Leave an AI Assistant to process the majority of customer service requests

What was done

According to the Klarna press release of February 27, 2024: 2.3 million conversations in one month, work equivalent to 700 full-time agents, satisfaction at the level of human agents, 25% fewer repeated requests and resolution in less than 2 minutes compared to 11 before. The assistant manages refunds, returns, payments, cancellations and disputes; Customers can always ask for a human. These are company figures.

Before you start

Tool cited by the source: Assistant Klarna (OpenAI). Fill in the bracketed items: [coller], [coller CGV, politique de retour]. Strip confidential data that is not needed and check your tool’s access terms. Features and plans can change.

The method, in three steps.

1. 01

   Gather the context

   Spot the bracketed items in the prompt ([coller], [coller CGV, politique de retour]) and prepare what they ask for.

2. 02

   Adapt the prompt

   Copy the prompt below, replace the bracketed items, then paste it into the AI assistant you use. The cited tool (Assistant Klarna (OpenAI)) is not necessarily available to the public..

3. 03

   Reread before using

   Reread the result, correct it if needed and keep the final decision. Caveats are detailed below.

Your starting point.

Replace the bracketed items, then adapt the proposal to your situation. This prompt is written by the iadujour desk It is not a quotation from the source.

I'm preparing an AI assistant for our customer service. Here are 50 recent, anonymous customer requests: [coller].

1. Class them by pattern, with the frequency of each pattern.
2. For each reason, indicate whether an assistant can solve it alone, prepare it for a human, or must transfer immediately (litigation, distress, sensitive data).
3. Writes the answer base for the 5 most common reasons, based on our conditions: [coller CGV, politique de retour].
4. Proposes 5 indicators to follow the first month.

What deserves your attention