My summary of Yellow AI before using it — I believe Yellow AI offers a machine learning model that underpins chatbot conversations. Not sure yet about the use cases are for this machine learning model.
In the interests of full disclosure, I work at Intercom where we enable our customers to engage effectively and satisfactorily with their customers, through proactive, self-serve and human support engagements. Enabling consumers to self-serve their questions or issues through the use of bots is a daily focus for me, and I need to stress that my views on Yellow AI are solely my own and do not represent my employer.
How does Yellow AI explain itself in the first minute? — When I go to https://yellow.ai/ I read that it Yellow AIO is an “enterprise-grade conversational AI platform”. This description is followed by a number of messages that rotate above the fold: “Offering Resolution Focused Automation”, “Unlocking Business Potential at Scale” and “Delivering the Best of AI+Human”.
However, the message that resonates more with me reads “For Delighted Customers and Happier Employees”. This to me suggests that Yellow AI specialises in improving user experience through AI.
Indeed, when I scroll down the homepage I then understand that Yellow AI offers an actual chatbot, which is positioned as a “Dynamic AI agent”.
The Yellow AI platform offers a number of features that will serve both businesses and their customers. There’s the ability to convert conversation insights into a bot (Insights Engine) and businesses can integrate Yellow AI with ecommerce platforms like Shopify or social channels like WhatsApp (Integrations).
How does Yellow AI work? — Recently, Yellow AI released DynamicNLP which aims to change the way in which bots are trained. The goal here is to reduce the amount of human time and effort involved in training a bot. Companies can thus provide more efficient and seamless customer support.
Artificial Intelligence (AI) can be broken down into supervised and unsupervised machine learning. Unsupervised learning is a type of algorithm that learns from untagged data. In contrast, with supervised learning the data is tagged by an expert. These labeled data sets are used to train algorithms to classify data or predict outcomes accurately.
- Zero-shot learning: DynamicNLP is based on Zero-shot learning; a predictive technique whereby a ‘learner’ observes samples from classes — which aren’t observed during training — and needs to predict the class that these samples belong to. For example, given a set of images of animals to be classified, along with supplementary textual descriptions of what animals look like, an AI model which has been trained to recognise horses, but has never been given a zebra, can still recognise a zebra when it also knows that zebras look like striped horses.
- Learning on the fly: Yellow AI explains that with DynamicNLP learners or dynamic AI agents can ‘learn on the fly’ — being able to learn and self-adjust to new patterns as the machine spots them. To do this effectively, I can imagine that pre-defined and structured descriptions still need to be in place for machines to successfully learn on the fly. For example, if a company deals in different types of office furniture, specific attributes of office furniture descriptions need to be pre-defined to enable the machine to learn continuously from textual descriptions and improve its understanding. From what I’ve seen sofar, Yellow AI feels similar to the email composer Compose AI in this respect.
- Setting up conversational flows: According to Yellow AI, its DynamicNLP model has ‘seen’ a lot of different syntactic variations of sentences from billions of conversational data. Thus making it easier for the machine to understand conversation context and the intent of a user’s query.
- Seamless customer support: Yellow AI’s overarching proposition is its conversational AI platform through which companies can provide automated customer support. A conversational platform is a type of AI that allows people to interact with computers in a way that mimics human conversation. A chatbot is the most well known and tangible output of such a platform. With DynamicNLP, Yellow AI says that businesses can “deliver a seamless customer experience with a better understanding of the context and intention of their queries.”
Main learning point: Keen to see the impact of Yellow AI’s DynamicNLP approach; the results it achieves in reducing the need for human effort to train a machine as well as the accuracy of its self-adjusted learning approach. From a systems point of view, I’d like to better understand how Yellow AI’s bot interacts with 3rd party systems and data repositories.