How Mars By GHC stopped 70% support queries from manual review
Mars By GHC is a $50M health and wellness brand selling supplements and personal care products on Shopify. They handle thousands of customer support tickets monthly through Freshdesk, covering subscription management, shipping inquiries, payment issues, and product questions.
Mars By GHC serves health-conscious consumers purchasing supplements and wellness products through subscription plans. Their support team handles recurring billing questions, shipping delays, and subscription modifications on a daily basis.
The Challenge
The goal was to reduce the volume of repetitive support tickets that required manual agent review - without building complex workflows, writing rules, or changing the team's existing tools. Mars By GHC wanted their support agents to focus on edge cases and complex issues rather than answering the same 10–20 questions every day.
The support team was manually reviewing and responding to every incoming ticket, even though the majority were repetitive - subscription pauses, payment failures, shipping status updates, and standard product questions.
Agents were spending most of their time on queries that had already been solved many times before. There were no automated workflows in place, and building traditional automation (decision trees, FAQ bots, intent mapping) felt too complex and brittle to maintain.
Solution
We connected Frontlore to Mars By GHC's existing Freshdesk account. There was no setup, no configuration, and no workflow building. Frontlore read the team's resolved tickets and began learning how agents respond to different situations. Within the first week, it started drafting responses for incoming tickets. When an agent corrected or edited a draft, that correction became a permanent rule - automatically applied to all future tickets across the entire team.
- Connected Frontlore as a sidebar inside Freshdesk - no new tools for agents to learn
- AI drafts responses based on patterns from previously resolved tickets
- Every agent correction is captured and converted into a reusable rule (27 rules learned in 3 weeks)
- Automated actions like pausing subscriptions, looking up orders, and applying tags - before agents even read the ticket
- Policy rules emerge naturally from team feedback (e.g., "don't disclose internal AI usage," "use standard closing on thank-you replies")
Result
Within 3 weeks of connecting Frontlore, 70% of Mars By GHC's support tickets were handled without manual agent review. The team taught Frontlore 27 rules through natural corrections - no rules were written manually and no workflows were built. The most common 10–20 question types were fully automated, dramatically reducing ticket volume that required human attention.