Sotheby’s sold the original source code for the World Wide Web for $5.4 million — © AFP Martin BUREAU
No-code AI offers an AI-driven, machine learning-fuelled conversational interface for performing data analytics. Not needing to write code for queries coupled with an intuitive interface enables non-technical professionals to access data to explore, query, analyse, test hypotheses, and make predictions based on their data.
This flexibility reduces the need to rely on a team of data scientists to provide insights, allowing businesses to make more seamless data-driven business decisions. To gain an insight into the development of the technology, Digital Journal spoke with Kevin Chen, Chief Data Scientist of Experian DataLabs in North America.
Digital Journal: What is No-Code AI (Artificial Intelligence) System? And why is it important?
Kevin Chen: No-code systems provide a user interface that enables users to create application functionality without writing programming code. Conceptually, it’s like a “what you see is what you get” (WYSIWYG) website design HTML editor, offering a visualization layer on top of the code that makes manipulating that code much easier. No-code AI employs the same approach to data analysis and AI/machine learning, except in a far more sophisticated manner.
One of the breakthrough technologies from our research and development arm Experian DataLabs, no-code AI offers an AI-driven, machine-learning-fuelled conversational interface for performing data analytics. Not needing to write code for queries coupled with an intuitive interface enables non-technical professionals to access data to explore, query, analyze, test hypotheses, and make predictions based on their data. This flexibility reduces the need to rely on a team of data scientists to provide insights. The idea is to make data available to product managers, sales teams, store owners, or anyone else who would benefit from it.
DJ: Does no-code AI remove the need for data scientists?
Chen: No, because no-code AI still requires upfront coding to enable the behind-the-scenes presentation of data and intelligence. For example, it needs ML data scientists in the areas of natural processing/understanding/generation to develop the capability of understanding a user’s intent for any requested analysis. Also, no-code AI benefits from algorithms developed by data scientists to automatically identify trends and patterns or perform root cause analysis.
In short, no-code AI is only as good as the level of intelligence that is built (i.e., coded) into it. Even with practical business problems where an organization has a significant amount of domain knowledge about it, writing the backend code for the no-code-AI user interface can be a challenge due to the nuances in the data, how the data should be used, or some other unique application within the organization’s operations.
Until true AI that independently conducts analyses exists, users will still need basic awareness regarding what no-code AI can and cannot do. Asking it to predict consumer attrition is seemingly straightforward, except that the definition for attrition can vary based on the business application. For example, a bank customer virtually or literally stops using their bank account, yet it remains open. In practice, this scenario should be considered as attrition, but it is often not marked in the bank’s system as closed. Having no-code AI simply use a closed-account flag in the data as the intended target definition for predicting attrition would yield, in this instance, a less desirable outcome for the bank.
As no-code AI empowers non-technical users to gain insights from their data to make data-driven decisions, data scientists should remain a key partner in the journey to help validate any data assumptions that are made and guide the analysis. At the same time, the efficiency introduced by no-code AI will free the data scientists from performing mundane data analysis and routine reporting, so their valuable time can be directed to focus on complex, high-value business problems.
DJ: How do you properly set up a no-code AI system to enable this type of open use?
Chen: There are a few fundamental functional building blocks that are required, beginning with the capability to conduct common ML/AI tasks such as data ingestion, data cleansing, data QA, feature extraction, training models with various ML techniques, parameter search, and model evaluation.
Next, there needs to be a business-logic layer that enables no-code AI to assist users in solving specific types of business problems as well as performing simple generic ML tasks to make a prediction from a collection of data points, identify the contributing factors to a business opportunity or challenge, etc.
Then, the system should have an intelligent presentation layer that automatically creates the most appropriate data visualization to present requested information to the users.
To bridge these AI capabilities with voice activation for querying, no-code AI needs powerful natural language processing/understanding/generation (NLP/NLU/NLG) capabilities to interact with the user using plain English or other desired spoken language. This feature is necessary to understand a user’s question and invoke the right analysis and/or even directly translate the request into executable codes.
This functionality extends way beyond Alexa and Siri-level voice recognition and activation, or even the existing task-oriented conversational AI/chatbot solutions that can – as examples – look up weather conditions or book flights. It requires deep learning-based solutions to attempt to understand want users want, then correlate that user’s intent with the available data. The challenge is matching the no-code AI system’s domain knowledge of the data at its disposal with that of the user making a query. If a user believes the system should have the appropriate data and present it yet the query yields insufficient results, there will be a disconnect.
DJ: Please provide an example of how no-code AI can work.
Chen: Experian DataLabs continually looks at disruptive forces within the market that can benefit from the fusion of raw data and machine learning, such as offering greater accessibility to advanced data analysis, research and development for non-technical users.
To help facilitate this type of access and enable users to experiment with querying data, Experian DataLabs developed the Ascend Analytical Sandbox. It’s an advanced analytical tool that leverages nearly two decades of anonymized credit data from more than 220 million consumers, along with commercial data, property data, and other alternative data sources.
This deep, expansive repository of anonymized consumer data enables data scientists and non-technical decision-makers alike to explore behavior, trends, demographics, and much more. Experian’s Ascend Interact is a no-code AI conversational analytics agent that Experian DataLabs is developing to help users query and perform data analytics within this sandbox using plain English.
Ascend Interact uses deep-learning NLU and NLP to give business decision-makers the ability to interact directly with a massive trove of data, and potentially join it with their organization’s data, without having to pass it through a team of data scientists.
From a business intelligence perspective, the product can deliver dynamic data-driven charting and analysis based on what a user needs at the time.
DJ: Where does no-code AI go from here?
Chen: The promise of no-code AI and its applications is significant and will continue to power data analytics and insights for business decision-makers. The key will be integrating the technology into more and more business applications across different industries. Early adopters will likely come from areas where there is shortage of data scientists/data analysts and domain knowledge needed for analysis can be easily embedded in the no-code AI. As the technology evolves and a new wave of non-technical professionals begin to understand its potential to drive business growth and stay ahead of potential risks, adoption of no-code AI systems will flourish.