EN
The innovation arm of a Fortune 500 retailer wanted to use AI and Natural Language Processing to enhance their ordering and customer service systems. The goal was to use natural human speech as means to interact with multiple backend systems of an enterprise organization. This included order management systems, inventory management systems, package tracking systems, and customer service interfaces.
consumer goods
2 years
NLP models, AI, speech-to-text, arduino/pi
The vision was to be able to say:
Hey, System, order me more blue pens!
Or
Hey, System, where is my order!?
The system had to interpret human spoken or written words, interface with back-end enterprise systems and respond in a natural way.
There were a multitude of challenges with this unique deployment. These ranged from hardware interfaces to take human speech, speech to text translation, natural language models.
Some of the most challenging aspects of the project included the many back end enterprise systems that needed to be integrated with. In order to fulfill on meaning requests from the customer, nearly “all” of the core systems, some of them legacy required integration.
Additionally the language taxonomy of the million plus customer SKU database was an extreme challenge. Most Natural Language models and process do well with “well known words”. However unique NLP configuration was required to create a language model so that the customers product catalogue could be understood by the system.
We created a “natural language” orchestration engine. This orchestration engine took in human speech from a variety of channels (Phone, Custom Hardware, Web Chat, call center recordings, etc).
The orchestration engine then took the customers speech and passed it to a natural language processing model to understand intent.
Complex dialogue flows were configured in a conversation decision tree that KIS engineered.
When required, the orchestration engine would interface with backend enterprise systems (to lookup products or orders for example.). The enterprise API responses would be interpreted by the system, and then a response would be formulated and returned to the end user over the required channel.
This system currently automates thousands of chat conversations on a daily basis, and auto-classifying call center contacts for a 3000+ person call center operation.
Like many of our projects, this one was an absolute pleasure to work on!
This was a truly “once in a lifetime” project, and one of the most challenging and interesting initiatives that we have worked on.
tech strategy & consulting.
Empowering companies, investing in IT, tech training, scalable squads, and delivering the right value.
data
management.
Data migration, conversion, analysis and integration.
Database consolidation and management.
machine learning models.
Use AI algorithms and data to improve the accuracy of your analysis and make better predictions that impact the decision making processes.
web services development.
From discovery to delivery, we solve issues that go beyond the basics.
