Customer-AI orchestrator · for Tradetron
CallFunnel.ai calls your users, decides what to say in plain English, and never breaks character. Voice + WhatsApp + Telegram + email — one brain, one timeline per customer, real authority to offer concessions with your approval. Built for 5 lakh users from day one.
The status quo
Voice has been dead since IVR. WhatsApp is fragmented from email. The bot can't make a real decision. You end up just hiring more humans.
IVR menus are a productivity tax. Users hang up. The 2-3% that don't, get routed wrong half the time. The bot doesn't know who you are.
WhatsApp doesn't know what voice said. Email doesn't know what Slack escalated. Every channel runs its own context window and forgets the customer between contacts.
When a caller says "₹999 nahi de paunga right now" — the bot recites a fallback script. A human would offer a free strategy build. The bot doesn't have that authority. It should.
What Conductor does differently
A natural-language rule book replaces the workflow builder. One cross-channel memory replaces ten siloed CRMs. Real authority — with hard guardrails — replaces fallback scripts.
You type what you want in plain English. Claude Opus writes a structured rule book — goal, audience, opening, intents, dispositions, safety. No nodes, no dropdowns, no clicks. The dropdown era is dead the moment Opus + tool-use existed.
referral_or_publish, hostile-DNC path, and a do-not-collect-friend's-details safety line.
Mongo holds an append-only feed of every voice call, WhatsApp message, Telegram reply, email open, Slack escalation, billing event, strategy event — keyed to the user. When the bot calls Manish for the third time, it doesn't start cold.
Bot can offer free strategy builds, plan extensions, even patch the call to a creator. Below your ₹X cap it decides per policy. Above, it asks you on Slack with the caller's full context — and waits for one tap. Every decision is audited. One kill-switch stops everything.
Hear actual openings
Every opening below was rendered by Cartesia from the workflow's rule book, just like the bot would render it on the actual call. Pre-rendered openings cost ~80 ms less than live TTS — meaningful when the first impression hangs on it.
From idea to first dial
No engineers. No workflow editors. No drag-and-drop. You type, you listen, you ship.
One paragraph. Plain English. Who to call, what's the goal, what the bot should do when the caller pushes back.
Claude writes the rule book. Conductor renders the opening line with the picked voice. You hear it in your browser. Tweak the words or swap voices until it sounds right.
Upload audience (CSV or segment query). Hit Run. Watch the live thread in Slack. Approve concessions with one tap when they ping you. Read the weekly digest Monday morning.
The voice cast
Pick per workflow. Or set a default in Settings and let each campaign override. Hindi · Marathi · Tamil · Bengali land day 2 via Sarvam Bulbul.
Built in, not bolted on
Most outbound platforms ship guardrails as v3. CallFunnel.ai ships with them mandatory in v1 — because the bot has real authority, so the floor matters.
At 5 lakh users
Conservative numbers. Renewal recovery + win-back + refund prevention alone. Add onboarding-stuck and margin-rescue and the cost line moves down too.
It's already running.
CallFunnel.ai is live on y-mcp. The dashboard, build flow, library, approval queue, settings, reports — all yours.