Building a Culture of Experimentation

Dataiku Company, Scaling AI Christina Hsiao

As more and more organizations continue to invest in their data and analytics practices, the question we repeatedly hear from analytics leaders is, “How can I streamline and scale my teams’ efforts in order to drive even more impact?”

In 2020, I had the honor of hosting more than 10 executive forums focused on Enterprise AI. During these events, dozens of leaders across many industry sectors assessed their organization’s current position on the AI maturity model:

Dataiku AI Maturity Model: A 5-Step Journey for AI Adoption in the Enterprise

Dataiku AI Maturity Model: A 5-Step Journey for AI Adoption in the Enterprise

The average score among leaders attending the forums was 2.6 — around halfway between Experiment and Establish, meaning that pockets of users have already derived value from AI through a few initial use cases, but that AI was not yet embedded across their organization as a whole. Interestingly, many attendees mentioned feeling that their company specifically was behind the curve when it came to widespread adoption of analytics...even though they were all at about the same place.

These findings are consistent with a 2019 Accenture report showing that while 84% of C-suite executives believe they must leverage AI to achieve their growth objectives, 76% also report that they struggle with how to scale AI. 

Although unfortunately there is no magic wand for digital transformation, during these forums I noticed that leaders who are successfully advancing their AI practices consistently expressed a few common themes. These themes included: a laser focus on value, establishing rigor around internal processes, and a culture of experimentation

It’s this last theme — the culture — that we’ll dive into here, as it’s a prerequisite to long term AI success, yet perhaps the trickiest of the three to link to tangible actions. As we know, mindset in individuals and collective corporate culture are difficult things to change overnight. 

Culture shift is a transformation, not a destination. Culture changes are never done...if you think there's a destination, you've missed the point."

- Director of Platforms & Engineering, Enterprise Data Office, Major U.S. Bank

Empower and Reward

One major bank leaned into its commitment to cultural transformation by overturning the traditional mindset that big decisions are made at the top. Rather than imposing top-down IT mandates for analytical tooling and processes, the company now intentionally empowers the people closest to the business problems to make vendor and solution decisions. For its data and analytics professionals, this bank also ties performance and compensation metrics to innovation, team collaboration, and the business outcomes his or her contributions deliver.

Compensation speaks, but there are other ways to encourage exploration and reward employees who trail-blaze; for example, hosting internal expos where teams can showcase their work to broader audiences or pitch their big ideas, Shark Tank-style. Being inspired by others’ innovations — and receiving public recognition and praise for their own — can be a strong motivator for many data professionals.

multiple arms reaching out to place post-its on a project table

Think Bigger

For organizations just beginning their Enterprise AI journey, the efficiencies gained from automation mean that streamlining processes for routine reporting is often the lowest-hanging fruit when it comes to project selection. But as teams become more experienced, it’s important to ensure innovation isn’t limited to simply improving existing processes. One leader at a major health insurance carrier promotes “proactive innovation,” challenging his data and analytics teams to create new, disruptive data products.

Think of the data lake as a forest. Historically, the data and analytics team would get a request. So we’d go to the forest, search for the right trees and chop them down, and prepare and deliver the wood. Back and forth, over and over. With the Data Exchange, the goal is to instead systematically harvest the entire forest, so that we have a warehouse full of standardized, prepared lumber ready to meet a diverse range of needs. 

And then go beyond that; to create finished products like hardwood flooring or bedroom sets — things that our customers didn’t even yet know that they needed! — and thereby create market demand for it.”

- Director, Data Strategy and Operations, Major Health Insurance Carrier

By extending AI’s benefits from merely automation efficiencies to include new value derived from pattern detection and predictive modeling, organizations can move from a climate of reactive, ad hoc reporting to one of proactive insights and recommendations. 

Redefine Success

A final piece of advice shared over and over amongst our forum speakers is to embrace curiosity and re-think the definition of failure. In a culture of experimentation, there are no failures, only learnings. 

This is not to say we shouldn’t try to positively influence outcomes by designing AI experiments with the appropriate rigor and a clear understanding of the value they might deliver. But when it comes to scaling AI, sometimes the most important question you can answer is “Is this project worth continuing to invest in?” Going back to practical ways you can enable this mindset, one executive advised, “Reward people who can shrink the time it takes to come to a conclusion.”

Short cycle times and regular group reviews are critical for AI experiments, so that the development team and stakeholders can quickly make go/no-go decisions and change course if needed. So failing fast — it’s not really failing. It’s about learning fast, and although what you learn might not be what you expected, sometimes it might be better! 

In the early stages of a tech lifecycle, it’s very important to keep an open mind. We aren’t afraid to try different approaches, and we don’t get married to one approach too early in the process. 

Success [in one of these early experimentation projects]... usually means that we find a solution that was not imagined at the beginning of the project; that we discovered something together as a team.”

- Global Digital Innovation Projects Manager, Leading Oilfield Services Provider

minds linked to different types of technologies

By engaging your data professionals to help make strategic technology decisions, you empower them. By rewarding them for innovation and effective collaboration, you invest in them. By redefining the boundaries of what’s considered success, you give them a safe space to be curious and encourage them to dream big. Implementing these practices will lay the foundation for an internal culture that is ready to tackle new challenges and invent better solutions, faster. 

You May Also Like

Democratizing Access to AI: SLB and Deloitte

Read More

Secure and Scalable Enterprise AI: TitanML & the Dataiku LLM Mesh

Read More

Solving the Ocean Plastic Pollution Problem With Data

Read More

Revolutionizing Renault: AI's Impact on Supply Chain Efficiency

Read More