Gist

AI-powered customer support and marketing automation platform

Gist is a US-based SaaS platform built as an all-in-one solution for customer support and marketing automation. Mallow worked with the client from the early stages to design and develop the platform, enabling businesses to manage customer interactions, marketing campaigns, and engagement from a single system. 

As the platform grew, the need to improve support efficiency became critical. To address this, an AI-powered chatbot and support system was developed to handle routine queries, assist support teams, and ensure faster and more consistent customer interactions. 

Gist
Gist
Gist
Technologies Used
  • Ruby on Rails
  • React
  • PostgresSQL
  • Hetzner
  • Bunny CDN
  • AWS
  • Elasticsearch
  • Clickhouse
  • Kafka
  • OpenAI
OUR WORKS

Major features implemented

The platform combines strong marketing automation capabilities with an AI-driven support system, creating a unified solution for customer engagement and support. 

  • Email marketing and lead engagement - Enables businesses to identify and nurture potential leads through targeted email campaigns, including broadcasts, scheduled emails, and follow-ups for non-openers.
  • Drip campaigns and behavioural targeting - Supports automated email sequences and user segmentation based on behaviour to deliver personalized and relevant communication.
  • Event tracking and customer intelligence - Tracks user activity across the platform to provide insights into customer behaviour and help identify high-value prospects.
  • Automation rules and workflow management - Allows businesses to automate repetitive tasks using predefined triggers and actions, with visual workflows to map and optimize customer journeys.
  • Forms, contacts, and lead capture - Provides multiple form types to capture leads and store them in a centralized system with advanced search and filtering capabilities.
  • Meetings and scheduling integration - Integrates with calendar systems to manage availability, schedule meetings, and streamline interactions with prospects and customers.
  • Knowledge base and self-service support - Enables businesses to create and manage help articles, allowing customers to find answers independently and reduce support dependency.
  • Messenger and live chat engagement - Offers real-time messaging to connect with visitors instantly and identify highly engaged users for proactive communication.
  • AI-powered chatbot and assistant - Automates query handling by capturing, qualifying, and routing conversations, ensuring 24/7 support and improved engagement.
  • Intelligent query filtering and cost control - Uses intent detection to classify queries and avoid unnecessary AI usage, optimizing operational costs.
  • Embedding-based knowledge retrieval - Retrieves relevant information from multiple sources using a vector-based system for faster and more accurate responses.
  • AI Copilot for support teams - Assists agents by suggesting context-aware responses based on historical interactions and knowledge sources.
  • Automated query routing and escalation - Routes queries based on complexity, resolving simple queries instantly and escalating complex ones to human agents.
SYSTEM ARCHITECTURE

Architecture overview

  • The solution follows a layered architecture designed for scalability, reliability, and controlled AI interaction. The system integrates marketing automation, customer engagement, and AI-driven support into a cohesive platform while maintaining clear separation between components.
  • At the core, a Large Language Model handles natural language understanding and response generation. The model operates only on structured context provided by the system and does not directly access any data sources.
  • Customer data, including knowledge base content, marketing data, event tracking logs, and past conversations, is processed through an ingestion pipeline. This data is normalized, segmented, and converted into vector embeddings for efficient semantic search.
  • These embeddings are stored in a vector-enabled data store, allowing the system to retrieve contextually relevant information quickly. Each dataset is scoped per customer to ensure proper isolation in a multi-tenant environment.
  • When a query is received, the system evaluates intent and determines whether it should be handled through automation, knowledge retrieval, or AI generation. Relevant information is retrieved and re-ranked before being passed to the model.
  • An orchestration layer combines user input, conversation history, retrieved knowledge, and system instructions to generate accurate and context-aware responses. Responses are then delivered through the chatbot or support interface.
  • The platform also includes monitoring, logging, and analytics layers to track performance across both marketing and support functions, enabling continuous optimization.
AI APPROACH

AI design approach

  • The platform follows a retrieval-augmented generation approach, ensuring that AI responses are grounded in verified knowledge rather than relying solely on pre-trained data.
  • The AI layer is designed to complement the existing marketing and communication system, not replace it. While the model handles reasoning and conversational responses, knowledge retrieval, customer data, and automation logic remain controlled within the application layer.
  • Knowledge sources are continuously updated through ingestion pipelines, incorporating documents, user interactions, campaign data, and support history. This ensures that both marketing and support systems benefit from shared intelligence.
  • A re-ranking mechanism improves the accuracy of retrieved information by selecting the most relevant context before passing it to the model. This enhances response quality and reduces irrelevant outputs.
  • The system follows a controlled design approach where AI is used selectively. Intent filtering, confidence thresholds, and validation layers ensure that AI is invoked only when necessary and that responses remain accurate, consistent, and aligned with business logic.
  • Safety mechanisms, contextual validation, and prompt-level controls are implemented to maintain response quality, reduce hallucinations, and ensure reliable customer interactions across both support and marketing use cases.
SOLUTIONS OFFERED

Works done in the project till date

The following phases have been diligently pursued in this project.

Consultation
Analysis
Frontend Development
Backend Development
iOS App Development
Android App Development
Web Development
Chatbot Development
DevOps
Third party Integrations
Testing
Live Support & Maintenance
TALK TO US

Ready to transform your customer support and engagement with AI?

Join us in building smarter, scalable solutions that automate interactions, improve response times, and deliver consistent customer experiences.

Connect with our team