Skip to main content

Q&A: Craig Stewart, SnapLogic – Academia and AI innovation

(Image credit: Image Credit: PHOTOCREO Michal Bednarek / Shutterstock)

What role do you think academia plays in AI innovation?

Academia plays an absolute pivotal role. Unlike in industry, where we’re constrained to some degree with return on investment, whether that be financial or time, when thinking about AI projects, academics are freer to pursue out-of-the box ideas and loftier applications. When academics are experimenting with AI there’s no real downside; if something doesn’t work, they’ve at least gained the knowledge of why it doesn’t work, and added that to the wider conversation. On the corporate side – with explicit and near-term expectations from investors, customers and other stakeholders – we often have to play within certain bounds and ensure we deliver the goals of the project like efficiencies, cost-savings, and so on.

This isn’t to say that corporate activity around AI is devoid of innovation, or the capacity to deliver innovation, but the ultimate goal is delivering something tangible, usable, with immediate value, so often there’s less scope for experimentation in that context.

Why did SnapLogic choose to hire a professor to create its own AI platform?

Well this really follows on from my previous answer. I think in order to really drive innovation in the field faster, there has to be synergy between the ambition and out-of-the-box thinking of academia and the can-do attitude, urgency, resources, and importantly, real world data held by industry. AI runs on data after all. 

That was the thinking behind getting Dr Greg Benson on-board in the early days of SnapLogic, and more recently for the development of Iris (opens in new tab), our company’s AI initiative. We wanted someone with outside AI experience to bring a different perspective to the project. It’s not really something that’s unique to the tech industry, and I’m sure you could ask people in many fields and they’d say the same, but when you’re working in a particular environment for long periods of time you can get comfortable with a certain mode of thinking, and that can be a hindrance. Greg was essential in preventing that from happening, and bringing that more adventurous and inquisitive academic spirit into the project. 

I don’t think I’m putting words in his mouth when I say I think Greg has similarly benefitted from straddling the gap between industry and academia. I think that the practical, goals-driven mentality of the corporate sphere has informed his way of thinking, and I know it’s certainly leaked into his teaching style as he has a better understanding of the skills needed to work in industry, and is passing that on to his pupils.

Not to trumpet it too loudly, but we’ve actually had his students come and work on projects at SnapLogic as part of their course work, and we’ve ended up hiring many of them as full-time employees after graduation, so I think it’s benefitting all involved.

Do you think the hype surrounding AI (both in industry and media) is justified?

In a word: yes. No question AI will transform our work and personal lives, in some predictable ways and in many ways largely unknown to us today.

I think it’s still quite early days for AI in business, and a lot of what’s being called AI is really just machine learning. We should take a deep breath when we’re thinking about the impact AI is going to have, for better or worse, as so much is just beginning to take shape. I think there is a huge amount of potential in AI technology but we’ve barely scratched the surface.

That said, it wasn’t long ago that only the largest companies, with deep pockets and large numbers of skilled resources, were investing in meaningful AI projects -- IBM Watson comes to mind as an example. But with recent advances in AI technologies, the availability of data processing in the cloud and on demand, and the associated cost reductions that have resulted, we’re now seeing companies of all sizes across all industries begin to experiment with AI and machine learning.

Bottom line, I’m an optimist, and my belief in tech’s ability to completely transform our way of living is one of the reasons I’m working in this industry. And I’m confident that we, as good, smart, well-intentioned humans, will push the technology to its limits but also reign it in if we sense it has the potential to do real harm.

How will automation drive digital transformation?

In my experience, one of the main reasons why businesses’ digital transformation projects underperform is that the time balance between implementing the shift and strategizing around it is off-kilter. For too many companies, more time is spent on the grunt work than actually first thinking about the strategic goals and outcomes.

Digital transformation isn’t an overnight process, and there has to be a certain level of trial and error, as well as adaptability when things go awry. Automation helps eliminate some of the heavy lifting, enabling more of a refocus for the IT department around how the strategy needs to change, or even what the eventual goals should be.

Time-to-value for big projects like digital transformation is critical. It’s important to get quick wins and build on those successes over time versus the years-long, big-bang projects common in the nineties. Applying AI and machine learning to automate rote, repetitive, manual tasks can be a tremendous accelerator on the road to value and ROI. It’s all part of taking agile development methodologies and applying them to business transformation projects.

What are the benefits of applying machine learning to enterprise integration?

Like with a lot of machine learning and automation systems, what we’ve discovered with Iris is that it’s really about speed and accuracy.

Just to give you an example: the first capability we delivered with Iris was the Integration Assistant, a recommendation engine that uses machine learning to deliver step-by-step guidance for building data pipelines — with accuracy of up to 90 per cent. 

Through the SnapLogic Enterprise Integration Cloud (opens in new tab), Iris learns from millions of metadata elements and billions of data flows, and then uses that learning to improve the speed and quality of integrations across data, applications, and business processes. 

The end result is that users are more efficient and productive. Complex app and data integrations end up being completed in a fraction of the time and cost, so IT and business teams can stay focused on delivering meaningful business outcomes, rather than getting bogged down in hand-coding.

And we’re just getting started with Iris and our AI and machine learning initiatives. We’re finding that all the development work that went into Iris holds potential value beyond our own core products. In discussions with our customers, we’ve found that they too can apply our machine learning capabilities, to their own business processes, in new and unexpected ways. We’re now working with select customers to determine how we can productize these capabilities and make them available to all of our customers.

What challenges do you see businesses facing when trying to implement AI systems?

One of the biggest issues is around a lack of talent to actually build and implement these systems. I think, as I alluded to earlier, students are coming out of their studies with a lot of AI knowledge, but not necessarily an understanding of the realities of working in the corporate world and translating that knowledge into business outcomes that align with strategy.

That’s not their fault, of course, and I believe working in conjunction with Greg and his students at the University of San Francisco is in some small way helping to remedy this issue. I’m certain other companies are doing similar things, but I still feel it should become more standard practice in tech as a whole. 

A lack of business strategy around AI is also contributing to this talent issue. I saw an EY poll of tech professionals recently, and it showed that 53% blame the AI skills gap on a lack of AI insight in business practices, and 48% blame a lack of managerial understanding of the technology. From some of my personal experiences I can see a reasonable argument here.

Essentially, established developers looking to upskill need to be given more direction from businesses around what AI skills are going to be most valuable for them, to incentivise them to actually make that leap. Otherwise, why take the risk when more traditional development skills are still in demand and sought after?

Lastly, lack of skilled talent is one thing, lack of imagination is another. Being able to spot business opportunities, or threats, and then determine where and how to apply AI to achieve a desired outcome is essential. For many companies, who maybe have had the same leadership in place for a number of years, it can be hard when you’ve been doing the same things for years to suddenly see things differently. I can sympathize. We try to use our outside-in perspective and help our customers find those opportunities and bring new technologies such as AI and machine learning to help them make that step change. 

Craig Stewart, VP Product Management at SnapLogic (opens in new tab)   

Image Credit: Michal Bednarek / Shutterstock

Craig Stewart is VP Product Management SnapLogic.