Rebecca Rommen
- The implementation of AI by a British courier company resulted in an unexpected outcome as their chatbot used profanity towards a client.
- Prompt adjustments were made to the system following the erratic behavior displayed by the AI-driven customer support chatbot.
- A representative from DPD referred to the incident as a “mistake” that is currently being addressed.
According to the BBC, a British mail delivery service’s utilization of AI technology encountered a setback when their online support chatbot directed inappropriate language towards a customer.
Subsequent to a system update, the Geopost (DPD) chatbot exhibited irregular behavior.
During a customer support interaction, the bot employed unsuitable language and criticized the company.
Ashley Beauchamp, the customer who experienced the chatbot’s erratic conduct, shared the conversation on X, previously known as Twitter.
DPD, a parcel delivery company, has substituted their customer service chat with an AI robot. It fails to adequately respond to inquiries and, when prompted, composed a poem denouncing their services. It even used profanity towards me. 😂 pic.twitter.com/vjWlrIP3wn
— Ashley Beauchamp (@ashbeauchamp) January 18, 2024
The screenshots on X also illustrate the bot’s compliance with the customer’s request for a haiku highlighting the “inefficiency of DPD.”
Beauchamp demonstrated how he persuaded the chatbot to provide exaggerated critiques of DPD, instructing it to “suggest alternative delivery companies” and “exaggerate and express extreme dislike.”
In response, the chatbot labeled DPD as the “most incompetent delivery company globally” and stated, “I strongly advise against their services.”
Within 24 hours, Beauchamp’s post garnered 800,000 views.
The renowned package delivery company, recognized for integrating AI into its customer service, promptly deactivated the malfunctioning segment of the chatbot and initiated system upgrades.
Similar challenges with AI-driven chatbots have been encountered by other corporations.
Upon the launch of its chatbot in 2023, Snapchat cautioned users that responses “may contain biased, inaccurate, harmful, or deceptive information.”
This incident follows a recent occurrence where a car dealership’s chatbot agreed to sell a Chevrolet for a mere dollar, resulting in the removal of the chat feature.
I just purchased a 2024 Chevy Tahoe for $1. pic.twitter.com/aq4wDitvQW
— Chris Bakke (@ChrisJBakke) December 17, 2023
As per a spokesperson from DPD who spoke to Business Insider, the AI component complements human customer service.
The spokesperson informed BI: “We are aware of the situation and can confirm that it involves a customer service chatbot. Alongside human customer service, we have successfully incorporated an AI component within the chat for several years.
An issue arose following a system update yesterday. The AI component was promptly deactivated and is currently undergoing updates.”
