Customer support · Production

How AI Agents Are Transforming Customer Service with Freshworks Freddy AI

The problem

Customer service teams face rising ticket volumes, repetitive queries, and rigid legacy chatbots that break when conversations deviate from preset scripts, leaving customers frustrated with incomplete or impersonal responses.

First attempt

Traditional chatbots relying on rigid scripts and flows fail to handle multi-part or off-script queries, forcing customers to repeat themselves and escalating frustration.

Workflow diagram · grounded in source
1
Customer query received
trigger
“A customer types a question in natural language on any channel (chat, email, WhatsApp, etc.). The AI Agent instantly detects the intent, tone, and context using large language models (LLMs).”
2
Intelligent knowledge search
ai_action
“the AI Agent searches across pre-approved knowledge sources, such as solution articles, offline files (.PDF, .DOCX, .TXT), structured Q&As, and public URLs (like pricing pages or return policies)”
3
Guardrails and compliance check
validation
“Strict safeguards are in place to guarantee compliance, with no access to personal or proprietary customer data by default, unless granted access by the Admin based on the business use case. Continuous checks are enforced to prevent any …”
4
RAG-grounded response generated
ai_action
“With RAG, the AI Agent assembles a fluent, human-like response fully anchored in the retrieved material, avoiding hallucinations and ensuring reliability.”
5
Action execution and response delivery
output
“If resolution requires action in an external system, like initiating a refund, triggering a workflow, or updating CRM data, the AI Agent seamlessly executes it via connected APIs before responding. The AI Agent delivers a response instan…”
6
Human agent escalation
human_review
“Human agents take over queries that require the human touch, where empathy, judgment, or exception handling is essential. L1 support ensures a smooth transition, with full context carried over from the customer service AI agent.”
Reported outcome

PhonePe automates 60% of its queries to serve 150 million customers using Freshworks.
Freshworks positions Freddy AI Agent as delivering significant cost savings compared to per-resolution pricing competitors.

Reported metrics
PhonePe query automation rate60%
PhonePe customers served150 million
Cost savings vs Zendesk at 50% resolution rate86%
Cost savings vs Fin AI Agent at 50% resolution rate~80%
Show all 5 reported metrics
PhonePe query automation rate60%
PhonePe customers served150 million
Cost savings vs Zendesk at 50% resolution rate86%
Cost savings vs Fin AI Agent at 50% resolution rate~80%
Cost savings vs Agentforce~95%
Reported stack
Freddy AI AgentFreddy AI Agent StudioFreddy AI CopilotRAGLLMs
Source
https://www.freshworks.com/freshdesk/ai-agents/customer-service/
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

PhonePe automates 60% of its queries to serve 150 million customers using Freshworks.

What tools did this team use?

Freddy AI Agent, Freddy AI Agent Studio, Freddy AI Copilot, RAG, LLMs.

What results were reported?

PhonePe query automation rate: 60%; PhonePe customers served: 150 million; Cost savings vs Zendesk at 50% resolution rate: 86%; Cost savings vs Fin AI Agent at 50% resolution rate: ~80% (source-reported, not independently verified).

What failed first in this deployment?

Traditional chatbots relying on rigid scripts and flows fail to handle multi-part or off-script queries, forcing customers to repeat themselves and escalating frustration.

How is this customer support AI workflow structured?

Customer query received → Intelligent knowledge search → Guardrails and compliance check → RAG-grounded response generated → Action execution and response delivery → Human agent escalation.

Related customer support cases