How Raaz.app built a digital clinic on WhatsApp - without a single developer
Five specialised AI agents. One patient journey. From first "Hello" to post-treatment recovery — in Hinglish, with full human oversight, deployed in 3 days.
The Company
Raaz.app helps men with PE and ED — sensitive issues that most men find hard to talk about openly. Their biggest competitor is not another startup. It is embarrassment. When a patient feels judged, they close the app and never come back. Raaz needed AI that could listen, understand local slang, and guide a patient from first contact to post-treatment recovery. A single chatbot could not do it well enough. So they built a crew.
58%
Increase in lead-to-booking conversion
41%
Patient queries handled entirely by AI
41%
Better appointment show-up rate with ₹49 token
3days
To build and deploy each new agent - no developer
The Problem
A "Press 1 for Appointments" menu kills the conversation before it starts.
When a man is nervous about a sensitive health issue, a robotic flow feels cold. He sees a menu, feels like a ticket number, and leaves. The clinical friction is the conversion killer — not the product, not the price.
Trust is the barrier
Men reaching out to Raaz are already anxious. Any moment of friction — a confusing menu, a generic question, a delayed reply — breaks the fragile trust needed to continue.
One bot cannot do everything
Triage, scheduling, anxiety management, payment nudging, and post-care follow-up are four completely different jobs. A single agent trained to do all of them does none of them well.
Language and culture matter
Patients communicate in Hinglish — a natural Hindi-English mix. A bot that does not understand Devanagari script, local slang, or even a rushed typo loses the patient immediately.
Scale vs empathy tradeoff
Hiring enough human agents for 24/7 empathetic support is expensive and impossible to scale. But pure automation without warmth is worse than no automation at all.
The Solution
Five agents. Each with one job. No developer required.
Instead of one bot trying to do everything, Raaz built a crew. Each agent hands off to the next when their role is done. Every agent was built in plain English — no code, no flowcharts. A junior intern manages the flows.
1
Raaz Mitra
The Empath — First contact & triage
+58% conversion
What it does
Greets every user as "Sir." Never asks for their name straight away. Communicates in natural Hinglish using Devanagari script. Identifies whether the user has a problem, then asks about their situation — gently, without judgment.
Why it works
Removing the judgment barrier increased lead-to-booking conversion by 58% vs the previous static menu flow. Patients feel safe — not processed.
Plain English instruction used
"You are a helpful friend named Kundan. Assist men with PE and ED only. Communicate naturally in Hindi using Devanagari script. Never ask for their name immediately — keep them anonymous until they feel safe."
2
Scheduler
Appointment booking
Slot to confirm <2 min
What it does
Once the patient admits they have a problem, this agent takes over. Connects to Zoho CRM. Offers three specific 1-hour slots in AM/PM format to cut decision fatigue. Confirms without asking for re-confirmation. Tells the patient to save the doctor's number as "Raaz Incoming."
Why it works
Specific slot options remove the paralysis of open-ended scheduling. The "save as Raaz Incoming" instruction means patients recognise the call and pick up. Slot to booking under 2 minutes.
3
Nurturer
FAQ & pre-appointment anxiety
Reduces cancellations
What it does
Between booking and the appointment, patients get nervous and cancel. This agent manages the anxiety window. Answers safety and privacy questions by reading from an uploaded FAQ PDF. Handles Hinglish typos — "binaries" instead of "bimari" — without breaking.
Why it works
A patient in a hurry makes typos. The agent understands intent, not just spelling. Pre-appointment cancellations dropped because patients got answers fast — without waiting for a human.
How the FAQ was added
The team uploaded a PDF of their FAQs into Peach. The agent reads the file to answer questions. Adding new topics takes minutes — upload a new PDF and the agent handles it the same day.
4
Closer
Order & payment
+41% show-up rate
What it does
After consultation, patients hesitate to pay for medicine. This agent pulls the patient's age from Zoho CRM. If they are young, it says: "For men in your age group, we see results in 4 days." Asks for a small ₹49 token to confirm serious intent.
Why it works
Personalised nudge beats a generic buy link every time. The ₹49 token is not about the money — it filters serious patients. Appointment show-up rate improved 41% after introducing this step.
5
Raaz Mitra
The Empath — First contact & triage
+58% conversion
What it does
Greets every user as "Sir." Never asks for their name straight away. Communicates in natural Hinglish using Devanagari script. Identifies whether the user has a problem, then asks about their situation — gently, without judgment.
Why it works
Removing the judgment barrier increased lead-to-booking conversion by 58% vs the previous static menu flow. Patients feel safe — not processed.
The Difference
What Peach says vs what a standard bot says.
The same patient. The same message. A completely different experience.
Patient says
Peach agent crew
Standard bot (WATI / Interakt)
"I'm scared and in pain"
"I hear you, and I'm here to help. Is the pain sharp or dull?"
"Error: Please press 1 for
"Can I come later?"
"No problem. Dr. Sharma has a gap at 5 PM. Does that work?"
"Invalid date format."
Types "binaries" (typo for "bimari")
Understands the intent. Answers the illness question correctly.
"I don't understand your query."
"Is this private? Will my family know?"
Reads the privacy FAQ and gives a specific, reassuring answer.
"Please visit our website for more information."
"I feel dizzy after taking the medicine"
Sends a comforting video. If symptoms are severe - stops and alerts a human immediately.
"Your query has been received. We will respond in 24 hours."
How It Was Built
No developer. No flowchart. Plain English.
The Raaz team did not write code. They wrote instructions. Peach's Natural Language Programming means any team member — including a junior intern — can build, test, and deploy a new agent in days.
Three things that made this possible without a developer
Old platforms required giant decision trees. Peach replaced that with plain English instructions.
No decision trees
On WATI or Gallabox, every Yes/No needs a drawn flowchart. On Peach, the intern wrote: "If the user asks to reschedule, hand off to the Booking Agent." The AI figures out the rest.
Plain English
Simple guardrails
To stop the AI from going off-script, they added one rule: "Do not call the get_record tool twice." That was enough. No complex logic required.
One-line rules
Instant testing
A simulator on the right side of the screen lets the team test changes in real time — before going live. Change a prompt. Test it. Ship it. 3 days per new agent.
No deployment risk
Human Oversight
The safety net behind every AI agent.
Founders worry: what if the AI gets it wrong? Peach solves this with a shared inbox where humans are always one click away.
The Watchtower
Human agents watch the AI talking to patients in real time. If the AI gets stuck or a patient gets upset, a human clicks Take Over. The AI stops. The human takes the conversation. The patient never knows the switch happened.
Smart escalation
Raaz set specific triggers. If a patient mentions "bleeding," "suicide," or "emergency" — the agent stops talking and tags a human supervisor immediately. No delay, no AI attempt to handle it.
Learning from analytics
The shared inbox showed that patients were asking about diet — not in the training data. The team uploaded a Diet PDF. The agent started answering diet questions the next day. No developer needed.
The Results
First 30 days.
58%
Increase in lead-to-booking after Raaz Mitra removed the judgment barrier
41%
Better show-up rate after the ₹49 token payment via the Closer agent
41%
Better show-up rate after the ₹49 token payment via the Closer agent
67%
Patient queries handled entirely by AI — zero human involvement needed
28%
Increase in second-month refills via the Caretaker post-delivery check-in
3 days
To build and deploy each new agent — managed by a non-technical intern
In Their Words
"We didn't build a bot. We gave our patients a friend. One that speaks their language, remembers their situation, and checks in after they've gone home. That's what Peach made possible."
FOUNDER · RAAZ.APP
Building AI for a high-trust, nuanced vertical?
Peach lets non-technical teams build, test, and deploy specialised AI agents in plain English — with human oversight built in. If your customers need empathy, not menus, book a 15-minute call.
