Believe the hype. AI has become one of the hottest topics in the business world for good reason: 47% of organizations have embedded at least one AI capability into their business processes, according to a 2018 McKinsey & Co. report.
In fact, AI adoption in organizations has tripled in the past year, and AI is a top priority for chief information officers, according to Gartner.
“Your businesses can’t survive and thrive unless you adopt artificial intelligence,” says Dinesh Nirmal, VP of development in IBM’s new data and AI business unit. Nirmal oversees more than a dozen global AI labs and design studios.
Why is AI so vital? Because it can deliver time and cost savings by automating manual tasks, identifying risks and enabling leaders to make data-driven decisions that can fuel business success.
AI has already become an integral part of consumer life. We use voice assistants to set timers and turn on air conditioners. We receive personalized ads based on our internet searches. However, despite rising adoption of AI overall, companies outside of tech, telecom and financial services have been slower to embrace the technology.
“Everyone talks about enterprise AI, but it’s still in the early stages,” Nirmal says. The sluggish pace of adoption can be blamed, in part, on limited access to the right talent, deployment challenges and the very real risk of getting it wrong.
“AI in the enterprise can be much more complex,” Nirmal says. “But it can also add a lot of value.”
AI in Action: Real-World Results
IBM’s supply chain team, which manages roughly $40 billion in materials and spend per year, recognized that the organization could improve efficiency, responsiveness and oversight by using AI. Working with IBM Watson, the team launched an initiative that uses cognitive technology, predictive analytics and automated reports to improve end-to-end supply chain visibility and speed up information sharing and decision-making.
The results: IBM’s supply chain division reduced its time to gather and analyze data by more than 75%. This allows the team to respond to critical supply chain disruptions, such as natural disasters, in hours instead of weeks—and cut costs in half in the process.
It’s not a one-and-done process, either. As the algorithms learn from each decision and action, the system improves its ability to quickly identify trends and respond to issues.
In another example, IBM worked with the Centre Hospitalier Chrétien (CHC) in Belgium. The hospital can support 1,700 patients at a time, each of whom requires detailed tracking of their care and follow-up. A team of 15 people was needed to review every discharge letter, medical record and lab result to assign the proper codes from a list of 69,000 options. Needless to say, it was time-consuming manual work.
So CHC automated its coding process with AI. Like all AI projects, it started with a sizable amount of “clean” data. This meant fixing errors, removing duplicates and clearing up irregularities. It also required filling in missing data and adopting a consistent language and format across all entries.
“Having good data is the most critical piece of AI,” Nirmal says. But it’s also where most AI projects fall short: Just 3% of companies have acceptable data quality standards, according to a University College Cork study.
When the data is clean, the next step is less daunting. “Building an AI algorithm is fairly straightforward,” Nirmal says.
Over the course of 18 months, CHC used a large set of pre-coded medical records to teach the AI bot to understand the terminology. Within a year, the bot achieved an 80% accuracy rate.
Now, the bot suggests codes, which are then reviewed for quality assurance. As a result, productivity at the hospital is up by nearly a third.
The AI Ladder
Every organization follows a different path to adopting AI, depending on its size, industry, goals and technical skills. To figure out what that journey will look like and how long it will take, IBM describes it as “the AI Ladder.”
“Every organization enters into AI on a specific rung and scales at their own rate,” Nirmal says. “Providing the ladder analogy helps to create a realistic AI implementation plan, and accelerates the planning and development of AI architectures.”
Rung one: Collect. Capture a large enough data set to derive meaningful insights. Then make sure the data is clean.
Rung two: Organize. Set parameters for how the data will be organized and analyzed to deliver results against a business goal.
Rung three: Analyze. Create dashboards and other tools that will display the results for easy interpretation. (In other words, is it working?)
Rung four: Infuse AI. Deploy the machine learning algorithms, implement automation tools and start analyzing the data.
Building an AI model in isolation is a good first step because it doesn’t need to conform to any of the rules or legacy applications that make up most enterprise systems. “But once you try to infuse your model into an existing system, all those complexities become apparent,” Nirmal says.
Even a simple AI application may need to access data from multiple platforms. Keep in mind that those platforms might have been built by different teams, each with its own security rules and structures. AI also may have to accommodate different levels of data privacy, service level agreements and key performance indicators—all while not interfering with the speed and efficiency of existing processes.
The key to setting expectations: “Think of it as an evolution rather than a destination,” Nirmal says. AI learns and improves with each bit of data it consumes. Likewise, a company that implements AI needs to be willing and able to adapt. But the effort is worth it for improved customer relationships, increased efficiency, fewer mistakes and lower costs.
“It’s a progressive journey that will continue to make your business, and the lives of your customers, better,” Nirmal says.
3 Reasons to Adopt AI
“It’s easy to feel overwhelmed by the idea of deploying AI in your organization, but the end results make the effort worthwhile,” says IBM’s Dinesh Nirmal. Here’s why.
AI has the potential to speed up processes across organizations, which can lead to measurable savings. Some examples:
AI tools can review thousands of pages of constantly changing regulations to determine how and when any new rules affect a particular business based on its industry.
Traditional manufacturing lines could be replaced with smart machines that alert staff when they require maintenance or repairs and even order their own parts.
Large companies have millions of pieces of customer data, including what they’ve bought in the past, when they make purchases, and what deals or incentives they’re most drawn to. AI tools can mine that data to discern preferences and provide insights into what customers will most likely want next and when.
3. Enterprise knowledge
This is the most valuable application of AI, Nirmal says. “When seasoned employees leave, they take their knowledge with them, and it’s gone,” he warns. Capturing that knowledge is important, but AI goes one step further: It can offer suggestions to improve processes.
Brave New World
91% of all enterprises expect AI to deliver new business growth by 2023.
But only 5% have extensively incorporated AI into current processes and offerings.
Secret to success: Find the right innovative thinkers in your organization and give them the power and resources to test. Identify a specific use case for AI and deploy a small pilot project to prove it delivers business results.
Source: MIT Sloan Management Review
Photo credits: PeopleImages/Getty Images