Get started with the Sales Qualification Agent
The Sales Qualification Agent (SQA) is an AI agent that automates the most time-consuming parts of lead qualification. Rather than waiting for a seller to research each lead manually, the agent takes over as soon as a lead is assigned to it.
The agent works through a research loop for every lead it processes:
- When a lead matches the selection criteria you configure (for example, leads with source "Web form" rated "Hot"), the lead gets automatically assigned to the agent.
- The agent researches the lead using public web data, existing Dataverse records, and any knowledge sources configured by your organization.
- The agent evaluates how well the lead fits your target customer profile (TCP).
- Based on that evaluation, the agent either hands the lead to a seller with full research insights, or disqualifies it and notifies a supervisor.
The agent doesn't replace seller judgment on handed-over leads; it removes the manual research overhead so sellers can focus on high-value conversations rather than background investigation.
Research-only and Research and engage modes
The Sales Qualification Agent is available in two modes. The mode you deploy determines how much of the lead qualification process the agent handles autonomously.
| Capability | Research-only | Research and engage |
|---|---|---|
| Research leads | ✅ | ✅ |
| Check target customer profile criteria | ✅ | ✅ |
| Check BANT criteria | ❌ | ✅ |
| Generate outreach emails | ✅ | ✅ |
| Send outreach emails | ❌ | ✅ |
| Detect positive intent based on responses | ❌ | ✅ |
| Send follow-up emails and clarify questions | ❌ | ✅ |
| Hand over promising leads to sellers | ✅ | ✅ |
| Notify supervisors about disqualified leads | ✅ | ✅ |
Research-only mode automates the research phase. The agent researches assigned leads, evaluates them against your TCP, generates a draft outreach email, and hands the lead to a seller. The seller reviews the research, decides whether to send the email, and makes the final qualify or disqualify decision.
Research and engage mode takes automation further. The agent researches leads, sends outreach emails, follows up based on lead responses, and evaluates purchase intent using the BANT framework. Only leads that demonstrate positive intent and meet your handoff criteria are passed to sellers—with the full email thread, research insights, and BANT evaluation already documented on the lead record.
How the agent researches leads
To build a comprehensive picture of each lead, the agent draws on three sources of information:
- Bing: The agent searches the public web for up-to-date information about the lead's company. This includes company overviews, recent news, financial summaries, and industry context.
- Dataverse: The agent checks your Dynamics 365 environment for existing account and contact records associated with the lead, drawing on relationship history and prior engagement data.
- Knowledge sources: Administrators configure custom knowledge sources, such as SharePoint documents and public URLs, to provide the agent with internal context. Common examples include competitor battle cards, product positioning briefs, outreach templates, and follow-up email guides.
The agent is built on Copilot Studio and uses a set of specialized subagents for different research tasks. A competitor research agent identifies and analyzes key competitors. A stakeholder research agent investigates the individual lead contact. An email validation agent checks whether the lead's email address is reachable before engagement begins.
These subagents work together to surface a complete, structured view of each lead—the kind of research that would otherwise take a seller 30 to 60 minutes to gather manually.
Evaluation frameworks
Two evaluation frameworks shape how the agent qualifies leads: the target customer profile and BANT.
Target customer profile (TCP) defines the characteristics of your ideal customer, like industry, company size, geography, and revenue range. You configure the TCP when setting up the agent, and it applies in both modes. The agent uses the TCP to score each lead and generate a fit assessment.
Defining a clear TCP is critical. If your TCP is too broad, the agent hands over large numbers of low-quality leads, overwhelming sellers and defeating the purpose of automation. If it's too narrow, the agent disqualifies leads that sellers would have pursued. A well-calibrated TCP is the single most important factor in determining agent effectiveness.
Tip
Start with fewer than 25 TCP criteria for initial calibration. Begin with 10–15 high-confidence criteria—such as industry, company size, and geography—and expand only after reviewing the first 2–4 weeks of agent performance data. Over-specifying TCP criteria from day one is the most common deployment mistake.
Budget, Authority, Need, Timeline (BANT) is a qualification framework used only in Research and engage mode. After the agent engages with a lead through email, it evaluates the lead's responses against BANT criteria to assess purchase readiness. A lead that demonstrates positive buying intent and meets your TCP and BANT criteria is handed over to a seller with a full qualification summary on the lead record.
One mode per organization
Before you deploy the agent, understand a key constraint: your organization can only run one Sales Qualification Agent at a time. You can't operate Research-only mode for one team and Research and engage mode for another—it's one instance, one mode, one organization.
You can upgrade from Research-only to Research and engage mode later, but you can't downgrade.
Important
Mode selection is a one-way decision. Once you upgrade to Research and engage mode, you can't revert to Research-only mode. Plan your mode selection carefully before you begin configuration, and consider your team's readiness to handle the extra prerequisites for Research and engage mode, including shared mailbox setup, server-side synchronization, and BANT criteria configuration.
In the next unit, you'll walk through the full configuration sequence for the Sales Qualification Agent, starting with prerequisites.