What you're building

An Agentforce SDR agent is an autonomous AI worker inside your Salesforce org that engages inbound leads, qualifies them against your criteria, books discovery calls, and hands off to human AEs — without a rep involved at any step of the process.

This guide covers the exact configuration we use when building SDR agents for clients. By the end, you'll have a working agent that responds to inbound leads in seconds, qualifies with precision, and delivers briefed opportunities to your sales team.

Prerequisites before you start:

  • Salesforce Sales Cloud (Enterprise or Unlimited edition)
  • Agentforce add-on license (or Einstein 1 Sales edition)
  • Agent Builder enabled in Setup
  • A clean Lead object with standard qualification fields (Company, Title, Phone, Annual Revenue, or equivalent custom fields)
  • At least one knowledge article or product overview document for grounding

Step 1: Define your qualification criteria

Before touching Salesforce, get alignment on what "qualified" means for your business. This is not a technical step — it's a business policy step, and skipping it is the most expensive mistake you can make.

Write your qualification criteria as explicit rules, not general principles. Bad: "Good fit companies." Good: "Companies with 50–5,000 employees, using Salesforce Sales Cloud, with a stated pain point around lead response time or pipeline visibility, and an active buying timeline within 12 months."

Your criteria should cover:

  • Firmographic fit: Company size, industry, geography, tech stack
  • Role fit: Is the lead a decision-maker or influencer? What titles qualify?
  • Need: What pain points or business objectives indicate genuine need?
  • Timeline: What buying horizon makes a lead worth pursuing now vs. nurturing?
  • Disqualifiers: What immediately removes a lead? (e.g., existing contract with a competitor, company size too small, wrong industry)

Document this in a qualification scorecard. You'll use it to write the agent's role description and topic instructions in the next steps.

Step 2: Write the agent role description

In Agent Builder (Setup → Agentforce Agents → New Agent), the role description is the most important configuration you'll write. It defines the agent's scope, capabilities, limits, and behavior. Be specific.

Here's a template for an SDR agent role:

"You are [Company]'s AI Sales Development Representative. Your job is to engage inbound leads who have submitted our website contact form and qualify them for a discovery call with one of our account executives.

Qualification criteria: The lead works at a company with 50–5,000 employees, uses Salesforce (any edition), and has a stated challenge related to [your value prop]. They should be open to a 30-minute discovery call within the next 30 days.

You have access to the lead's CRM record, our product knowledge base, and our AE calendar availability. You can update lead fields, book discovery calls, and enroll leads in nurture campaigns. You cannot make promises about pricing, discounts, implementation timelines, or specific ROI outcomes. If a lead asks about pricing, tell them it will be discussed on the discovery call.

If a lead qualifies, book a 30-minute discovery call. If they don't qualify, thank them for their interest and enroll them in the appropriate nurture sequence. If a lead asks to speak with a human, transfer immediately to a live chat rep or create a callback task for the assigned rep."

Notice what this includes: scope, qualification criteria, available capabilities, explicit limits, and escalation conditions. Every element reduces ambiguity and produces more consistent agent behavior.

Step 3: Configure topics

Topics tell the agent what categories of conversations it handles. Each topic has its own instructions and action library. The LLM uses topic descriptions to classify each incoming message — so write them precisely.

For an SDR agent, configure these four topics:

Topic 1: Initial Lead Engagement

Description: "The lead has just submitted the contact form and this is their first interaction. Greet them, confirm their interest, and begin gathering qualification information. Start with their primary challenge or goal, then move to company size and Salesforce usage."

Topic 2: Qualification and Booking

Description: "The lead has provided enough information to make a qualification decision. Evaluate against the qualification criteria. If qualified, offer specific meeting times and book the discovery call. If not qualified, acknowledge their interest and route to the appropriate nurture sequence."

Topic 3: Product and Company Questions

Description: "The lead is asking questions about our services, process, case studies, or company before deciding whether to book a call. Answer using the knowledge base. Do not speculate on pricing or specific outcomes."

Topic 4: Objection Handling

Description: "The lead has expressed hesitation — they're happy with their current solution, not in an active buying cycle, or concerned about switching cost or implementation complexity. Acknowledge their concern, provide relevant context, and offer to send follow-up resources without pressuring for a meeting."

Step 4: Build the action library

Actions are what your agent can actually do. Every action needs a clear name, description, and input/output definition. The LLM reads the description to decide when to invoke each action — vague descriptions cause incorrect invocations.

Action 1: Get Lead Record

  • Description: "Retrieves the lead's existing CRM record including company name, title, phone, annual revenue, and any previous activity. Use this at the start of every conversation to avoid asking for information already in the system."
  • Input: Lead ID (automatically available from the conversation context)
  • Output: Lead fields and recent activity summary

Action 2: Update Lead Qualification Fields

  • Description: "Updates the lead's CRM record with qualification information gathered during the conversation. Use this after each qualifying question is answered to keep the record current."
  • Input: Field name(s), new value(s)
  • Output: Confirmation of update

Action 3: Check AE Calendar Availability

  • Description: "Returns available 30-minute time slots for discovery calls in the next 5 business days, filtered by the assigned AE for this lead's territory. Use this only after the lead has been confirmed as qualified."
  • Input: Territory or assigned rep ID
  • Output: List of available time slots (date, time, timezone)

Action 4: Book Discovery Call

  • Description: "Creates a Salesforce Event for the discovery call, invites the lead and assigned AE, and sends calendar confirmations to both. Use this only when the lead has explicitly agreed to a specific time slot."
  • Input: Lead ID, AE ID, selected date/time, meeting subject
  • Output: Event ID and confirmation message

Action 5: Enroll in Nurture Campaign

  • Description: "Adds the lead to a Salesforce Campaign based on their disqualification reason. Use this when the lead does not meet qualification criteria but has expressed some interest."
  • Input: Lead ID, disqualification reason (maps to campaign)
  • Output: Campaign enrollment confirmation

Action 6: Create Callback Task

  • Description: "Creates a high-priority task for the assigned rep to call the lead manually. Use this for high-value leads who prefer a phone call, or when the lead requests to speak with a human."
  • Input: Lead ID, task priority, notes summary, due date
  • Output: Task creation confirmation

Step 5: Set up the channel

Your SDR agent needs to be reachable where your leads are. For a first deployment, start with web chat — it is the easiest to test and offers the fastest feedback loop.

  1. Go to Setup → Messaging Settings → New Messaging Channel
  2. Select "Enhanced Messaging" and choose Web Chat
  3. Associate the channel with your SDR agent
  4. Configure the pre-chat form to capture Email and Company (the agent uses these to retrieve the CRM record)
  5. Copy the embed code and add it to your contact form confirmation page or your main website header

For follow-on deployments, connect the agent to your email channel so it can engage leads who reply to automated outbound sequences, and to SMS for mobile-first lead engagement.

Step 6: Connect the trigger

Your agent should engage leads automatically — not wait for them to find the chat widget. Set up a Flow trigger that fires the agent conversation when a new Lead record is created from your web form:

  1. Create a Record-Triggered Flow on the Lead object (trigger: Created)
  2. Add a condition: LeadSource = "Web" (or whichever source your form uses)
  3. Add an action: "Send Agentforce Message" — this initiates the agent conversation via the lead's email address or phone number
  4. Set a 30-second delay before triggering (gives the lead time to land on your confirmation page before the chat appears)

This creates a seamless experience: a lead submits your form, lands on the confirmation page, and within 30 seconds receives a personalized follow-up from your AI SDR — already knowing their name and the reason they reached out.

Step 7: Test with your 20 hardest leads

Before going live, run the agent through your 20 most challenging lead scenarios from the last 90 days. Include:

  • 5 clearly qualified leads — the agent should book a meeting every time
  • 5 clearly unqualified leads — the agent should route to nurture without pushing for a meeting
  • 5 borderline leads where your team historically disagrees — see how the agent applies your criteria and refine its instructions based on the results
  • 3 adversarial tests — leads who try to get the agent to make pricing promises or bypass qualification
  • 2 high-value edge cases — leads from unusually large companies or unusual industries

Document every case where the agent's behavior was wrong or surprising. For each, decide whether to fix it via role description refinement, updated topic instructions, or a new guardrail. Don't launch until your pass rate on the 20 scenarios is above 90%.

Step 8: Launch and monitor

Start with 25% of your inbound lead volume routed to the agent. Review every transcript for the first 3 days. Look for:

  • Qualification decisions you disagree with — refine the criteria in the topic instructions
  • Questions the agent couldn't answer — add answers to your knowledge base
  • Conversations that went too long without resolution — tighten the topic descriptions to help the agent reach a decision faster
  • Positive patterns — things the agent did better than expected that you can reinforce

Increase to 100% after 5–7 days if qualification accuracy and CSAT are on target. Set up a Salesforce report to track the four core metrics: speed-to-engagement, qualification rate, meeting booking rate, and pipeline generated.

At Maple Wave AI, our SDR agent implementations take 6–8 weeks from kickoff to production launch. The payoff: a lead generation engine that never misses an inbound lead, qualifies with consistent rigor, and delivers your sales team a pipeline of briefed, ready-to-close opportunities — 24 hours a day.