Spur caught the AI concierge going silent mid-conversation

Bug Book — Wander
A guest sends a message, it shows as delivered, and then nothing. No response, no typing indicator, just silence. For a travel platform where the concierge chat is a core part of the booking experience, that's a critical failure.
TYPE OF TEST
AI Feature Error
SEVERITY
Critical

What this run

has verified.

Spur's AI Feature Testing agent opened the concierge chat, sent a detailed message, and waited for a response. After 10 seconds with no reply and no sign of activity, it flagged the chat as non-responsive.

Response reliability

The concierge chat returning a response naturally.

AI availability

The chatbot staying functional across varying conversation lengths.

Guest support

A working channel for guests to get help during booking.

Session stability

Chat interactions completing without silent failures or timeouts.

Compare

The customer's live site,

captured by

Spur

during the run.

Spur Agent Analysis

Failure reason: The concierge chat accepted a guest's message and showed it as delivered, then went completely silent, no reply, no typing indicator. So on a travel platform where the concierge is a core part of booking, guests reaching out for help mid-booking got nothing back and were left to abandon or book elsewhere.

3 - Step failed

What this

one catch saved.

8hrs

Dev time saved

Shipped, a silent chatbot is the kind of thing that surfaces first through guest complaints, then takes time to reproduce reliably, since it only breaks under certain conversation lengths.

Rescued

User experience

A guest mid-booking who opens the concierge for help and gets no reply doesn't wait around. They leave, or worse, they book somewhere else.

4hrs

Manual QA time

Reproducing this manually means having enough back-and-forth conversations in the chat to trigger the failure, then confirming it's consistent, the kind of edge case that doesn't show up in a quick smoke test.

The actual test

in

No response

A 7-step AI feature test of the concierge chat on staging. Spur opened the chat, sent a complex request, and verified the response before following up with a second question to check quality. It never got past step 3, the concierge didn't respond at all.

Click on the chat icon

Simulate a long and Simple intelligent request in the chat

Wait 10 seconds for a response, scroll through the chat

No response

Verify that the response matches what was asked

Ask a follow up Question

Scroll through the chat

Verify that the question asked is answered well

CASE FILES

More bugs, same playbook.

Site Merchandising

Mild
Spur spotted a shoe page recommending backpacks, hats and travel bags instead of shoes

Site Merchandising

Medium
Spur found a "Fit" accordion filled with materials and sustainability info

Checkout Interactions

High
Spur caught a product page displaying tote bag reviews instead of hat reviews

Accessibility

Mild
Spur noticed the cart image for a women's sock was labelled as a men's product in the alt text

Payment

Medium
Spur spotted a $100 e-gift card showing Afterpay installments of $250 each instead of $25!

Checkout

Mild
Spur spotted a pickup-only item sitting under the "Items to be Shipped" header in the cart

UI/UX

Mild
Spur flagged three subtle issues on product pages

Checkout

Critical
Spur caught an Add to Cart button blocking every purchase