While we may be at the beginning of the predictive AI revolution, IBM, which is at the forefront of the movement, has already seen its fair share of challenges. The US tech giant has developed its processes and engagement capabilities – both internally and externally – to such a degree that users of its platforms and programs are able to analyze data for almost any business or organizational requirement.
"Data is the fuel for digital transformation and those companies that are going to be most successful are those that can take that data and use it predictively to shape future events, and enable people to conduct higher-value work and transform business processes," says Scott Hebner, vice-president of marketing at IBM Data and AI.
While IBM may not be the only big player out there enhancing its analytical abilities, it has been able to differentiate its services by understanding what its clients are hoping to achieve with their data and why. And it is a data-driven cultural transformation agenda that will likely pierce most sectors of society and business.
"In marketing, the world in which I live in, we have become a highly data-driven culture where we know everything about all the touchpoints that can help us optimize a client’s experience with us," Hebner explains. "Many of the different professions we work with are data-driven, which helps when we talk to customers, as we can share our own experiences to optimize outcomes."
More than the basics
Hebner oversees the marketing function of a recently integrated part of IBM's business that brought together the tech giant's data and analytics group with its IBM Watson unit into one integrated organization. But this new business unit is about more than the huge potential of IBM's data and AI platforms, products and solutions.
As Hebner points out: "It's one thing to provide customers with technologies, tools and platforms, but it's another thing to help them make the cultural transformation into a data-driven digital world. And it's a lot more than just capabilities – it's about the skills and the leadership capabilities of data teams – and, in many ways, it's cultural."
To encourage clients to embrace a cultural transformation focused on how they use their data, there is never a better place to start than with your own systems and infrastructure. And, in order to foster acceptance and willingness, Hebner asserts that it is essential to focus on facts that are backed up by evidence.
"The one thing I've noticed about IBM over the past decade is that there has been a cultural shift," Hebner notes. "In our world, facts and evidence rule. We are required to have evidence-based opinions in data-driven arguments when we want to accomplish something, whether that is attaining additional funding or investing in new offerings.
"We have evolved to a point where data is critically important, because while opinions are a dime-a-dozen, when you have the facts behind you and you're making evidence-based decisions, it makes all the difference in the world."
IBM is, according to Hebner, committed to using data in the appropriate ways – an especially fraught topic considering recent, much publicized debates and discussions on data privacy and AI ethics and democratization .
"As part of our data-driven cultural transformation, we have to value the data and understand its context," he says. "We're only able to use the data that the compliance and company policy will allow us to use, and that often refers back to what the user or the client wants us to have access to."
Moving from opinions to facts
The move from making data-driven decisions based on opinion to using evidence-based data has been supported over the past few years by the shift from companies within the AI and data space toward a predictive, rather than reactive, approach to data.
"We're now able to get a better sense of what may happen in different scenarios," Hebner remarks. "But this notion of predicting what's going to happen in the future based on trends brings us into the world of AI. And, of course, this brings with it yet more compliance and policy issues, such as making sure there is no bias in the AI models to ensure there is explainability and lineage, so that a decision based on a recommendation can be explained."
Climbing the ladder
Hebner describes IBM's methodology as the "AI Ladder", which both it and its clients must climb together. The first step, the collection of data, is about making that data simple and accessible regardless of where it lives or what type it is. But with some 80% of an enterprise's data either untrusted, unanalyzed or inaccessible, Hebner advises IBM's clients to start with ensuring they are unlocking the value of all that data they have, both internal and external; especially so in today's world where data is exploding everywhere.
"Just because you can collect data, doesn't mean that it's useful," he notes. "This is where compliance controls and policies come in, and where catalogues – making data business-ready – can help users understand what the data actually is, trust it and have self-service access to it."
Once data has been integrated and cleansed, and can be trusted, the next step of the AI Ladder is to begin on the analytics.
"You can use traditional analytics such as dashboarding, or you can use AI-based models which many companies are already experimenting with," explains Hebner. "And, finally, you have to operationalize all of this, because ultimately the analytical or AI models have to get to the people who are doing all of the work. From there, it becomes part of a business process or a workflow, and joins a continuous loop, as these models are generating even more data.
"The overriding challenge for most companies is that their data is everywhere, and from an analytics' perspective, the data has to be federated and democratized for an AI and multicloud world," he adds.
Keep it open
Hebner argues that the technology used to process data should be open-source based, as it offers more choice and flexibility to users.
"Users might have different preferences for different tools, but you want it all based on a common, open, multicloud platform," he says. "In a nutshell, that's the AI Ladder; that's what we've been doing at IBM and what we've been working with customers on.
"There's no AI without an IA", says Hebner, denoting that without the implementation of a solid and dependable information architecture, there can be no AI.
"What is really important here is that you have the ability to virtualize and federate all of your data sources, and then make it available for different types of business units and users with different skillsets," he notes. "And, in our world especially, you have to globalize it because we have to operate it in 170 counties with different languages and data regulatory policies."
To overcome the challenges here, Hebner says that data scientists should build infrastructure bottom up, so whatever platform or tools are being used to access the data – whether it is customer service, customer support, marketing or sales – there is one information architecture and that data is from one version of truth.
"In the old days, different versions of data would often say different things. We've solved that issue and that's the notion that there is no AI without an IA," Hebner concludes.