Building a next generation AI platform

How CA Jarvis is using robotics to enhance UX, simplify human tasks

Creating Outstanding Customer Experiences Requires Data Analytics

Everyone is unique. Creating outstanding customer experiences means understanding these differences, and providing personalized connections to build trust and develop positive relationships.

While many elements of the customer experience are already automated, nowadays companies are taking it a step further. Recent advances in Big Data have enabled the processing of tremendous amount data, needed to personalize services that precisely target customer needs. Thus, many businesses are exploring how to use artificial intelligence (AI) and machine learning (ML) to generate a whole new type of experience – one that involves adaptive, self-learning robots.

Is this the right time to use robots for personalized customer service, to minimize low-value and repetitive tasks? Possibly. The answer lies in whether Big Data, AI and ML can create memorable customer experiences.

Figure 1: Bill-B1 welcomes the customers at the entrance of the bank.

The CA Jarvis-Leo Robot Experiment

Driven by this possibility, CA Jarvis recently partnered with Leo Robot Ltd. to explore whether Jarvis-empowered Leo robots could enhance the experience of customers – in this case, at a local bank.

Leo’s service robot, called Bill-B1 ( high 55 inch, weight 97 pound), comes equipped with many useful technologies, such as infrared credit card readers, an HD camera, LED eyes, a thermal printer, LCD screen, laser ranging radar and even crash sensors. Bill-B1 can speak with customers in multiple languages, engage them with basic questions, read bank cards, perform facial recognition and even manage data queueing. These attributes make Bill-B1 a quite useful store assistant. But is he capable of more?

In a large banking corporation, new promotions arise almost every day. The business office managers are normally tasked with making recommendations to customers based on their preferences and qualifications. However, without comprehensively analyzing each customer profile, it’s hard to give personalized recommendations to customers. Could Bill-B1 do a better job?

  
Figure 2: (left) shows the integration of CA Jarvis dashboards on Leo robot’s LED display (10.1’’IPS LCD Screen).  (right) Bill-B1 (empowered by Jarvis) is having conversations with customers in bank.

 

The short answer, is yes.

When engaging with exsisting customers, Bill-B1 is able to read  their banking cards and access their user profiles. The conversations between customers are streamed into Jarvis and then useful information can be processed by Cloud based NLP services. Then multiple pre-trained models ( including neural networks and ensemble learning based algorithms ) that have been triggered to generate recommedations to users. The prediction results are further constructured to robots’s language for response.  Also, CA Jarvis’ dashboard empowers Bill-B1 to present useful insights, such as the estimated customer wait time, hourly store traffic volume, exchange rates and investment trends… not to mention the personalized promotion recommendations.

 Data Science Project Learnings

In launching this project, we have learned that the major effort in creating a successful data science application is not just optimizing the models but also putting these models into production. Choosing the right AI platform is a critical and ubiquitous for not only data scientists, data engineers and architects, but also for operation teams and project managers. The next generation of AI platforms should be able to handle various types of data flows, support both real-time and batch-mode applications, manage entire model life cycles, as well as elastically scale run-time resources on demand of data and applications.

Specifically, a successful AI platform should:

  • have comprehensive support from data collection, data integration, and data processing for both structured and unstructured data.
  • target model lifecycle management including model development, model turning, model deployment, and model replacement (online)
  • provide flexible (Cloud vs. On-premise) and elastic scaling (up and down) runtime for data science applications on both development and on-production settings
  • have standard service APIs for the integration between AI/ML algorithms and common robot controlling services like security, multi-media and networks.

Our AI Platform mission

The CA Jarvis team is working to address all the challenges inherent in creating a top-notch AI platform. We are excited for upcoming enhancements, such as Microservices JAF, which will provide flexible architecture to support various deep learning frameworks; the integration of Jupyter notebooks with JAF, which will allow data scientists to operate data and models directly with Jarvis, enabling  model evaluations and performance metrics in both notebooks and CA Jarvis Dashboard.

Figure 3: (left) The exploration of new customer experiences continues, as does the research of robotics reasoning and emotion modeling. (right) Leo Robot is tackling new challenges by leveraging CA Jarvis’ next-gen AI platform.

 

Meanwhile, Leo Robot continues to work with banks to advance their exploration of robot-led banking services. They are also exploring new business lines from health care, to retail, transportation and tourism, etc.  Our dream is that one day, the next generation of CA Jarvis will power Leo Robot to revolutionize the customer experience.


Cui Lin, PhD, is a Senior Data Scientist at CA Technologies, working on predictive analytics,…

Comments

Modern Software Factory Hub

Your source for the tips, tools and insights to power your digital transformation.
Read more >
RECOMMENDED
How to Stay in the PinkWeWork's VP of Engineering on the Big Difference the Little Things MakeManufacturer's Reimagined Processes Enable a Step Change in Growth