Vocily AI

What Vocily AI is

Vocily AI is a voice agent builder for businesses that talk to customers in Indian languages. The company is based in Bengaluru, India, and its platform page is titled "AI Voice Agent Builder, Inbound and Outbound". The core idea is stated on that page: "One agent, every customer conversation. Define an agent once. Vocily AI applies the same role, voice, knowledge, and workflow rules across chat, browser voice, and outbound calls."

It is aimed at operations, support and revenue teams. The pages list industries including financial services, healthcare and clinics, EdTech and coaching, real estate, logistics and last-mile delivery, D2C and e-commerce, and public sector and NGOs. The homepage also describes the product as running "campaigns across phone and WhatsApp with AI agents that listen, respond, and take action".

What you build and how

One agent definition. The platform page says every agent is "built from a small set of explicit fields. No hidden behaviour". The six groups are identity and role, language and voice, knowledge and tools, listening behavior, escalation and handoff, and lifecycle and versioning.

The identity group is the most detailed. You set a display name, a role prompt of "Up to 64 KB" with a built-in caller number variable for inbound calls, and custom runtime variables you can inject per call through the API. The prompt supports section editing for purpose, response style and workflow. You also write the greeting, choose whether the bot speaks first or waits for the caller, and set a widget-side welcome for chat and browser voice. A voice style preamble is applied automatically with the guardrail "speak naturally, no markdown, spelled-out numbers".

Role and scope. The platform page describes a "Job-scoped role": define what the agent is hired to do, the tone it should take and the topics it must refuse. The knowledge boundary is explicit as well. An agent answers from the knowledge base attached to it, and other agents' knowledge bases are not visible, so there is "no cross-talk between billing and sales".

Workflow rules. The page says when to use the knowledge base, when to call your APIs, when to book a meeting and when to transfer are "declared once, applied to every call". Escalation is a rule, not a model judgment: "The agent doesn't decide; the rule does." The FAQ says a transfer can be triggered when a customer asks for a person, sentiment changes, a sensitive topic appears, or the request is outside the agent's approved scope.

Languages and voices. The platform covers "17+ Indian languages, code-mix, per-language voices". The platform page names Indian English, Hindi, Hinglish auto-switch, Bengali, Gujarati, Kannada, Malayalam, Marathi, Tamil, Telugu "and more", picked per agent and switched live by the caller. The homepage says you can choose the speech, voice and reasoning provider for each agent, with "Vendor retry plus automatic STT fallback if a provider hiccups".

Channels. The same agent serves a chat widget, browser voice and outbound calls, and the homepage and FAQ add inbound phone on a real number, web widgets and WhatsApp. That is the reason it belongs on the chatbot-builders shelf as well. The chat side is a widget and WhatsApp, not a visual bot flow editor.

Knowledge. Agents ground answers in your own content: "websites, PDFs, FAQs". The homepage describes deterministic retrieval for exact-match questions and semantic retrieval for conceptual ones, and says that when the knowledge base returns nothing, the agent says "I don't know" instead of inventing an answer.

Actions and scheduling. Agents can book a meeting, look up the caller, transfer to a teammate and send a follow-up SMS. Scheduling happens inside the call: agents check availability, book, reschedule and confirm appointment details.

Outbound. Outbound calls can be triggered programmatically through the REST API. Outbound batches run controlled recipient lists "with progress, reruns, and exportable outcomes". The FAQ adds that when a customer does not answer, the outcome stays visible per recipient, and operations can use controlled retries or focused reruns without calling people who already completed the workflow.

After the call. Each call becomes an execution record with a transcript, recording and call status, tool results such as bookings, lookups and transfers, and follow-up automations. The FAQ lists what a team can review: transcripts, recordings where applicable, actions taken, provider events, costs, sentiment, outcomes and custom post-conversation analysis.

API. Vocily AI ships "a REST + realtime API right now". You can place calls, run batches, attach knowledge, register your own tools and stream every conversation event.

What the pages say about the product

Pricing. Every account includes "the full platform and 10 concurrent voice calls". Usage is prepaid and drawn from a wallet you top up. Voice starts at ₹3.10 ($0.0410) a minute, WhatsApp platform fees start at ₹0.04 ($0.0004) a message, and web chat is charged per LLM response. Lily, the builder assistant, costs ₹1.00 ($0.01) a turn after 30 free turns a day. If you bring your own carrier account, the carrier bills you for the call. If you rent a number from Vocily, its minutes pass through at ₹0.40 ($0.0040) a minute. The only monthly charge is for extra lines, and the pages do not give the price of those lines.

Compliance. The wording here is unusually plain. The compliance primitives are recording consent announcements, PII redaction, topic fencing and hallucination guardrails, and the homepage says "We don't claim regulator certification". You get the controls and your own compliance team does the rest. No SOC 2, ISO or HIPAA claim appears.

Examples. An insurance renewals agent shown on the platform page picks up renewal-due policyholders in Hindi or English, answers premium questions, and books a callback inside the call after hours instead of leaving a voicemail. The use case list runs from missed-call follow-up and appointment reminders to debt and NBFC collections, order verification, clinic receptionist, restaurants and cloud kitchens, automotive dealerships and citizen helplines.

The pages publish no customer names, counts or case studies.

Who it suits

Vocily AI suits Indian businesses and teams serving Indian-language callers, where code-mixing between languages is the normal way people speak. Collections, renewals, reminders and clinic or dealership front desks fit well. The explicit role, scope and escalation fields suit a compliance-minded operations lead who wants the agent held to a job. The published wallet pricing helps a small team estimate spend before talking to anyone.

It suits teams outside India less, since the language emphasis is on Indian languages. It is not a visual flow chart builder or a general website chatbot designer. The pages describe prompts, fields and rules plus a REST API. A buyer who needs a named certification should note the company's own statement that it does not claim one.