Artificial intelligence is ready. But is the data ready?
How can businesses prepare to take full advantage of the insights AI can provide? Tools may be in place and talented people may be onboarded, but there may be gaps in the data. Yes, there is a lot of data flowing through the enterprise, but harnessing that data in an efficient and impartial manner is another story.
A new survey of senior data and analytics executives by Wavestone NewVantage Partners found that only 24% of organizations currently consider themselves data-driven, and only 21% have what could be considered a “data culture.” Furthermore, only 24 percent of companies say they have taken sufficient steps to ensure that data is used responsibly and ethically within their organization and industry. “Becoming data-driven is a long and difficult journey, and organizations increasingly recognize that it continues for years or decades,” note the study’s authors, Tom Davenport and Randy Bean. “Companies continue to lack focus and commitment to data ethics policies and practices.”
Mona Chadha, director of category management at Amazon Web Services, said the data gap may be the most pressing issue affecting AI success. “Companies need to be aware of things like poor data quality, unfair bias and lax security,” she said. “The predictive quality of an AI model depends heavily on the data used to train the model. Poor data quality can lead to inaccurate results and inconsistent model behavior, leading to a lack of trust from customers and internal stakeholders.”
Data bias and security are also issues that AI needs to address, Chadha continued. “It’s easy to fall into the assumption that AI can make decisions more impartially than humans. Unfair biases in the data used to train AI models can lead to discriminatory behavior that puts businesses at risk. Attackers continue to Attempts to exploit AI vulnerabilities. Businesses must ensure that AI systems are protected from adversarial attacks on their data and algorithms.”
When it comes to data quality, organizations need to focus on the processes used to oversee their data assets. “Existing data often resides in multiple databases and data warehouses, which often contain duplicate data, outliers, and irrelevant data points,” Chadha said. “There are also gaps in existing datasets. Organizations need better tools to clean and label data. Poor data quality can lead to inaccurate results and inconsistent model behavior, leading to a lack of trust from customers and internal stakeholders.”
Once the data gap is closed, organizations can start building the business case for advancing AI. “As AI becomes more popular, many business use cases across industries are seeing results,” Chadha said. “Examples include driving product innovation by accelerating drug discovery and training self-driving cars to navigate complex traffic scenarios. AI supports risk mitigation by helping fight financial fraud and reduce unplanned downtime of industrial equipment. Consumers also see AI Improve user experience by driving content engagement through recommendation engines or by using AI to assist human agents in improving customer service. Finally, AI is helping manufacturing through computer vision, making strides in overall efficiency and safety improvements.”
(Disclosure: Over the past year, I have worked as an independent analyst on projects for AWS mentioned in this article.)
