Canadian organizations are increasingly looking for ways to gain value from their data, but many aren’t sure where to start.
“All roads lead to artificial intelligence (AI),” said Filip Draskovic, Information Architecture Domain Expert, IBM Canada. He was talking with IT leaders at a recent CanadianCIO virtual roundtable.
“It’s like driving a car and looking through the windshield, rather than in the rearview mirror. It helps to steer the company better to save money or do things more efficiently.”
The first step for any organization is a self-assessment to evaluate where it is today in its AI journey and where it wants to go, Draskovic said. This is the best way to pinpoint areas that need attention.
From there, Draskovic said that its essential to get the fundamentals right to gain value from AI. He outlined the top five steps to success:
- Clean up your data. Success in AI projects begins with a strong data architecture according to Draskovic. “You must have a strong foundation,” he said. That foundation comes from an information architecture. One participant noted that his organization invested heavily in its data infrastructure and as a result is capable of processing 200 million data points per day on one project.
But you do not have to sacrifice agility. Another participant mentioned that they had a “DevOps for data” approach which allowed them to generate results quickly. Draskovic agreed. “It is a mental mindset that you have to manage the workflow process of the data lifecycle.”
To ensure trust and transparency, it’s important to track the entire data lineage. “You need to have systems in place that trace the data and prove what was done to the data,” he said. “That is part of the foundation. If you don’t set the base, you don’t go to the next level.”
- Identify new tools and solutions that can help. The most time-consuming aspect of AI projects is collecting and integrating the data. To address that, it’s ideal to consolidate tools into one integrated architecture, said Draskovic. Otherwise, people can spend too much time on many different tools.
Data virtualization can help organizations move beyond the data collection stage. It provides an access layer to an inventory of data, wherever it resides, like a “one-stop shop,” Draskovic said. “It will help you move faster to achieve high AI value.”
- Get a quick win to build momentum. Once the foundation is in place, Draskovic stressed that organizations should seek a quick win. “Be selective on the low hanging fruit,” he said. “It has to be something cheap and quick.”
Some participants said that finding a “quick win” project can be a challenge. They asked for some suggestions. Predictive maintenance is becoming very common in many sectors, said Draskovic. As well, he sees opportunities in human resources and customer care. For example, organizations are using AI to analyze the issues that are contributing to the current issue of high employee turnover. Smart chatbots is another area that had advanced quickly to help relieve stress on call centres.
- Automate as much as possible. Having efficient processes in place counts for one-third of the success of any AI project, said Draskovic. He urged participants to look for every opportunity to automate systems for data processing and governance.
- Demonstrate your early success.
It can take time to build trust in the results of AI projects, said Draskovic. He encouraged the IT leaders to build on their initial wins with internal marketing campaigns. “Write up the metrics of success and get the business sponsor who benefited from the project to speak up for it,” he said. “We call it an internal roadshow. It generates envy among other departments so that they will want it too. It’s all about creating a buzz about the benefits.”