Customer support · Production

Rohlik Group deploys agentic shopping and support for 3 million e-grocery customers

The problem

Rohlik's customer support was fully human-led, and as the company scaled across five markets and six languages, maintaining the same speed and quality became increasingly complex, with longer wait times at peak hours and no live coverage overnight or in the early morning. Customers could also only shop and get support through the app or website.

Workflow diagram · grounded in source
1
Deploy Maia across channels
trigger
“Rohlik configured Maia on ElevenAgents and deployed it across four channels.”
2
Voice agent answers phone calls
ai_action
“On the phone hotline, the agent handles 90% of incoming calls with a natural-sounding voice.”
3
MCP backend integration
integration
“ElevenAgents connects directly to Rohlik's backend systems via Model Context Protocol (MCP) - an open standard that lets AI agents securely call APIs and take actions in external systems.”
4
Autonomous backend actions
ai_action
“Through MCP server integrations, Maia can perform more than 30 distinct actions, including: - Checking real-time order status and delivery locations - Making live modifications to existing orders - Issuing credits and resolving delivery …”
5
Voice-based shopping and checkout
ai_action
“customers can search for products, get recipe recommendations, build a shopping list, and complete a full checkout - all through conversation”
6
Escalation to human operator
routing
“When Maia cannot resolve an issue - or when a customer asks for a human - the platform routes the call immediately and passes the full conversation transcript to the human operator, so the customer never repeats themselves.”
7
Asynchronous ticket filing
output
“For less urgent issues, customers can ask Maia to file a support ticket on their behalf, and the request is handled asynchronously.”
Reported outcome

Maia now handles 90% of incoming phone calls and 90% of overall customer communications automatically across phone, web, app, and WhatsApp, delivering over 2x faster resolution, 24/7 availability across five markets and six languages, and is on track to cut operational costs by 50% while maintaining CSAT.

Reported metrics
Incoming phone calls handled by voice agent90%
Customer communications handled automatically90%
Distinct backend actions availablemore than 30 distinct actions
Languages supportedsix languages total
Show all 7 reported metrics
incoming phone calls handled by voice agent90%
customer communications handled automatically90%
distinct backend actions availablemore than 30 distinct actions
languages supportedsix languages total
resolution speedOver 2x faster resolution
support availability24/7 availability across 5 markets and 6 languages
operational cost reduction (projected)on track to reduce operational costs by 50%
Reported stack
ElevenAgentsMaiaModel Context Protocol (MCP)WhatsApp
Source
https://elevenlabs.io/blog/rohlik
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Maia now handles 90% of incoming phone calls and 90% of overall customer communications automatically across phone, web, app, and WhatsApp, delivering over 2x faster resolution, 24/7 availability across five markets a…

What tools did this team use?

ElevenAgents, Maia, Model Context Protocol (MCP), WhatsApp.

What results were reported?

Incoming phone calls handled by voice agent: 90%; Customer communications handled automatically: 90%; Distinct backend actions available: more than 30 distinct actions; Languages supported: six languages total (source-reported, not independently verified).

How is this customer support AI workflow structured?

Deploy Maia across channels → Voice agent answers phone calls → MCP backend integration → Autonomous backend actions → Voice-based shopping and checkout → Escalation to human operator → Asynchronous ticket filing.

Related customer support cases