Customer support · Production
Wayfair Agent Co-pilot: Generative AI assistant reduces digital sales agent handle time by 10%
The problem
Digital sales agents had to manually hunt down product information, policy details, and other relevant data during live customer chats, slowing response times and reducing service quality.
Workflow diagram · grounded in source
1
Customer initiates chat
trigger
“whether they have a quick question about a specific product or want to collaborate with one of our agents to find the perfect pieces for their home”
2
Prompt assembly
ai_action
“The prompt consists of several crucial components:
- Task Description: This clearly defines what the LLM should do
- Guidelines: These outline internal rules and processes agents need to follow
- Policies: This provides the LLM with up-t…”
3
LLM generates response
ai_action
“the LLM assigns probabilities to different possible tokens that could follow the sequence present in the prompt. The token with the highest probability is selected, and this process is repeated iteratively, adding one token at a time, un…”
4
Agent reviews and sends
human_review
“This response is then presented to the sales agent where it might be sent to the customer as is, or modified to further refine its accuracy and suitability for the conversation”
5
QA LLM quality assessment
feedback_loop
“employs a second LLM – a "QA LLM" – to assess the quality of Co-pilot's responses. We're constantly refining the system to provide the best possible support to our agents and customers”
Reported outcome
Initial test results show a 10% reduction in agent handle time, allowing customers to be served more quickly.
Reported metrics
agent handle time (AHT) reduction10%
Agent time saved on information lookupimmediately surfacing information they would typically have to hunt down
Reported stack
Large Language ModelsQA LLM
Source
https://www.aboutwayfair.com/careers/tech-blog/agent-co-pilot-wayfairs-gen-ai-assistant-for-digital-sales-agents
Read source ↗Frequently asked questions
What did this team achieve with this AI workflow?
Initial test results show a 10% reduction in agent handle time, allowing customers to be served more quickly.
What tools did this team use?
Large Language Models, QA LLM.
What results were reported?
agent handle time (AHT) reduction: 10%; Agent time saved on information lookup: immediately surfacing information they would typically have to hunt down (source-reported, not independently verified).
How is this customer support AI workflow structured?
Customer initiates chat → Prompt assembly → LLM generates response → Agent reviews and sends → QA LLM quality assessment.