You'll design a complete automation for a daily sales lead qualification workflow — from the trigger that starts it, to the 7 agent steps that run it, to the metric that proves it works. Then you'll run the full agentic loop live.
Before you design the automation, understand exactly what you're replacing — and where the 2 hours per day disappears.
Apex is a B2B SaaS company with a 6-person sales team. Every morning, 40–120 new leads arrive from three sources: their website contact form, LinkedIn lead-gen ads, and a data provider (ZoomInfo). Each lead requires the same 7-step manual checklist before a rep can touch it:
① Check CRM for duplicates — to prevent embarrassing double-outreach
② Verify the email — bounced mail burns domain reputation with ISPs
③ Enrich company data — employee count, funding stage, industry from a third-party API
④ Score against ICP — job title, company size, industry, funding stage, intent signals
⑤ Route to the right rep — by territory (region) and vertical (industry)
⑥ Draft personalized outreach — reference recent company news and job posting signals
⑦ Log everything to Salesforce — score, routing decision, draft email attached
The SDR team lead does steps ①–④ every morning. It takes 90–120 minutes. Research shows leads older than 4 hours have a 5× lower response rate. By the time they're processed, most leads are cold.
The trigger is the event that fires the agent. It determines lead freshness, duplicate risk, and whether the automation actually solves the response-time problem. Choose the right architecture — then answer two configuration questions.
Below are 10 candidate steps. Your pipeline needs exactly 7 — in the right order. Click a step from the library, then click a numbered slot to place it. Three options are decoys that violate the human-in-the-loop boundary or don't belong in this workflow.
Each of the 7 pipeline steps needs a specific tool. Match the right tool to each step. Two tools appear more than once — that's intentional. Two listed tools are decoys that don't belong in any step.
Five leads just arrived. Click Run All to process each one through your 7-step pipeline. Watch the step-by-step reasoning trace, then see the final verdict for each lead.
Your agent is running. Now you need to prove it's working. Choose the one key efficiency metric you'd track — the one that maps directly to the business outcome this automation was built to deliver.
Watch the complete Perceive → Reason → Act → Observe cycle run on one final lead. This is the conceptual architecture behind every autonomous agent — not a linear script, but a loop that evaluates and adapts.
Drawn from every section of this lab. 70% to pass.