The costs of random acts of AI: The right culture is the new strategy imperative

In my experience leading transformation initiatives, I have seen many similarities between broader business transformations and AI transformations, but there is one key difference – in AI, the world is still highly experimental. In fact, 3 out of 4 of AI projects fail to show positive returns on investment. Because of this, there hasn’t been a map of what AI success at scale really looks like. It remains uncharted territory for enterprises.

In October 2019, we commissioned a study with Inc.digital, where we talked to >550 executives across a variety of industries to understand why enterprises don’t more frequently see tangible, measurable results from AI.

Through this research, we were able to determine the DNA of the few that are successful with their AI strategy – the one common thread is keeping the AI strategy, data, technology, people and processes close to the core and controlled. These organizations dominate the highest ROI across financial metrics including 208% change in revenue; 245% better performance in product and service design; and 500% less likely to fail. 

If keeping AI controlled and close to the core is the secret to ROI, then organizational structure and culture need to better support these future-forward, strategic initiatives. Without foundational organizational building blocks in place, leaders ended up working overtime to clean up the mess of well-intentioned first movers and visionaries. Most rogue AI experiments and programs lead to mediocrity – excess spend, data put at risk, potential bias, and an inability to scale across the organization. And these random acts of AI cannot be measured for impact across the organization. 

Starting at the top avoids these stumbling blocks. Setting an executive mandate and making the transformation a business priority sets the tone across the teams. On top of that, I’ve found that the basic foundational elements of transformational success include a clearly articulated objective, with a framework and collaboration model that drives cross-organization focus and collaboration.  

But in many organizations, there is no executive mandate for the AI strategy. It’s often run via skunkworks projects that start within the data science team, with no broader business collaboration or oversight. For example, once IT gets wind of the applications, frameworks and hardware that sits on the desktop, or under the desk of, the data scientists, they often go full ‘IT police’ to crack down on the spend – remember the early days of the cloud and IT’s worry about shadow IT and non-approved purchases of applications like Tableau or Slack? Well, in the AI era, these rogue technology programs pose even more concern – not just from a manageability and cost management perspective, but also now from a data privacy and security standpoint. When IT swoops in and begins confronting the data scientists and working to standardize the AI stack, they end up starting an internal war instead of fostering a collaboration across the organization. This type of cross-team conflict becomes counter-productive and takes the focus off the broader transformation goals. 

AI transformation could benefit from a more systematic, standardized approach and a center of excellence model with the following six key elements. 

1. Business orientation and focus that can positively impact more broadly than just an isolated one-off experiment
2. Cross-team collaboration model that ensure implementable, measurable and scalable outcomes
3. Coherent information architecture approach that enables data access, governance and privacy
4. On premise applications and infrastructure plus skilled in-house resources that sets smart defaults, ensures control and manages costs
5. Bias oversight committee that addresses proactively addresses biases (and potential biases) in datasets and in training
6. Documented processes and rules of engagement that delivers efficiency through scale and repetition

Embracing a center of excellence model requires a cultural shift across the organization; new ways of working together; and a mindset of exploration, openness and curiosity. It’s about balancing the need for structure and defaults, without squashing the data scientists’ out-of-the box thinking. It’s about building an agile organization, but not necessarily an Agile organization. The CIO of Red Bull Racing recently shared with me that the key to AI success in his organization is having “agility embedded into the DNA of the organization, as well as a singular focus across the collaborating teams."

This is more work up front for the leadership team, but significantly less policing, less internal strife and better results on the back end. And, at the end of the day, it leads to more measurable ROI from AI investments. 

Hear more about the art and science of AI in my recent podcast with Forbes Futures in Focus.