Talking the Talk: Retailer Uses Conversational AI to Help Call Center Agents Increase Customer Satisfaction


With more than 11,000 retailers throughout Thailand serving millions of consumers, CP All, the country’s sole certified operator of 7-Eleven benefit retailers, recently turned to AI to dial up its contact centers’ assistance capabilities.

Built on the NVIDIA conversational AI platform, the Bangkok-dependent company’s consumer service bots support get in touch with-centre brokers reply routinely requested concerns and track consumer orders. The bots understand and communicate Thai with 97 percent accuracy, according to Areoll Wu, deputy basic supervisor of CP All.

This variety of innovation is a important benefit for CP All, which partners with a number of field teams and countrywide businesses on an yearly awards system to persuade investigate and entrepreneurship in Thailand.

CP All’s technique makes use of NVIDIA DGX techniques and the NVIDIA NeMo framework for purely natural language processing teaching, and the NVIDIA Triton Inference Server for AI inference and design deployment.

Preserving Up With the Phone calls

CP All’s seven-Eleven get in touch with facilities regulate shopper inquiries in quite a few enterprise domains — like e-commerce, finance and retail — which each have region-specific professional reps. The facilities commonly get nearly 250,000 calls a working day, in accordance to Kritima Klomnoi, venture supervisor at Gosoft, a subsidiary of CP All.

“Reducing keep time for prospects is a critical evaluate of our support efficiency,” Klomnoi explained. “NVIDIA technologies supply us a 60 percent reduction in the contact load that human brokers have to tackle, allowing workforce to efficiently tackle a lot more exclusive and elaborate issues raised by shoppers.”

Applying AI-pushed automated speech recognition services, CP All’s customer phone phone calls are transcribed in true time. When a consumer company bot recognizes a dilemma based on the NVIDIA-powered clever FAQ program, it instantly supplies an solution utilizing text-to-speech technologies.

Or else, the AI immediately analyzes and routes phone calls to the suitable staff who can aid in resolving the question in its particular company area. CP All has also automated all e-commerce order-monitoring inquiries using AI.

Adapting to the Thai Language

When first discovering conversational AI, the CP All team confronted the problem of having the model to acknowledge the nuances of the Thai, Wu claimed.

Common Thai uses 21 consonants, 18 pure vowel appears, 3 diphthongs and 5 tones — producing it a complicated language. NVIDIA NeMo — a framework for setting up, training and fine-tuning GPU-accelerated speech and natural language being familiar with versions — served CP All function by means of the intricacies.

“The toolkit’s pretrained models and resources created the system of deploying our services significantly considerably less daunting,” reported Wu. “With the enable of NeMo, we have been equipped to quickly make and improve our AI language models, which are now optimized to fully grasp and discuss the special Thai language.”

According to Wu, the NeMo framework enabled a 97 percent accuracy in CP All’s Thai language versions, far more than tenfold the accuracy reached formerly.

On the lookout ahead, CP All ideas to grow its AI services to additional organization domains and scale to hundreds of thousands of concurrent periods on NVIDIA GPU inference architecture.

Learn extra at CP All’s panel at GTC, working March 21-24.

Leave a comment

Your email address will not be published.