Providing Superior Customer Service in Hospitality with Generative AI

Mad Mobile scales customer support and reduces operational costs by speeding up ticket resolution with a next-generation Generative AI-powered Customer Service Agent

Home » Case Study » Providing Superior Customer Service with Generative AI

Mad Mobile is a leader in point-of-sale (POS) devices, serving the hospitality industry. Mad Mobile's platform and products empower businesses to seamlessly integrate mobility into their existing systems, digitize guest experiences, and boost sales. The platform’s technology is designed to meet evolving industry demands, enabling businesses to offer superior service to guests while enhancing staff efficiency. Mad Mobile's solutions have been adopted by over 21,000 stores worldwide, benefiting approximately 280,000 users across 26 countries.


Mad Mobile wanted to modernize its platform by incorporating a next-generation, Generative AI-powered customer service agent into its operations. Designed to provide quick, accurate, and contextually relevant responses to their human customer support agents, the agent would help Mad Mobile scale their customer support and reduce operational costs through faster ticket resolution, all while taking customer-client interactions to the next level.


Provectus developed a Gen AI-powered customer service agent utilizing a comprehensive technology stack. At its core are Amazon’s and Cohere’s Foundation Models (FMs), hosted on Amazon Bedrock, selected for their scalability, flexibility, and high-quality text generation. Designed for easy integration with and on AWS, the agent and its components serve as a foundation for various AI use cases.


Provectus successfully completed the Use Case Discovery & Design Workshop and the Proof of Concept (PoC) phases. In the PoC phase, the customer service agent achieved a 70% accuracy rate in responses, with a 30% improvement in response precision due to advanced prompting. Currently, Mad Mobile is evaluating the integration of this solution into its operations, while Provectus is gearing up for the next phase of collaboration.

6 weeks

From the project’s Discovery to a Proof of Concept (PoC) implementation


Accuracy in responses generated by the Customer Service Agent


Increase in the number of correct responses achieved with prompting


A New Opportunity for Transformation: More Efficient Customer Service as a Critical Component of Success in the Hospitality Industry

Mad Mobile, a leader in point-of-sale (POS) solutions, is known for its expertise in integrating seamless mobile experiences into existing customer- and client-oriented systems. Their technology caters to the evolving needs of the hospitality industry, enhancing service quality and operational efficiency.

generative ai use case for customer support

The leaders of Mad Mobile recognized the need to stay at the forefront of technological innovation, enhancing and modernizing their existing hospitality platforms and software with Artificial Intelligence (AI). Their vision prioritized cultivating deeper customer-client relationships through seamless service experiences, while enhancing the efficiency of their own customer support and sales teams through automation and data-driven insights.

Mad Mobile sought a starting point for adopting AI. Provectus, an AWS Premier Tier Services Partner, was introduced to Mad Mobile by AWS, to assist in exploring high-impact Generative AI use cases that could help Mad Mobile’s leaders realize their vision. In collaboration with AWS, Provectus delved into Mad Mobile’s business and solutions, identifying a tangible entry point: a Generative AI-powered Customer Service Agent.

By integrating the Gen AI Customer Service Agent into their operations, Mad Mobile would enable their customer support services to provide rapid, precise, and context-aware responses to their clients. For example, a customer service representative can now provide an accurate, fast response to a client struggling with Wi-Fi settings for a PoS device without taking up time searching for internal documentation and crafting a response from scratch.

For Mad Mobile, faster and more efficient ticket resolution means streamlined and scaled customer support operations, reduced operational costs, and superior customer service.

In addition to the Customer Service Agent, Provectus and Mad Mobile plan to develop a Conversational UI module that would allow for seamless integration with Salesforce, leveraging Generative AI for email interactions and helping sales associates in their communication with customers.


The Generative AI Journey with Provectus: From Discovering Possibilities to Building a Customer Service Agent on AWS and Cohere

The Mad Mobile project consisted of two primary phases: a Use Case Discovery & Design Workshop, followed by a Proof-of-Concept (PoC) development and implementation.

The Discovery & Design phase involved examining Mad Mobile’s high-level business context including the company’s objectives, challenges, and initiatives. It included identifying the needs of Mad Mobile’s customer support team, establishing the best value proposition, and defining success criteria for the project.

As an outcome, the Provectus team set out to deliver a comprehensive Gen AI Use Case adoption roadmap, outlining the expected business value (e.g. a 10% reduction in support agent response times), and the complexities and risks associated with the project. This also involved defining the data sets to be used, along with the responsible teams, and designing a high-level solution architecture on AWS and Cohere. The roadmap included a development strategy for the next phase and preliminary Total Cost of Ownership (TCO) estimates for the AWS infrastructure and Cohere services.

During in-depth discovery sessions, Provectus reviewed multiple hypotheses to improve Mad Mobile’s customer support operations using Generative AI, and concluded that a Customer Service Agent would be the best initial application of this technology.

The Proof-of-Concept (PoC) development and implementation phase showcased how the Gen AI Customer Service Agent (and its most important components) could be implemented in a non-production environment. Once deployed, the Provectus team would assess the solution’s feasibility with Mad Mobile’s customer support team and evaluate the effectiveness of applying Generative AI technology for the selected use case with the available data.

The Customer Service agent and its accompanying components were designed, built, and implemented using a comprehensive suite of technologies, including Python, LLamaIndex, FastAPI, Swagger, among others. Adhering to Clean Architecture and Hexagonal Architecture principles, along with the Mediator Pattern, the Provectus team prioritized modularity and flexibility as keys that would help Mad Mobile scale AI across their organization.

At the core of the solution are Foundation Models (FMs), hosted and served via Amazon Bedrock, Amazon Titan, and Cohere’s LLM Embeddings on AWS. These models and services were selected for their scalability, flexibility, and high-quality text generation capabilities. The solution’s performance was evaluated using such metrics as ROUGE and BERTScore. In total, Provectus tested over ten different hypotheses and applied advanced prompting methods, which helped improve the correctness of Gen AI responses by 30%.

Recognizing that Mad Mobile was at the initial stages of its Generative AI adoption journey, Provectus tailored the solution for effortless integration with AWS services. This approach was complemented by extensive documentation, facilitating the solution’s deployment on AWS, and included Docker support for added versatility.

Overall, Mad Mobile was provided with a Proof of Concept (PoC) deployed in their development environment, TCO estimates for AWS and Cohere services, knowledge transfer materials, and a development plan for the next stage.


Transforming Customer Service with Generative AI: The Results of Mad Mobile’s Cooperation with Provectus and Next Steps

The first phase of collaboration between Mad Mobile and Provectus was primarily focused on speeding up and enhancing ticket resolution, to improve customer satisfaction. It also enabled Mad Mobile’s customer support to become more efficient in driving upsell and cross-sell initiatives. This collaboration resulted in a successfully completed PoC with a plan for going into production.

Over a span of six weeks, Provectus developed and tested the Generative AI-powered Customer Service Agent and further improved the correctness of responses using advanced prompting techniques to achieve 91% total accuracy in responses.

The in-house trial phase highlighted the solution’s capacity to:

  • Streamline and scale Mad Mobile’s customer support
  • Reduce operational costs by accelerating ticket resolution
  • Provide assistance in onboarding support team members
  • Benefit Mad Mobile’s internal teams and enhance customer-client interactions

The solution is anticipated to be a powerful addition to Mad Mobile’s sales toolkit, helping to enhance upselling and cross-selling strategies across different industry verticals.

The leaders of Mad Mobile were impressed by the Provectus team’s ability to execute quickly and their experience with the Generative AI suite on AWS including Amazon Bedrock, Amazon Titan, Amazon SageMaker, and Cohere’s LLMs.

For Mad Mobile, successful collaboration with Provectus set the stage for transformative advancements in customer service and operational efficiency. Looking ahead, Mad Mobile is well-positioned to leverage Generative AI & Data to build seamless mobile experiences, continuing their quest for innovation in the hospitality industry.


Moving Forward

  1. Learn more about Gen AI’s potential from business and technology perspectives in The CxO Guide to Generative AI: Threats and Opportunities
  2. Assess the readiness of your organization to adopt Gen AI solutions with Generative AI Readiness Assessment with Provectus
  3. Develop high-value Generative AI use cases with Generative AI PoC with Provectus and Cohere


Looking to explore the solution?

  • Hidden
  • Hidden
  • This field is for validation purposes and should be left unchanged.