The decision looked rational on paper. A home goods retailer with about 40 staff replaced its five-person phone support team with an automated chat system costing $180 per month. The expected savings were roughly $14,000 monthly in wages and overhead.
What the spreadsheet did not capture
The chatbot handled order status queries adequately. Anything involving damaged goods, missing items, or billing disputes sent customers into a loop of scripted dead ends. Refund requests that previously resolved in under eight minutes now averaged four days.
The support inbox accumulated unanswered tickets. The team that remained, two part-time staff managing social media, had no authority to issue refunds or escalate to logistics. Complaints began appearing on Google reviews and a consumer forum dedicated to Canadian home retailers.
Six months of silent damage
The owner, Petra Wulff, noticed the savings clearly. She did not notice the repeat purchase rate dropping from 34% to 19% over the same period until an accountant flagged declining cohort revenue in a quarterly review.
Restoring a hybrid model, one dedicated support agent plus the chatbot for basic queries, cost more than the original team because hiring and retraining took three months of overlap pay. The net loss for the year exceeded the 12-month savings the chatbot was supposed to generate.
The decision point that mattered
Petra later identified the specific mistake. She compared vendor quotes for the chatbot against current staff wages. She did not map which customer problems actually required human judgment. Roughly 60% of incoming contacts involved exceptions the chatbot could never handle.
Budget decisions in customer service work differently than supply chain cuts. The cost of a failed interaction does not appear on an invoice. It appears months later in a customer who simply does not return.