Premium Concierge Service John Paul Delivers Unforgettable Customer Experiences with Snowflake

VIP passes to a fashion week show. Tasting sessions with a top chef. Front-row seats at a Grand Prix race. Since 2008, John Paul, now a wholly-owned subsidiary of the ACCOR group, has been a world leader in premium concierge services and custom relationship marketing programs. Its 400-strong global workforce is committed to helping brands develop meaningful, tailor-made experiences with their clients.

With service quality at the heart of its business, John Paul relies on qualitative and quantitative data to ensure faultless operational efficiency and continuing customer satisfaction. To help concierges improve their responses to requests and queries, John Paul manually analyzed certain types of interactions using random samples of communications, or certain instances of extreme customer satisfaction or dissatisfaction.

To automate and systematize their analysis, and study them en masse, the data team decided to explore a solution based on AI, capitalizing on existing data assets with Snowflake which was capable both of transcribing telephone conversations (speech-to-text) and of analyzing interactions on all communication channels available to users of their concierge services (mail, voice, WhatsApp, etc.).

As a provider of high-end concierge services, John Paul focuses on major accounts covering many sectors, such as banking and car sales. And these companies entrust John Paul with the needs and requirements of its own clients, essentially integrating the  concierge service into the brand. So service quality is vital to ensure accurate reporting for clients, but also to guide its operations, which impact everything from the continuing professional development of its teams to the training of newly recruited concierges. 

The tenor of communications between end customers and concierge, among other factors, are very detailed and provide a fairly accurate measure of customer satisfaction. Until recently, this had been based on random sampling or on specific instances relating to unusual feedback from a client. John Paul wanted to be able to automatically analyze all client interactions on all communications channels, especially voice ones as number one preferred channel.

For the data team responsible for the technical aspects of the project, relying on Snowflake was a given.  “We have used the Snowflake platform to manage all our data for several years,” explains Sanyat Mohamed, Data Scientist at John Paul. “Using the built-in, prefab Snowflake solutions, we have been able to respond fully to business needs.”

The first challenge to confront was getting written transcripts of telephone conversations. The team used Whisper from OpenAI, with Snowpark Container Services helping with its integration. Similarly, the availability of all data in the Snowflake platform itself meant that testing of the various large language models (LLMs) was simplified in order to identify the best solution to meet their needs.

“Conversations, especially spoken ones, can be very lengthy,” explains Sanyat Mohamed. “The LLM we chose had to be able to handle that.” After testing both Mistral Large and Llama, the team opted for Llama. “Payment-on-request via Snowflake — and the availability of our data already on the platform — made testing easier and quicker. It was also easier to assess the costs of each LLM we tested,” Mohamed says.

Since so much data was already on the Snowflake platform, it only took two months for the data and professional operations teams to build an initial version of a tool, which naturally led on to other projects running in parallel. 

For Mohamed, the autonomy that the Snowflake platform offered was key to delivering quickly on the project. Its artificial intelligence solutions like LLMs and the ease to deploy a speech-to-text solution with Snowflake Container Services, have the advantage of use-on-demand, significantly reducing both costs and complications for users.

But the data scientist also emphasised data security. “We are working in an especially restrictive environment in terms of regulations,” she says. “Being able to process data entirely within our Snowflake environment means their security and our regulatory compliance are guaranteed.” And this includes creating interfaces, which the John Paul data team did using Streamlit, also part of the Snowflake environment.

This environment and these fully-integrated tools allowed the luxury concierge service’s data and business teams to focus first and foremost on their project value without wasting time or energy on technical design.

John Paul has seen immediate benefits. Communications transcription and sentiment analysis are now on the way to be fully automated thanks to Snowflake, which is going to free four hours a week for the Quality Insurance team so that they can re-invest in contacting customers and further elevating their experiences.

Additionally, the depth of the available analyses are having a positive impact on the service quality work. From now on, all customer interactions are monitored continually, allowing operational teams to respond quickly in the event of any malfunction — the team has already seen a more than ten percent increase in call volume made by the quality team. This improves training quality for current concierge teams or new recruits, thereby providing premium B2B clients with a service that fully matches their brand image at all times.

Looking ahead, John Paul is planning to increase the presence of AI in its organization to further optimize their operations and augment their Concierges with AI copilots, as well as more personalized customer experiences through newsletters and digital portals — all in service of John Paul’s “money-can’t-buy-this” hospitality approach.