Nabugu -

Advanced analytics challenges limit analytics, AI benefits

Traditional advanced analytics are not going to disappear with the rise of generative AI. Cultural and quality challenges hinder organizations as attention shifts to new trends.

Generative AI is poised to transform many facets of data and analytics, but it won't provide the value organizations expect if their data management practices are lacking. They must solve current advanced analytics challenges first before incorporating the latest AI tools.

Large language models (LLMs), such as GPT-4, are good at understanding ideas and coming up with readable text. They're not well suited to tasks that legacy systems and advanced analytics tools are already good at, such as managing transactional or structured data or doing high-speed data processing.

As a result, some organizations are considering a hybrid approach. They might use generative AI to automate some tasks and continue to use traditional analytics and data tools for others, said Traci Gusher, data and analytics leader at EY Americas. Consider a common use case for LLMs: customer service chatbots.

"Generative AI can answer questions in a chatbot for a more seamless and better user experience," Gusher said. But, to answer specific questions about order status, the chatbot must be able to access accurate customer data. Traditional data processes create those clean data sets, she said.

As exciting and sexy as the AI is, if we're not thinking about how we manage the data to make the AI useful, organizations are going to spend a lot of money on AI they can't use.
Traci GusherData and analytics leader, EY Americas

Traditional analytics tools require accurate, clean, well-governed data. But new generative AI platforms require clean data as well, whether it's for training data sets to create custom models or powering the back-end systems that provide data to customer service chatbots in real time.

"Organizations that haven't invested in robust data governance projects will quickly realize that the AI they're trying to apply is garbage because the data they're tapping into is garbage," Gusher said. "As exciting and sexy as the AI is, if we're not thinking about how we manage the data to make the AI useful, organizations are going to spend a lot of money on AI they can't use."

Lack of data quality

Most companies are still in the process of building secure, governable and accessible data infrastructure, said Bradley Shimmin, chief analyst for AI and data analytics at Omdia.

Data quality problems have a big impact on performance and are getting worse. On average, 31% of a company's revenue is affected by poor data quality, according to a May 2023 Wakefield Research survey of 200 data professionals. That figure is up from 26% in last year's survey. It also showed that data quality issues take longer to resolve, with a 166% increase in average time to resolution.

Lack of focus

Many companies aren't focused on benefit to the business and the value that advanced analytics can provide. That applies to traditional advanced analytics projects and to new generative AI use cases. When IT decision-makers ignore what the company needs, it can result in wasted time and money, cause analytics projects to fail, and make users and executives lose faith in analytics altogether.

"They focus on the capabilities instead of what they can do with them," said Donncha Carroll, partner and head of the data science team at Lotis Blue Consulting.

Carroll encourages a human-centered design approach, which focuses on meeting the needs of the people who will use the tools and working with them through the problem-solving and implementation processes. This also means keeping lines of communication open going forward to continue work with business users after the original implementation phase is over.

Advanced analytics projects need a mechanism for monitoring usage, as well as a follow-up process to capture user feedback and to confirm that the project is providing its intended benefits over the long term.

Lack of culture

There are many cultural and organizational issues that companies aren't dealing with when it comes to putting data to good use. Only one-quarter of companies described themselves as data-driven, and 21% said they have a data culture in their organizations, according to a January 2023 survey of data and information executives from NewVantage Partners.

Employees often don't know where to go to ask questions about the data or how to get access to a new data set that they might need to work with. Even though most companies now have communication and collaboration software, many still lack internal communication channels specific to data and analytics.

"Do they have a community that they can turn to? Adult learners need not just the education, but community support," said Kim Herrington, senior analyst on the business insights team at Forrester Research.

When employees do know where to go to find the data or ask questions, they often do not feel safe doing so. They need to be able to have honest conversations about data.

"I'm talking about psychological safety. This is something the tech industry doesn't like talking about: the ability to speak up without fear of rejection or ridicule," Herrington said.

Dig Deeper on Data science and analytics

Data Management
Content Management