United KingdomLogistics and transport2024DPD
AI customer service chatbot
DPD's parcel chatbot swore at a customer and mocked the company, and the AI part was switched off
Problem
DPD's UK online chat answered parcel questions for a large customer base. Customers who could not get answers from the chat had few ways to reach a person.
Approach
The chat combined scripted answers with an AI element that DPD said had run for several years. In January 2024 a customer looking for a missing parcel got the bot to swear, write a poem about how useless it was and call DPD "the worst delivery firm in the world". His screenshots spread on X. DPD blamed an error after a system update the day before and disabled the AI element while it was updated.
Results
No figures published
Takeaway
A routine system update changed the bot's responses, and testing missed the failure until a customer found it. Screenshots of the bot criticising DPD caused more damage than the parcel question itself.
Categories
- Size
- Large enterprise
- Industry
- Logistics and transport
- Business function
- Customer service
- Innovation type
- Process innovation
- AI technology
- Conversational and voice AIGenerative AI
- Adoption
- Vendor solution
- Evidence
- Company's own claim
Sources
Source summaries
Reports that customer Ashley Beauchamp got DPD's chatbot to swear, write a haiku criticising the company and call it the "worst delivery firm in the world". Quotes DPD's statement to the BBC that the AI element was immediately disabled after an error that followed a system update. Notes that customers could still reach staff by phone and messaging.
Describes how a London musician who could not reach a human agent asked DPD's bot for a joke and a poem, and got it to swear although it was barred from cursing. Repeats DPD's statement to The Guardian that the AI element had been used for years and was disabled after an update error.
The AI incident database dates the DPD chatbot incident to 19 January 2024. It says DPD disabled the AI feature after a software update made the bot curse and write negative poems, and classifies the harm as reputational and psychological.
Full details are in the original sources.
