How to deliver a scalable AI pilot in just 8 weeks
In business, aspiring to world-class is not enough when your competitors are already there. About half of the companies listed on the S&P 500 will be replaced over the next 10 years. Compared to the past, what’s unique about the disruption happening today is the rapid pace of change. During past revolutions, businesses had many years – even decades – to adapt. Today, that is no longer the case.
One indicator of just how fast the landscape is changing is the longevity of companies on the S&P 500, an index of leading U.S. companies. Roughly a hundred years ago, the average lifespan of a company listed on the S&P 500 was 67 years. Today, that the average lifespan of a company on the S&P 500 has decreased to just 15 years. This demonstrates that it is becoming harder for companies to stay in the lead, or even in business, for very long.
Experts believe that we’ll see this trend continue. Richard Foster, a Yale University professor, predicted that in the next decade, just 25 percent of the companies currently listed on the S&P 500 will remain listed. New companies will make up the other 75 percent, and it’s likely they’ll be the trailblazers using innovative technologies like artificial intelligence and hybrid cloud. The companies that will remain on the index will be ones that choose to embrace AI and rapidly evolve their businesses.
The bottom line: today’s companies must adapt quickly to change, using new technologies that fuel competitive advantage, or risk getting left behind. That’s why it’s imperative to make the most of data, cloud and AI—all of which help companies accelerate the speed of their business through smarter decision-making and faster execution.
AI at the Speed of the startup at the scale of an enterprise
By 2020, 95 percent of the top 100 largest enterprise software companies will have integrated cognitive technologies into their products, according to Deloitte. It can be deceptively easy to start an AI pilot, deliver incremental results, show some solid business outcomes. However, it is fiendishly hard to move toward “AI @ scale” and enable all horizontals, like marketing, HR, sales, customer support, customer experience, risk management and fraud management. All sorts of problems arise, threatening to undercut the AI revolution at its inception.
8-week AI Pilot at one large ANZ Bank
We were recently in the following scenario with a client, one of the biggest banks in ANZ region. The bank had set up data science team to start a few pilots and deliver on some of prioritized use cases. The big question they faced: how do they scale these experiments across their company, divisions and business subsidiaries? Let me go one level deeper into the AI pilot they carried out and what we learned together while helping them to scale AI.
The bank’s newly-formed data science team delivered an AI model which reduces the risk of non-performing loans (NPL) by 10 to 15 percent. This metric is one of the KPIs of their risk officer. The data science team faces several issues after the AI model is deployed in production. How do they keep track of rapidly changing data sources; how do they automate continuous learning of AI models; and how do they address the possible bias problems in the model once they scale the application?
Scaling the AI Pilot to AI @ Scale
Along with this ANZ client, the IBM team started out by drawing a roadmap to do AI @ Scale powered by IBM Cloud Pak for Data—a leading enterprise insights platform that allows you to simplify and automate how your organization turns data into insights. We built an entire strategy encompassing three core pillars—culture, architecture and technology. It began with the bank’s risk management division delivering an AI-powered dashboard that shows estimated risk of each loan in the region. Users can analyze data by type of client, loan and branch office.
The same set of data scientists were also able to build AI-powered apps for customer acquisition and customer experience leaders under marketing by leveraging intelligent, collaborative services of the Cloud Pak for Data platform. The outcome: the bank’s customers can get a new, personal microcredit through the customer virtual agent after the bot has run the risk model. In less than five minutes, customers can access the money in their accounts due to the use of a predictive risk model with a bot. In this way, AI is steadily infused through the bank’s processes – both back-end and front-end – to help make the bank a true digital player.
How did we do this?
Again , three core focus areas helped IBM and our client deliver AI @ Scale: culture, architecture and technology. Culture can promote experimentation, underpinned by trust and respect for data privacy. The architecture was based on hybrid cloud, open standards and built to leverage the cutting-edge technologies like the latest AI frameworks and the latest innovations in hardware acceleration.
I am going to cover these three critical focus areas in more detail in my next blog post. Meanwhile, you can learn more about our end to end AI platform, Cloud Pak for Data. And you can enroll in out Data and AI forum to get the latest and greatest updates about the world of AI.