Businesses everywhere are striving to become more data-centric, whether to improve customer experience, develop smarter products and services, or fine-tune decision-making. But while these initiatives rely heavily on technologies such as big data analytics and artificial intelligence (AI), developing a successful data-driven culture is not a technology problem to solve.

The main ingredient of a data-driven culture, in fact, is not technology but humans. And, as many businesses have found, if the human factor is not considered upfront, data-centric initiatives will go nowhere.

This is easy to overlook, especially as businesses turn to AI-driven technologies to derive insight from their increasingly unmanageable data volumes. Given that these systems are specifically designed for self-learning using machine-learning algorithms – and ultimately for discovering patterns to make autonomous decisions and predictions about future outcomes – it’s easy to imagine that humans might as well be taken out of the loop.

The Human Element of Machine Success

This, however, couldn’t be further from the truth. When businesses kick off an AI-driven data initiative, human workers are integral, from providing feedback to algorithmically-produced outputs so that the system can learn and improve over time, to remaining at the center of the system’s design. To do that, workers need to fully understand, buy into and trust the intent of the system.

When I was at a recent conference, for example, a speaker recalled an organization he’d worked with that had created a machine-learning algorithm and given it to a team of workers to validate the system’s output. With no context for what they were doing, team members assumed their role was to provide only positive feedback for the system’s conclusions. They incorrectly surmised that to play a supportive role in the program’s success, they basically had to say the machine was always right. But instead of improving the algorithm, they led it astray.

In other cases, workers may distrust the intent of the program (perhaps believing it would result in the loss of their own job) and intentionally bias the algorithm.  In either case, it’s essential for employees to understand what their role is in guiding the machine, and to know that the interplay between humans and the machine is essential to the program’s success.

We’ve seen similar dynamics play out when machine-learning systems are deployed in a call center organization. Here again, humans are integral to ensuring that the machine’s learnings are based on “good” customer service interactions in order to automate a positive experience. Because the current way of working may not necessarily provide the best inputs, a good amount of human input is needed to define a good interaction and ensure the machine is learning from good, not bad, examples.

Humans: The Focal Point of Design

Humans also need to be the focal point of solution design, especially when huge volumes of data are involved. At many businesses, IT works hard to wrangle reams of data into what they believe to be insightful reports. The problem is, for many employees, there are too many insights to take in.

By combining human-centered design thinking, advanced analytics and AI, businesses can produce a system that – like Waze for traffic navigation – provides a “best route” for leading employees through the data and decision options.

We worked with a global consumer products company on an initiative like this. Sales teams at the organizations were struggling to get a handle on the company’s data to improve client interactions and thereby sales. The data team was frustrated because they felt they’d provided a solid asset, complete with relevant data sets and dashboards. However, the sales teams felt they were drowning in data, with no clear idea of which data was most important to look at.

To resolve the issue, we orchestrated a design-thinking approach, in which we spent a “day in the life” with several stakeholders, working to understand their underlying needs. By developing journey maps, we were able to represent how all the data needed to be synthesized for specific user personas, and how to present it, through which channels, and at what point in time. 

Based on that work, we designed a daily sales digest that incorporated customized key indicators, interactive data visualization and inferences that bring the data to life through supporting text, using natural language processing. The digests now reflect the needs of each user; for instance, a top-performing sales person might be less worried about meeting his or her numbers because of a solid pipeline and more concerned about cross-selling or upselling recommendations.  The system also goes beyond showing critical data in chart format and describes the insights in a more easily understandable, natural language format, such as month-to-month regional insights.

All of these needs were uncovered through joint workshops performed with the sales community. Because users can absorb inferences and interpretations more quickly, system adoption increased, and decision-making has improved. 

Businesses need a solid foundation to enable faster intelligence across the enterprise. That can’t be achieved through technology alone. When it comes to strategic use of data – and successful data-centric cultures – humans are still very much a part of the solution.

Poornima Ramaswamy

Poornima Ramaswamy

Poornima Ramaswamy is Vice-President of Cognizant’s AI and Analytics Practice. With her 20 years of experience, she consults and works with clients across... Read more