Nikhil Siwan

Case study

Daisy SuperHuman Chat Bot

A human-like AI sales representative for SolidAppMaker — a persona engineered to chat naturally, understand a visitor's idea, and only move toward booking a call once they're genuinely ready.

AI ProjectsProject 06 / 07
Daisy SuperHuman Chat Bot project showcase

Overview

What this is

Daisy is an AI-driven "SuperHuman" representative built for SolidAppMaker's website — a persona-driven chat agent designed to feel like a real team member rather than a bot. She opens every conversation naturally, takes a genuine interest in the visitor's idea, and only moves toward scheduling once they're confident SolidAppMaker can help — then hands off to Calendly and logs the finished conversation as a lead in SolidAppMaker's CRM.

01

Problem

2 key points

02

Research

2 key points

03

Solution

4 key points

01

Problem

Off-the-shelf AI chat widgets read as scripted, and every change to Daisy's persona or boundaries meant pulling in a developer instead of the team that actually talks to customers.

01

Bot Language Kills Trust

Default AI widgets lean on phrases like "as an AI" or "virtual assistant," which immediately signals to a prospect that they're talking to a script instead of a person.

02

Interrogation-Style Intake Forms

Most chat widgets open by demanding a name and email before they'll say anything useful, which reads as a form wearing a chat bubble rather than a conversation.

02

Research

The conversation design started from a working reference tool and the client's own internal voice guidelines, rather than a blank slate.

01

HubSpot Chatflows as the Starting Reference

Early conversation logic was sketched out inside a HubSpot chatflow as a working reference point, before the persona moved to a more natural, context-driven conversation instead of a rigid step-by-step script.

02

Codifying the Client's Own Voice

SolidAppMaker's engagement rules, business ethics, and an MBTI/Enneagram personality profile for Daisy were folded directly into the system prompt, so the persona reflects how the company actually wants to sound.

03

Solution

I engineered a persona that leads with genuine conversation, and only asks for contact details once the visitor is actually ready to move forward.

01

Strict Engagement Rules

A locked-down language layer bans phrases like "as an AI" and "virtual assistant" and defines exactly how Daisy redirects off-topic or personal requests, keeping the human illusion consistent.

02

Conversation Before Qualification

Daisy opens by taking real interest in the visitor's idea and explaining how SolidAppMaker can help, rather than leading with a name-and-email intake form.

03

Calendly Handoff, Once They're Ready

Only after the visitor is engaged does Daisy ask for a name and email, then hands over a Calendly link so they can pick a slot on their own time instead of negotiating one in the chat.

04

Automatic CRM Lead Capture

The finished conversation is logged straight into SolidAppMaker's CRM as a new lead, with no manual re-entry.