How to go beyond hype to succeed with AI?

How to go beyond hype to succeed with AI?

Canadian CIOs are still trying to find their way through the hype to get real value out of artificial intelligence (AI) and machine learning.

"There is a lack of understanding," said a CIO of the financial sector at a recent CanadianCIO virtual roundtable. "We need a more abstract look at what AI can do for us and how it ties into our business objectives."

Many organizations aren't sure where to start with AI, acknowledged Philip Draskovic, information architecture executive, IBM Canada. “It should start with a self-assessment to evaluate where they are and then plan small steps forward to move up the analytics and AI maturity curve.” Draskovic suggests that businesses look for "low-hanging fruit improvements" that will produce quick results.

There was a general consensus among participants that, at the end of the day, success with analytics and AI would depend on the data.

You can't have AI without it

There is no quick fix for improving data quality, Draskovic said. "It's not a one-time thing. It's an ongoing process and two-thirds of it is people and processes."

Data quality is important to be successful with AI, but improving information architecture is equally important. "We have a saying that without IA there is no AI," Draskovic said. Information silos have to be dismantled. Organizations must ensure that users can access and share data securely from a single location.

If employees have to copy information to share it or if ongoing improvements to the data aren't tracked, you're losing time and productivity, Draskovic said. Making copies also poses a security risk and increases storage costs.

This approach is not the same as moving the data to a data warehouse. Instead, data virtualization provides an access layer to an inventory of data, no matter where it resides. "It's a one-stop-shop for any data in the organization," Draskovic said. This solves the biggest problem when starting a new project, which is knowing where to find the data. It also simplifies data governance.

Another advantage is that centralized access can put data in the hands of line-of-business experts who can become "citizen data scientists," said a public sector CIO. They are in a good position to identify use cases that support the business objectives.

Are we ready to take the machines into our own hands?

Many participants admitted that their organizations were not prepared to turn decision-making into machines. "There is resistance from business leaders because of concerns over privacy and access to data," said one IT leader. "Proceeding too quickly on this front leads to loud banging."

Organizations should also monitor machine learning to ensure that the models they are using are not biased. It will soon be required by law, Draskovic said. An IT leader said she sees a risk in allowing machines to make sensitive decisions that impact people's lives.

"The bottom line is whether you trust the data and the governance," Draskovic said. "And it all comes down to the strength of the architecture."