Are data infrastructures ready for (AI) prime time? Survey says no
The inability to access all information across organizations, ensconced in silos, rises as a major impediment achieving strategies and ROI.
Whether consumers trust artificial intelligence enough to jump on board is one well-covered challenge.
Whether healthcare organizations have the data “gasoline” to power the engine is an entire other concern.
A recent study suggests that a variety of organizations worry that the quality of data that’s feeding AI is suspect and that they lack sufficient data readiness.
The survey, by Cloudera, surveyed 1,200 IT leaders worldwide, working in a variety of industries, including healthcare. It identified a crucial gap affecting the effective use of AI – while organizations are ready to embrace new frameworks to improve data readiness and their infrastructures. However, nearly four out of five respondents say their data initiatives are struggling because their organizations can’t access 100 percent of the data needed across all their environments.
Infrastructure immaturity, lack of standardization and insufficient interoperability are factors that will need to be solved for AI to achieve the change that leaders envision. That’s particularly true for healthcare organizations, which need to draw data from various siloed systems to obtain the most trustworthy and impactful results.
Measuring data readiness
Cloudera researchers note the rush to implement AI across multiple industries. But while use of the technology is scaling fast, “many organizations are still searching for impact, wondering why projects aren’t generating the intended ROI … The actual challenge is a matter of data readiness.”
Researchers define data readiness as the ability to fully govern, access, integrate and trust data across all environments so it can support AI, analytics and operational decision making. They note that data readiness may mean different meanings depending on the industry – in healthcare, challenges come from complicated access requirements and processes (identified by 45 percent of healthcare respondents).
Using the cloud is one way to begin to achieve data readiness, but organizations are still struggling to access their own data “to feed into AI and analytics initiatives.”
Those conducting the survey were asked how well their infrastructures support self-service access for technical users – fewer than one-third (31 percent) said they fully supported the capability.
Even though respondents report struggles in achieving data readiness, they “signaled a strong awareness of data strategy, responsibility and structure within the broader enterprise,” researchers said, concluding that from the fact that they know which leaders are in charge of the effort. Most respondents (63 percent) said CIOs and CTOs were ultimately responsible for data readiness, far ahead of chief AI officers (15 percent) and chief data officers (12 percent). And nearly nine of 10 respondents said their organization’s senior leadership is prioritizing the necessary data infrastructure to scale AI.
Impediments to readiness
A previous survey by Cloudera found that organizations were embracing AI for critical business operations, but integrating AI isn’t a guarantee “of a meaningful return on that investment,” researchers contended. Roadblocks were identified by the current year’s survey that now stand in the way of achieving results.
Factors that impede ROI include data quality, cited by 22 percent of respondents; cost overruns (mentioned by 16 percent); and weak integration into workflows (cited by 15 percent).
Data silos are one of the most commonly cited impediments dragging down infrastructure performance needed for AI and analytics, respondents noted. More than a third (34 percent) of respondents cited these silos as “a top issue preventing them from collaborating, sharing, managing and using data effectively.”
Silo factors include many that are well-known within healthcare organizations, including complicated access requirements and processes, limited visibility into where data resides, insufficient training and data literacy, and cultural resistance to data sharing.
Respondents affirm that data quality, ensured by top-level governance, is crucial to achieving ROI with artificial intelligence. A key capability they cited is optimized management of unstructured data. “Any governance framework needs to account for all forms of data – structured, unstructured and semi-structured – to ensure accessibility and usability with data-backed initiatives.” Nearly one in five respondents said all, or almost all, of their data is unstructured.
Healthcare’s relative strength
Across various industries, healthcare appears to have achieved a degree of data readiness, as defined by Cloudera.
Some 87 percent of respondents in the healthcare segment said it was “extremely true or very true” that they have “complete visibility into where 100 percent of my organization’s data resides.”
In terms of data accessibility, four out of five healthcare respondents (80 percent) responded either “extremely true or very true” that they can access “100 percent of my organization’s data at any time, regardless of format or where it exists.”
As an industry segment, healthcare trails only the telecommunications industry.
However, data visibility and access haven’t necessarily translated to operational success in healthcare. ROI in the segment is likely to fall short because of weak integration into workflows, according to 20 percent of respondents.
While respondents overall generally reported they were confident that their current data infrastructures could support their organizations’ strategic priorities in the near future, they generally are finding operations hindered by infrastructure challenges and ROI hampered by longstanding complications with siloed data.
Overcoming these infrastructure-based and under-the-hood challenges are thus key to implementing AI and achieving measurable operational benefits, the researchers conclude.
“As enterprise AI shifts from experimentation to execution, data readiness is emerging as the defining factor separating leaders from laggards,” they note. “Organizations able to fully access and govern all their data, wherever it resides, are far better equipped to deliver trusted, scalable AI. Notably, every respondent in the report indicated their organization is at least somewhat willing to adapt existing frameworks to support true data readiness.”
Fred Bazzoli is the Editor in Chief of Health Data Management.