Build
Two agents. One qualifies the inbound, one runs the nurture and books the showing.
ADK · Agent Studio · Model Garden
BLUEPRINT · REAL ESTATE · PROFESSIONAL SERVICES
Lead Qualification & Follow-up
Qualify and nurture inbound leads in real time, book the showing, hand the broker a warm conversation.
Qualify and nurture inbound leads in real time, book the showing, hand the broker a warm conversation. Named scope, named timeline, named stack — ADK · A2A · Memory Bank · 10 weeks.
Design ceiling
Inbound response time from five hours to under five minutes. Close rate from 9% to 15%. Eight to ten more live negotiations per broker per month.
The agent system is designed to qualify the inbound lead in real time, run the nurture sequence across days or weeks as the lead warms, and hand the broker a booked showing with the conversation history already on the page.
The problem
Five hours late to an inbound lead is the same as not replying.
A regional brokerage carrying inbound from the listing site, the referral network, and the open-house sign-in sheet is paying its brokers to chase the same opening conversations on every lead: qualify the budget, qualify the timeline, qualify the area, find a slot for a showing. The first response window is small and the brokers are in the field, so the strong leads go cold before anyone replies, and the broker spends the evening writing follow-ups instead of closing the ones already in motion.
The job is not to replace the broker. The job is to answer the inbound inside the response window, qualify in real time, run the nurture sequence over days as the lead warms, and hand the broker a booked showing with the full conversation history. Two agents, A2A handoff between the qualifier and the nurture-and-scheduler, Memory Bank carrying the lead state across days, every booked showing linked back to the broker who owns the relationship.
Agent architecture
Four pillars, this blueprint’s stack.
The platform’s four pillars, mapped to the components this agent system actually exercises.
Scale
Qualifier and nurture agent talk over A2A. Lead state lives in Memory Bank across days.
Agent Runtime · A2A v1.2 · Memory Bank
Govern
Every contact policy-checked. Every booked showing tied to the broker who owns the relationship.
Agent Registry · Model Armor · Gateway · Identity
Optimize
Eval set walks the qualifying edges and the historical broker close-or-pass signal.
Evals · Observability · Agent Analytics
Engagement · 10 weeks
Week by week.
Fixed scope, fixed price, fixed timeline. Here is what happens when.
Week 1-2
Discovery and lead-source inventory.
Walk the current intake-and-nurture workflow with the broker-owner and two senior brokers. Inventory the lead sources, the listing feed, the CRM, the calendar. Agree the qualifying bar and the nurture cadences. Sign the SOW.
Week 3-4
Qualifier agent build.
Stand up the ADK qualifier agent against the inbound channels. First eval pass on a sampled quarter of historical leads against the broker close-or-pass signal. Tighten the qualifying conversation against the eval results.
Week 5-6
Nurture agent and A2A handoff.
Build the nurture-and-scheduler agent. Wire the A2A protocol between qualifier and nurture. Memory Bank carries the lead state across the days the lead spends warming.
Week 7-8
Governance and observability.
Register both agents in Agent Registry. Configure Model Armor policies for contact handling and outbound messaging. Stand up the Agent Analytics dashboard so the broker-owner can see response time, qualified-lead volume, and showing bookings live.
Week 9
Staging and shadow run.
Run the agent system in shadow on live inbound. Compare the qualified-and-nurtured shortlist to the broker queue. Tune the eval set and the nurture cadences against any drift.
Week 10
Production cutover and handoff.
Deploy to Agent Runtime with the broker landing on the showing invite. Walk the runbook with the broker-owner. Hand the team the repo, the eval harness, the dashboard, and the on-call playbook.
What it looks like in code
ADK Python — the agent definition.
The actual shape of the code your team owns at engagement end. Real ADK, real tools, real instruction copy.
agents/lead_qualifier/agent.py
python
from google.adk.agents import LlmAgentfrom google.adk.tools import FunctionToolfrom .tools import ( fetch_inbound_lead, qualify_against_criteria, handoff_to_nurture, record_lead_state,)qualifier = LlmAgent( name="lead_qualifier", model="gemini-2.0-pro", instruction=( "You answer inbound leads inside the first-response window. " "Run the qualifying conversation against the brokerage's bar: " "budget, timeline, area, intent. Hand qualified leads to the " "nurture agent with the full conversation attached. Surface " "showing-ready leads to the broker who owns the area." ), tools=[ FunctionTool(fetch_inbound_lead), FunctionTool(qualify_against_criteria), FunctionTool(handoff_to_nurture), FunctionTool(record_lead_state), ],)What you walk away with
Code your team owns. Governance baked in.
Every blueprint hands the engineering team a deployed agent and the artefacts to run it themselves. No black box, no lock-in.
- Two ADK agents (qualifier + nurture-and-scheduler) deployed on Agent Runtime, in your Google Cloud project.
- The Git repo your team owns, with the agent code, tools, prompts, and tests.
- CI/CD pipeline wired to your repo, deploying on merge.
- Eval harness with a sampled-quarter corpus across the qualifying edges and the broker close-or-pass signal.
- Governance bundle: Agent Registry entries linking each booked showing to the named broker, Model Armor contact-handling policies, Gateway access controls.
- Agent Analytics dashboard plus an on-call runbook for the broker-owner.
Book this blueprint as a pilot.
Two weeks. Named scope. Working agent on Agent Runtime at the end.
Code
Lives in your Git org, owned from commit one.
Governance
Model Armor and Agent Registry on day one.
Speed
Two weeks to a runnable pilot. Eight to production.
Not ready to talk? Take the 4-min readiness assessment