Conversational AI Outbound Intelligence System

About the Customer

Ankur Graphics is a founder-led printing press serving B2B clients across India. The business provides commercial printing services to enterprises and operates as a traditional print shop, where revenue generation primarily depends on order fulfillment and repeat business from existing customers. Like many small B2B operations, growth was constrained by the founder's limited bandwidth and reliance on word-of-mouth referrals rather than systematic lead generation.

Customer Challenge

The founder spent his entire day managing existing orders and fulfillment operations, leaving zero time for systematic lead generation or business development. Revenue came primarily through word of mouth, which meant growth was unpredictable and constrained by the founder's limited bandwidth. The business had a lead database with potential customers, but no one was calling them—every day without outbound calling represented direct revenue loss and missed deals.

The founder would not hire sales staff without proven ROI first, so the database sat unused and revenue stayed flat. Although he used ChatGPT daily for business tasks and understood AI could automate repetitive work, he could not figure out how to apply it to lead generation. When isaga explained AI outbound calling, he immediately saw the opportunity: AI could make calls while he focused on operations, unlocking a new revenue channel without adding headcount. Traditional outbound dialers were inadequate as they provided only static call automation without intelligent two-way conversational capability, while the use case required dynamic multi-turn dialogue, contextual retrieval of order and pricing information, payment negotiation handling, and consultative discussion for custom print specifications.

The founder was willing to start even at 40-50% accuracy because this represented revenue upside—an incremental opportunity where any conversion meant net-new business. The core problem was straightforward: he had leads, he had no time to call them, and he needed automation to convert that unused database into predictable revenue without hiring staff or waiting for free time that would never come.

Partner Solution

isaga partnered with AWS to design and deploy an AI-powered outbound calling platform that leverages Amazon Bedrock, Amazon Nova Sonic 2, Amazon ECS, Amazon S3 Vectors, and Exotel to deliver scalable, intelligent, and cost-effective automated conversational calls. The solution enables enterprises to run outbound campaigns for lead generation with dynamic, multi-turn dialogue capability.

The solution is deployed on AWS as a cloud-native, microservices-based platform running on Amazon ECS. An Agent Management Service orchestrates outbound call campaigns, manages conversation state, and coordinates real-time, two-way voice interactions. The AI agent autonomously reasons over customer intent, retrieves relevant business context using Retrieval-Augmented Generation (RAG), and conducts natural, multi-turn conversations during live calls.

Amazon Nova Sonic 2 provides real-time speech-to-speech capabilities with native support for Hindi and English, including Indian accent recognition critical for B2B communications across diverse customer bases. Amazon ECS hosts the LiveKit Server for WebRTC-based voice streaming and the Agent Management Service, managing outbound call workflows, campaign triggers, agent orchestration, and real-time conversation state. Amazon S3 serves as the database powering the RAG pipeline for knowledge base and calling details. Amazon Bedrock interprets customer intent and queries Amazon S3 via RAG to retrieve relevant business context such as order status, pricing, or payment history before generating responses.

Exotel handles telephony integration, programmatically initiating outbound calls, managing SIP connectivity, and supporting call routing, including seamless transfer to human sales when escalation is required. Post-implementation, isaga provides ongoing managed support including prompt engineering and tuning based on call transcripts, knowledge base updates, performance monitoring and optimization, and regular reviews to identify new automation opportunities.

Results and Benefits

The AI system makes 1,550+ calls per month, generating 45+ qualified demos. The founder then takes over and converts approximately 20% of these leads, closing 9 deals per month. This adds nearly one lakh in new monthly revenue that did not exist before—representing incremental business that was previously locked in an unused lead database.

The solution enabled intelligent multi-turn outbound conversations with contextual objection handling and automated demo scheduling powered by Amazon Bedrock. Amazon Nova Sonic 2 provided natural speech interaction optimized for Indian accents and Hindi-English code-switching, improving engagement quality during cold outreach. The Total Cost of Ownership analysis demonstrated measurable financial outcomes, including consistent monthly net new revenue after covering AI calling and data access costs, predictable funnel conversion performance, and the ability to increase outreach volume without incremental hiring.

Infrastructure costs remained tightly linked to actual usage, ensuring cost transparency and operational control. Return on investment was driven by reduced dependency on manual effort and unlocking net new revenue. The founder can now scale outbound calling volume independently of his available time, transforming a static lead database into a predictable revenue engine without hiring sales staff.