The Implications of Underutilizing AI-as-a-Service
It’s no secret that artificial intelligence (AI) is radically transforming how we do business. Although, the pace at which businesses are implementing AI in their day-to-day operations is far behind the pace at which the capabilities of AI technology are evolving. Businesses that don’t leverage AI’s potential will constantly be playing catch up with competitors, but what is causing so many businesses to delay embracing this disruptive new force?
The short answer: AI-as-a-Service (AIaaS) has a marketing problem. Many enterprises underutilize the technology due to a limited understanding and limited imagination of what it can do for businesses, how to deploy and incorporate it with existing processes, and concerns about the potential for harm if an AI-based solution “gets it wrong.” Many business leaders have not considered the short- and long-term implications of not embracing AIaaS—and that’s a problem. In data science, this condition is called counterfactual.
For example, consider the following statements; using today’s tools and processes, how do businesses measure each? How could AI improve the accuracy and timeliness of these metrics?
• Operations: Where are the bottlenecks in your critical business processes? Which teams in your organization do you see the most improvements needed?
• Supply Chain: Where is the greatest risk in your B2B, B2C or B2B2C supply chain? How much would specific actions improve that risk posture in areas such as time-to-delivery or out-of-stock?
• Churn/Customer Retention: What are the key factors driving long-term customer retention in your organization? How does this vary by product/geography? How does this vary by individual account?
• Data Governance/Democratization: How do your current data systems support equitable data access across your organization? What percentage of your organization do you consider “data-enabled” through data-driven decision making, self-service analytics and insights, and accessible AI solutions?
In these examples, successful AIaaS projects potentially enable increased revenues, reduced costs, improved margins and more rapid growth. In short, not embracing AIaaS equates to leaving money on the table.
At Kin + Carta, we use our proprietary, cloud-based, automated machine learning platform called Octain to build predictive, explainable models for clients that have accelerated data-driven decision making and digital transformation outcomes.
What does AIaaS look like in the real world? Some examples from our clients include:
• A multinational technology giant was able to make better predictions for future demand to get ahead of supply chain shortages and blockages, resulting in tighter manufacturing targets and better finance forecasts for this publicly traded Fortune 100 company.
• An international petroleum transporter and distributor was able to better predict wait times at 83 nationwide terminals resulting in a new, modernized workflow that automates the last mile of its supply chain.
• One of the largest food, snack, and beverage companies in the world was able to better predict units sold at the SKU level globally by retailers, reducing error by 50 percent in unit forecasting.
• A consumer electronics company utilized AIaaS to develop a content acquisition strategy and to separate content that earns revenue from other content that is rarely viewed.
One of the largest U.S. retail beverage companies used AIaaS to detect fraud and score propensity-to-pay—the key here, though, is ensuring these predictions are made without bias. One can’t discuss the use of AI without addressing bias. I believe that all machine learning algorithms are inherently biased since they’re trained on historical data. However, it’s up to us (as humans) to fix this issue. At Kin + Carta, we place equal value when it comes to people and profit and when building Octain, we wanted to ensure that these biases were cleansed so that the models produced were not only accurate, but ethical.
By selectively filtering on the data sources with known biases (like home zip code and area code) as well as other unknown biases (demographics such as age and gender) the model is able to focus on causal factors versus those that are residual and may represent systemic bias.
Utilizing ethical tools should be the standard for businesses to set up the foundation for a better future. At the head of innovation are people and technology and our mission at Kin + Carta is to build a world that works better for everyone through sustainable digital transformation.
The healthcare industry is a great example of where AIaaS can be particularly powerful when deployed ethically. We have used Octain to help clients identify interventions specific to certain patient segments to improve outcomes and continue to learn over time. In addition to acute care, we’ve developed methods to parse electronic health records (EHRs) using natural language processing, developed novel data products from health wearables and worked with the largest U.S. healthcare claims processors to identify opportunities to improve efficiency and profitability.
During a time where the world is constantly changing and different needs are being created and met, businesses need to act fast. One of the great things about AIaaS is its ability to cut manual processes that take up to three months, for example, and shorten it to a matter of minutes. It can automate repetitive tasks faced by data scientists such as data cleansing, model selection, visualization, and resource optimization, and can be deployed quickly and continuously.
AI-as-a-Service gives enterprises the ability to fill in the cracks as markets change, and since it’s AI, it’s always learning so it provides better, optimized outcomes.
About Cameron Turner
Cameron Turner is the Vice President of Data Science at Kin + Carta. Cameron is a seasoned technologist who has over twenty five years of experience delivering results through strategy, team development and full implementation of digital transformation projects. Most recently, he served as the CEO of Datorium, a machine learning and data science product company, while also leading multiple teams to deliver data products to Fortune 500 clients.