AiThority Interview with Arijit Sengupta, CEO and Founder at Aible

Hi, Arijit. Welcome to our Interview Series. Please tell us a little bit about your journey and what inspired you to start at Aible.

I came to the U.S. to study artificial intelligence at Stanford University because in those days you could not study AI at the bachelor level in India. Eventually, I went to Harvard Business School where my research led me to create BeyondCore – the first augmented analytics and data discovery company to automate analytics with AI – which was acquired by Salesforce and became Salesforce Einstein Discovery.

Post-acquisition and after completing over 1,000 AI projects throughout my career, I wrote a book called AI is a Waste of Money where I identified common points of AI failure. This led Harvard to bring me in as an executive fellow to co-create and co-teach a course on AI for Market Facing Functions. The course was designed to help MBA students figure out how to get economic value from AI. A concept that I carried over to the creation of my current company, Aible – the only enterprise AI solution that guarantees business impact from AI in 30 days or customers don’t have to pay.

Throughout this journey, my focus has been on how we can empower people through the use of AI. When I started Aible, all of the companies were paying lip service to the idea of the citizen data scientists. But in reality, all of the software was so hard that unless you were an expert data scientist, you couldn’t use it productively. Aible’s tagline is “I am Aible” because we believe everyone needs to be able to create and adjust AI to serve their needs, as opposed to being guided by AI created by others. In a world where only the ones creating the AI have the power, everyone else becomes disempowered. 

Recommended: AiThority Interview with Andy Pitre, Executive Vice President of Product at HubSpot

What are your core offerings? How do you incorporate various advanced technologies to create your products for large and mid-sized companies?

Aible is a cloud-based Auto-ML platform with three core offerings. The first one is Aible Sense – an augmented data engineering solution that lets the AI automatically handle all of the data cleansing, feature creation, and data prep work that is typically done manually. The human can adjust it as needed, and the system will learn from it so the human doesn’t have to do it a second time.

The second offering is Aible Explore – an augmented analytics product that automatically looks at millions of questions in your data and shows you the key insights you need to see. In the time a human can ask the analytical system a handful of questions, AI can ask a million and save the answers. Aible gives people the fundamental ability to find the unknown unknowns – the questions they didn’t think to ask – while also massively reducing their analytics costs. For example, we recently did a benchmark with Google where a Fortune 500 company was able to analyze 75 data sets, 100M rows, across 150M variable combinations, for a total cost of less than $80.

The third is Aible Optimize – an augmented data science and machine learning capability that helps users understand what a predictive model would do for their business outcomes such as revenue, costs, churn, etc. It understands the benefit of a correct prediction, the cost of an incorrect prediction, and capacity constraints such as budgets, and uses that information to craft a model that helps the business achieve its goals. Instead of forcing people to speak AI or data science, the product speaks the language of business.

All of these technologies have been used at several of the Fortune 500s – including seven of the Fortune 50 today – and many mid-sized companies.

What is the opportunity for organizations when it comes to utilizing AI and Machine learning in their operations? How does that specifically play out with supply chain optimization? 

We’ve entered a world of uncertainty where the business environment is constantly changing amid supply chain disruption, inflation, and political issues. It is very hard for people to notice and react to change quickly, and traditional AI actually isn’t designed for that either. Traditional AI assumes the future will look like the past and is predicting based on what it has learned, but this can’t help you if the situation is constantly changing.

The main opportunity for organizations in utilizing AI and machine learning is in scenario planning. Imagine an AI model that is optimal for a supply chain disruption out of your facility in China, or another one that is optimal when the exchange rate changes in a certain way. By scenario planning and building AI models that are optimal under different circumstances, you can be more prepared for the future and adjust your AI to the right setting for optimal value when disruptions occur. 

Recommended: AiThority Interview with Chris Maeda, Co-Founder & CTO at Botco.ai

ChatGPT and Google’s Bard have taken the tech industry by storm. Could you let us know how generative AIs could transform intelligent automation?

Generative AI is a transformative moment for this market and multimodal AI will completely change intelligent automation, but before you get over-excited, understand its limitations. Today, large language models like ChatGPT and Bard have hallucinations, meaning they can literally make up facts out of thin air. This was in full display in both the Google and Microsoft demos of their respective technologies. For instance, a Bing demo was intended to develop a pros and cons list for certain products and create an itinerary plan for a trip to Mexico City, but the product reviews were fabricated and the travel plans didn’t match up to the websites they were based on. 

The inaccuracy of LLMs can be extremely dangerous if it’s the human’s responsibility to double-check the facts before acting on them. In reality, people won’t fact-check their results. They’ll expect the technology to work since it works most of the time and act on it until they show up at a resort that has been shut down for several months, or worse.

Instead of placing the responsibility on the human, there is a different kind of AI called information models that can be used to double-check the facts of large language models. Information models are descriptive, diagnostic, and predictive, meaning they have an understanding of what has happened in the data, what is likely to happen, and why. By using information models to double-check large language models, the technology can scale with trust.

What are your views on the future of supply chain optimization with ChatGPT and other AI techniques?

In the past, the supply chain was optimized mainly on cost. The world was stable and we were trying to get every bit of marginal cost out. But one of the most important lessons we learned from COVID-19 and recent political disruptions is that as the world constantly changes some things matter more than just the cost of inventory. We need to balance it with the potential of stockout due to shifting demand or increased transportation costs from route closures. AI is very good at considering multiple, and potentially conflicting, objectives such as balancing inventory carrying vs stockout costs or low-cost vs low-risk providers. By scenario planning and testing the implications of different situations, you can create a more complex and responsive system rather than one that is static and fragile, so you can react to market changes. By using AI in combination with optimization of key business indicators and scenario planning, you’ll build a responsive and resilient system that can balance change faster than people alone.

Do you foresee any challenges brands will deal with as a result of AI based optimization and automation?

The biggest problem that people are going to run into is that they can’t say “the AI told me to do that.” You cannot just magically trust AI. This is why it’s very important to have an AI-first, as opposed to an AI-only, mindset. If you are doing magical thinking and assuming the AI will work without the human in the loop, you’re taking way too much brand and business risk.You could get into a world where your brand and your business might be gone because the AI completely misunderstood one moment in time and did something that a human could’ve obviously noticed and stopped if they were in the loop. 

With an AI-first approach, the AI will recommend what it suggests you do based on all the factors it can consider, and then the human can look at it and determine whether or not to proceed. This approach empowers the person. As Santiago Pino said, “AI will not replace you. A person using AI will.” That is really what is going to be happening, and that is why we must focus on how to empower the person using AI.

Recommended: AiThority Interview with Ruairi Adam, Director of Global Transformation at Majorel

Is there anything else you would like to add?

There is incredible power to AI, but what is really important to see is who can create and adjust it. A world where the few make the AI that affects the lives of the many is not a world I want to be in. That’s a very dystopian world. Think of the Gini Index and the wealth disparity. The AI knowledge disparity world would be far worse than the wealth disparity world. We must empower anyone and everyone to create their own AI and adjust it to their own preferences.

Thank you, Arijit! That was fun and we hope to see you back on AiThority.com soon.