The hardest part of any technology disruption isn’t ideation or experimentation. It’s moving from experimentation to real business value. With the massive hype around generative AI, there is more pressure from boards to implement AI thoughtfully.

At Summit in June, I was joined by Sasha Jory of Hastings Direct and Awinash Sinha of Zoom for a CIO Executive Panel focused on optimizing AI tech investments. Over the last year and a half, there’s been massive pressure for CTOs and CIOs to harness the potential of AI while containing costs and mitigating risk. 

“Gen AI is a new wave of technology — but we’ve seen many waves of technology,” Sinha said. “The key part is connecting technology to the best possible contextualized outcome for your business.” 

How do you take the buzz around generative AI and apply it to your world in a pragmatic way? 

“From a Hastings perspective, we’re not caught up in the hype around using AI to solve every problem in our organization,” Jory said. “We tend to focus on making sure data is in the right place, managed in the right way, and we have the foundation in place to explore that data with any tech we want to use.”

Snowflake approaches gen AI in the same way. We’ve been disciplined about data from the beginning. When we consider the building blocks that contribute to a solid AI strategy, it all starts and ends with your data.

How Snowflake, Hastings and Zoom use gen AI 

My team is customer zero for everything we take to market here at Snowflake. The best part of that is the ability to build custom applications that improve productivity throughout the organization:  

• Helping employees settle in faster: We aim to revolutionize the employee experience, from Day 1. Getting up to speed can take up to 90 days, which is frustrating for both a new hire and the overall organization. One obstacle is that we document and communicate many of our processes through knowledge-base articles designed for help desk use, which is cumbersome and not as intuitive or proactive a way to empower a new team member. My team addressed this by consolidating these articles and using the AI/ML capabilities of Snowflake Cortex to create the Employee AI Assistant. This Streamlit app lets employees easily access information, like “how do I configure a printer in the San Mateo office when visiting from London,” without needing to open a support ticket. The assistant, provided to new hires on their first day, also incorporates compliance, HR training and certification tailored to the individual's role. 

• Shifting away from QA drudgery: How many data engineers live and breathe quality assurance testing? I don’t think many do. Resources are always tight, and engineering talent is always in high demand and short supply, so we’re shifting the QA and testing part of software development to AI. 

Our internal app, QA Assistant, streamlines the software development lifecycle. Remarkably, we have implemented this capability in three major SaaS applications, reducing the workload by more than 35%. We’re applying this framework to more complex applications, confident in its repeatability. Now, software engineers (who typically dislike QA and testing) can focus more on design and development while exploring the potential of AI. 

Hastings Direct is one of the leading general service insurance providers in the UK. Insurance is all about risk — determining the risk factor for a driver, a homeowner or anything else. Their initial, successful AI projects include:

• Improving risk assessment: About 10 years ago, Hastings’ underwriters were limited to the standard 40 to 50 data points when evaluating insurance applications, making it difficult to stand out in the market. Now they use thousands of data points, sourced both internally and externally. In the UK in particular, potential customers use many price comparison websites that query Hastings to deliver a competitive quote, and speed is absolutely critical. Now, with the help of Snowflake, Hastings can assess a customer through those thousands of data points and deliver an estimate in less than 3 seconds. 

Zoom is a subscription-based collaboration platform that saw phenomenal growth over the last two-and-a-half years, including product line growth and geographical expansion. That's good news, but it required considerable, rapid upleveling of its business operations, Sinha explained. Now, with a cloud-based warehouse powering its AI/ML engine, the company is evaluating the customer lifecycle: 

• Sourcing cross-sell opportunities: For the past two years, Zoom has used ML to inform cross-sell opportunities, flag when a customer might be most receptive to expanding their Zoom licensing, and anticipate customer churn.

• Presenting insights in the local API: Traditionally, analytics are delivered through some sort of dashboard. Now, Zoom is bubbling up gen AI-powered insights directly in an API where sales reps work day in and day out. “We can access Salesforce, the Zoom client and mobile insights on Zoom Chat for our Customer Success team to get them information in less than 60 seconds,” Sinha said.

‘A copilot, not an autopilot’

Gen AI will impact jobs, but not in the fear-inducing way people were thinking. I don’t envision the technology replacing people, and neither did my fellow panelists. “I keep saying it to my team: This is a copilot; it’s not an autopilot,” Jory said. “You can use this technology to enhance productivity, but you need to review what’s coming back, keep training your models, and keep ensuring it’s giving the right answer.”

There’s so much you can go for with generative AI. From the vantage point of the CIO or CTO, the opportunities are almost overwhelming. You have to really contemplate which use cases provide the greatest business impact because everyone is clamoring to get their hands on it. Don’t lose sight of those value props. If your data is in the right place and you have the right partners, the sky is the limit. 

For the full conversation, watch the CIO Executive Panel: Optimizing AI Tech Investments in an Era of Global Uncertainty.