eMeals is a subscription-based meal planning platform that handles a high volume of customer support requests, including billing, refunds, subscriptions, login issues, and shopping assistance. As ticket volumes grew, manual categorization became time-consuming and inconsistent. eMeals needed an AI-powered solution to automate ticket classification while fitting seamlessly into its existing support workflow.
.png&w=3840&q=75)
Support triage is the kind of work that quietly consumes a team. Every ticket needs reading and categorizing before anyone can start solving it, different agents reach different conclusions on the same issue, and the only way to handle more volume is to hire more people. Meanwhile the categories themselves keep changing, and none of the resulting decisions are recorded anywhere useful.
Every ticket about cancellations, refunds, billing, login issues, or shopping assistance required an agent to read the subject and description before assigning a category. As volumes grew, that repetitive work consumed time better spent resolving issues, and handling more tickets meant hiring more staff.
Different agents interpreted customer requests differently, so similar issues were sometimes tagged under different categories. That inconsistency made reporting, trend analysis, and workflow automation harder to rely on, and holding a common standard became more difficult as the support team expanded.
Rather than replacing the customer support platform, the automation was built to sit inside the existing Zoho Desk environment and leave the agent workflow untouched. Event-driven architecture, a language model, and configurable business rules combine to handle the repetitive decision, while classification logic lives in configuration rather than code so it can move at the speed of the business.
Webhooks capture each new ticket the moment it is created in Zoho Desk, with no manual step required. Once the category is determined, the system writes tags and customer metadata straight back onto the ticket, so agents open pre-classified work and start resolving the issue rather than filing it first.
Rather than keyword matching, a large language model reads the ticket subject and description to work out what the customer actually wants. That means two people describing the same billing problem in completely different words still land in the same category, which keyword rules had consistently struggled with.
The system has classified more than 31,098 tickets in production, averaging 139 a day and handling 631 in a single day at peak, without adding support headcount in proportion.
Automatic classification into 15 configurable categories replaced manual interpretation, so similar customer issues now route through the same workflow whichever agent receives them.
A central record of every AI decision revealed that nearly 80 percent of requests fall into just three categories, which informs workforce planning and self-service investment.
Eternalight automated eMeals ticket triage with a serverless event-driven pipeline on AWS Lambda and Amazon SQS. Zoho Desk webhooks capture each ticket, a large language model determines intent, and configurable prompts held in MySQL and DynamoDB let business teams change categories without a deployment. Delivery followed a structured process from brainstorming and wireframes through development, integration, launch, and user feedback.
Brainstorming
Wireframe
Development
Send us a message, and we’ll promptly discuss your project with you.
Support categories and business policies evolved regularly as new products, subscription models, and customer scenarios appeared. Traditional rule-based systems needed an engineering change and a deployment every time classification logic moved, which slowed business agility and added maintenance effort.
Tickets were stored in Zoho Desk, but there was no centralized repository of classification results that analytics, auditing, or continuous improvement could draw on. That gap made it hard to understand which customer issues kept recurring, or to evaluate how support performance actually shifted over the course of time.
Classification prompts and business rules live in a centralized configuration database rather than inside application code. Business teams can enable, disable, or reword ticket categories themselves, so the AI evolves alongside new products and subscription models without an engineering change or a deployment.
Every classification the AI makes is recorded in an internal analytics database, creating a complete record of decisions. That gives operational transparency, supports reporting on which issues recur, and provides the historical data needed to refine accuracy as the category set continues to change.
User Feedback
Launch
Integration
Understands and processes complex customer conversations.
A cross‑platform, open‑source JavaScript runtime environment for building fast, scalable server‑side applications.
Structures and exchanges data between application components.
Powers backend logic and workflow automation.
Understands ticket intent and automates classification.
Enables reliable, scalable message processing between services.
Manages customer support tickets and support workflows.
A popular, fast relational database suited for web apps and read‑heavy workloads.
