Data Quality Remains Biggest Detriment To AI Success

In the past few years, organizations of all industries and sizes have looked to artificial intelligence as a way to optimize productivity, customer satisfaction, and other metrics. While many have succeeded, one of the biggest impediments to a successful AI deployment is the quality of data being collected and analyzed by the AI program. 

AI models are fed off of whatever data is inputted, and will spit out insights and analysis based on it. If the data is inaccurate or fails to take into account human biases, expect the AI model to provide similar inaccuracies to customers and clients. 

See Also: How Digital Onboarding Fuels Data-Driven Insights

“There are issues that companies need to be aware of, such as poor data quality, unfair bias, and lax security to name a few,” said Mona Chadha, director of category management at Amazon Web Services to Forbes. “Quality of predictions of AI models depends strongly on the data used to train the models. Poor data quality can result in inaccurate results and inconsistent model behavior, leading to lack of trust from customers and internal stakeholders.”

Good data practices have to start from the ground-up to have the most impact on AI model performance and sophistication. Having data governance, ethics, and standards also ensures that future AI development follows a similar data collection and quality structure, which can improve time to market and reduce the likelihood of failures.

“Becoming a data-driven organization requires time, persistence, and relentless execution and focus,” said NewVantage Partners innovation fellow, Randy Bean. “Those organizations that commit to the course while adapting over time tend to prevail – fail fast, learn faster. Change often comes slowly, however. We have noted that organizations must be receptive to change if they intend to make serious progress. Large, legacy businesses don’t easily change overnight. It is always a journey that unfolds over time, with progress and setbacks along the way.”

To build a data-driven culture inside a business, organizations can create C-suite executives directly responsible for data management and quality. Chief Data and Analytics Officers are key In a survey by NewVantage Partners, 84.6 percent of companies surveyed had a senior data leadership title within their organization, which is up 20 percent from five years ago. The number has also increased steadily outside of financial and banking industries, showing a greater diversity of businesses embracing data as a valuable resource. 

Even with improvement in the amount of organizations hiring data executives, the survey also indicated that data executives tend to have a broad remit with a lack of clear expectations for the role. This leads to dissatisfaction when the AI deployment is not going as expected, especially if it is to do with an issue of data quality. 

Hiring a data executive will not automatically make a business data-driven however, as there needs to be real change inside the organization to ensure that processes and decisions are backed up by data. It makes sense that in the same survey, just over 20 percent of businesses said they were data-driven, even though over 85 percent of them had hired a data executive.