Q: What does it mean, from both a technology and a cultural standpoint, to be ready for AI? What are the enablers?

Iain Brown: From a technology perspective, much of it comes down to maturity around data and capabilities and resources that exist internally. Many organizations want to do more with what they’ve got, but their ecosystems may not be ready to deal with the volume of data or the processing power needed to understand the data in the best way possible.

The technology enablers include the right data feeds coming in, having a consolidated view across the businesses, and having the right technological capabilities. There’s both a software and a hardware aspect to this: making sure you have the right hardware, localized or cloud-based, that can process the data that you need in a timely manner, and then the actual software, the analytical algorithms that you can utilize.

Culture is a harder piece to deal with. There’s still a lot of pushback within organizations around AI adoption. That may have to do with a misunderstanding of what AI is and what it can achieve, and, in some regards, how much AI will take over the human aspect of the role. There needs to be buy-in from the business, across the business, for successful AI adoption.

Read more from this series with MIT/SMR Connections

There also needs to be a clearly defined reason for AI’s use and a well-governed application of it, as well as a positive augmentation of existing human efforts. When organizations try to go to wholesale automation of a process, they typically struggle to, first, get buy-in internally, and second, deploy something that’s robust, fair, governed, and interpretable in certain regards — and that has the right kind of human oversight.

Q: What are you seeing in terms of SAS customers and AI adoption?

Brown: Typically, organizations have focused on structured data, so they use data that already exists in well-defined databases. It’s easy to manipulate. It’s numeric. It’s in a constant form. Where we’ve seen a big uptake in the last half-decade is in unstructured data usage. This could be as simple as textual documentation — think webchats, call center records — or more complex in the form of image, video, or audio.

The classic stat is that around 70% or 80% of all data is unstructured. But I’d say most organizations still aren’t getting the value out of that level, so they’re exploring how to generate value from it. This usually means starting small with aspirations to grow.

Culture is a harder piece to deal with. There’s still a lot of pushback within organizations around AI adoption.

We’ll typically go in and have a look from a value perspective for where it’s easy to start a project. That means the right level of data already exists; someone’s got a good understanding of it; there’s a business problem that’s clearly defined.

That’s often where you can grow utilization as well. If you can prove the value and show the business “this is how AI should be applied, and here’s a use case in your organization of how it’s working well,” you’ll see much more momentum behind the adoption of AI across other departments and applications.

Q: How is IT’s role evolving in response to AI?

Brown: In terms of operationalizing, a lot falls back on IT. You have data scientists who are building great innovative things. But unless they can be deployed in the ecosystem or the infrastructure that exists — and typically that involves IT — there’s no point in doing it.

IT’s role has been elevated in terms of decision-making, and that’s drawing IT and business closer together. So there needs to be a greater collaboration between the two. But I’d go further than that. I think the data science community and AI teams should be working very closely with IT and the business, being the conduit to join the two so there’s a clear idea and definition of the problem that’s being faced, a clear route to production. Without that, you’re going to have disjointed processes and issues with value generation.

Q: In terms of AI-related talent, what’s mission-critical and what’s “nice to have”?

Brown: Diversity of thought is extremely important. There’s a lot of debate about how data scientists should grow as a community. At the moment, unfortunately, there’s not a lot of diversity, and that causes all sorts of potential issues.

In terms of skills that are most applicable — this goes back to the STEM-type areas: science, technology, engineering, and mathematics. But I would say STEAM, including the arts, is also important. Having creative thinking alongside the very statistical and technical knowledge can differentiate organizations in terms of the way they utilize AI.

You also need strong capabilities and skills from a delivery and IT perspective. You need to have DevOps- and Model- Ops-type personas sitting in there as well.

Q: Same question in terms of applications, technologies, tools: What’s mission-critical, and what’s “nice to have”?

Brown: You need to have a platform-type approach. Having a centralized set of capabilities that could be used across multiple projects and departments — that’s where things really work well.

I’ve seen where you have different silos popping up within an organization, building their own wheels, as it were. That not only doesn’t make financial sense for the organization, it also doesn’t allow for efficiency or scalability gains when you start to tackle really difficult problems that touch multiple departments.

Diversity of thought is extremely important. There’s a lot of debate about how data scientists should grow as a community. At the moment, unfortunately, there’s not a lot of diversity, and that causes all sorts of potential issues.

Having a platform of capabilities, whatever they may be, is essential to making sure that an organization generates value as quickly as possible and accesses the underlying data to generate that easily as well.

I do think a cloud-based environment is the way forward. You can’t rely on localized versions of anything; that approach doesn’t scale. Relying on third-party cloud providers — the Microsofts, Googles, and the AWSs of the world — is a good way to lower the barriers to entry. But, again, you need the right capabilities sitting alongside those environments to make the best use of them and to get the value out of the data.

Interested in more about AI? Learn more about technical readiness for AI in the full report

Iain Brown is the Head of Data Science for SAS UK&I and Adjunct Professor of Marketing Analytics at the University of Southampton. Over the past decade, he has worked across a variety of sectors, providing thought leadership on the topics of Risk, AI and Machine Learning.