Deloitte AI Institute
Artificial intelligence, share the limelight with data leadership
By Craig Brabec and Beena Ammanath
Why do many AI projects fail?
Artificial intelligence (AI) has arrived, turning the corner and revealing itself in landmark successes, in multiple forms across industries.
Computer vision is driving transformation, from augmentation of expert patient diagnosis in health care to advancing autonomous vehicles. Chatbots and virtual assistants are everywhere, improving and personalizing the customer experience. Deep learning is advancing decision sciences, processing tremendous volumes in record time for faster, automated business actions that drive value.
Yet with these substantial accomplishments, as some companies gain advantage in pockets of their business, many are still struggling with wider adoption and success. Why?
Increasing your AI project’s chances for success
New platforms and solutions are abundant. Data scientists are better equipped than ever before. But the same data challenges remain—availability, quality, accuracy, completeness, and trust. Why? Effective data leadership often remains a challenge as companies typically relegate it to the back office with reduced senior leadership attention and investment focus. While many CXOs confirm the need for “good data” across the enterprise, there is a gap between what is needed and what is delivered at the operational level. With resources constrained, will the new AI technology project with potential revenue growth be selected over the data governance initiative with difficult-to-quantify benefits?
Companies can accelerate their AI journey by focusing equally on the data journey. These efforts can—actually should—run concurrently. Without an equal focus on data and AI, companies will likely not maximize their opportunity for advantage. To aid in elevating the focus on data leadership, three key themes may need a reboot:
- Develop a well-articulated data strategy
- Elevate and accelerate data governance
- Sunset data literacy, and launch data enablement
of surveyed respondents reported that AI initiatives are important for organizations to remain competitive over the next five years
of surveyed respondents completely agree that they have access to the necessary data sets for their AI initiatives
Develop a well-articulated data strategy
A data strategy primarily serves as a vehicle to achieve a vision or a set of objectives. This can be to maximize use of data assets; to achieve specific priorities with data as a driver, enabler, or utility; or to maximize the value of data. To do this, position the data strategy as an enabler for the business strategy.
Once developed and widely communicated, management systems at all levels of the organization should be adjusted to activate the data strategy, especially in business operations. This includes performance management metrics for treating data as an asset.
For example, customer representatives at call centers are often evaluated with two measures:
- Time to successfully resolve an issue, and
- Customer post-call surveys.
Accurate and complete data capture from the customer interaction should be measured as well. In the spirit of speed, representatives will sometimes skip entering key data or just plug in “xxx” or “999”. They are not aware of the value of that incorrect data to downstream algorithms for tools such as recommender systems or next best actions. Treating data as an asset means doing so at every point of entry. This would help bring the strategy to life.
Elevate and accelerate data governance
Effective data governance is required for the sustained success of AI initiatives, yet many organizations continue to struggle mastering it. All too often data governance is handled as an IT function, relegated to leaders under the CIO who are responsible for the data management activities of the organization. The business should sharing accountability by:
- Emphasizing policies, roles, and processes with business stakeholders. Decrease the priority on technology solutions. Adopting leading practice data governance behaviors and organizational principles is more important than the technology deployed.
- Building a data governance playbook, drive with change leadership, and measure adoption. Identify success criteria for key business outcomes, and directly connect these with the data and analytics assets that enable them.
- Building “light touch” data governance, balancing business area and technology needs. Help ensure governance has influential business leaders to drive the adoption across the enterprise.
- As data governance matures, merge efforts with AI model governance, as many of the stakeholders share equal engagement priority with both the data and the model usage.
Sunset data literacy, and launch data enablement
Gartner defines data literacy as the ability to read, write, and communicate data in context, including an understanding of data sources and constructs, analytical methods and techniques applied, and the ability to describe the use case, application, and resulting value. This is technically complete, yet for a business it may be insufficient, as there is no emphasis on application and usage.
Data literacy programs have a tendency to over-focus on technology training, offering online self-paced programs for the individual. While this may be an efficient means to deliver training, it is typically not how effective use of data occurs in organizations today.
Data is a team sport
Everyone should share a common understanding of the problem being solved, the data being used, and the analytical method deployed. Enablement should follow the same method, with group training using real-world problems with real data.
Lastly, enablement should include a change in the ways of working, which requires a support network for knowledge professionals. This can take on many forms, including establishment of new internal working groups focused on common data sets, communities of practice across functions, and even organizationwide education sessions introducing emerging data topics and priorities for the business.
Remember, creating a data-driven culture means creating sustainable networks across the organization focused on data. The chief data and analytics officer (CDAO) fuels this with evangelitics—communicating success stories across the organization, creating forums for sharing knowledge, and promoting the business strategy enabled with AI and data.
The promise of AI is being delivered today. We are still early in the journey, with so much more enterprise value to come, and that promise can be delivered with velocity by embracing a critical component: the data journey.
Pack your bags.