Case study — fig. 04
The client's website met five distinct kinds of visitor — each arriving with different questions, at any hour, in whatever state of readiness. The site could only do what sites do: sit there. After business hours, every question went unanswered and every ready-to-talk visitor left the way the Tuesday lead always leaves — quietly, for a competitor who answered first.
An AI agent, built on Agentforce, working the front door. It recognizes which of the five visitor personas it's talking to and adjusts accordingly. It answers from a knowledge library built from the client's own material — so it says what the business would say, not what a model guesses. It captures the lead inside the conversation, into Salesforce, not a side spreadsheet. And it's business-hours-aware: when a human is available, it hands the conversation over; when nobody is, it doesn't pretend — it does the intake properly and queues the handoff for morning.
Note what the agent is grammatically: an employee of the workflow. The workflow decides when it speaks, what it knows, and when it yields to a person.
The front door works the visitor's hours instead of the office's. Questions answered from approved knowledge, leads captured at the moment of interest, and humans entering conversations exactly when they add value — never impersonated, never bypassed.
An AI agent isn't a chatbot you bolt on. It's a hire. It needs a job description, source material, working hours, and an escalation path — and when it gets them, it stops being a novelty and starts being coverage.
Bought AI tools that never got wired into the actual work? Tell me where work is falling through — I'll tell you the first three places I'd look. Free, no pitch.
Find the leak or just message me →