see if we are a good fit? > > **Prospect**: Yes, I have a couple of minutes. > > **Agent**: Great. To route you to the correct specialist, how many endpoints does your team currently manage across your network? > > **Prospect**: We have roughly four hundred endpoints across two regional offices. > > **Agent**: Thank you. And are you looking to replace your current system this month, or are you in an early research phase? > > **Prospect**: We need a solution running before our annual audit next month. It is quite urgent. > > **Agent**: Understood. Because your project requires rapid deployment for four hundred endpoints, I am going to connect you directly with our senior infrastructure advisor. Please hold while I transfer the line. ## How to start setting up lead scoring and routing 1. Select your core qualification factors: Choose three or four questions covering project scope, urgency, company profile, and purchasing role. 2. Define outcome buckets: Establish clear categories such as Hot (live transfer or direct booking), Warm (SDR queue), and Support/Nurture. 3. Configure CRM custom fields: Build matching properties in HubSpot or your primary CRM to receive the data points. 4. Draft the agent script: Enter your qualification questions into DeepAgent and configure the mandatory AI disclosure sentence. 5. Set transfer rules: Add the transfer phone numbers or SIP endpoints in DeepAgent for accounts meeting your immediate handoff criteria. 6. Build CRM assignment workflows: Create workflows triggered by the incoming webhook data to assign owners, create tasks, and notify reps. 7. Run test calls: Place five test calls answering with different combinations of budget, timing, and needs to verify that every contact routes to the intended owner. ## DeepAgent pricing DeepAgent offers monthly and annual subscription tiers: * Starter: EUR 99/month * Pro: EUR 299/month * Growth: EUR 500/month * Scale: EUR 799/month * Enterprise: Custom quote Prices exclude VAT. Selecting annual billing saves 17% across all standard plans. 30 minutes free, no card. Visit our pricing page for complete plan details. ## Frequently asked questions ### What is lead scoring with AI? Lead scoring with AI uses a phone or chat agent to ask your qualification questions, store the prospect answers as structured fields, and sort contacts into tiers such as hot, warm, or cold based on your rules. ### Can an AI voice agent route leads to the right salesperson? Yes. The voice agent collects details like location, product line, or urgency. Your CRM or phone system then uses those data points to transfer the live call or assign the record to the correct account executive. ### Does the voice agent work with HubSpot lead routing? Yes. DeepAgent writes qualification fields, call outcomes, transcripts, and audio links directly into HubSpot contact and deal properties via API or webhook, allowing native HubSpot workflows to assign owners. ### What questions should the agent ask before routing to sales? You configure the questions based on your criteria, such as project timeline, budget range, role, and current software. BANT qualification before routing helps filter out contacts who are not ready for a rep. ### What happens to leads that are not ready? Leads that lack budget, authority, or immediate need are categorized as warm or cold. The system writes this status to your CRM, which sends them to an automated email nurture sequence or marks them closed. ## Automate your inbound and outbound qualification Talk to sales → · Create your agent →