entertainment (59 percent), higher education (53 percent), and aerospace and defense (48 percent) feel the data quality challenge particularly keenly. According to S&P Global analysts, the data quality challenge does not stem from a lack of data to build performant models. Rather, at issue is data not being set up in such a way that project teams can take full advantage of it. When asked specifically to rank the primary data challenges to move projects to production environments, respondents indicated that availability of quality data is a more notable impediment than identifying relevant data. “With 34 percent of organizations perceiving availability of quality data as a top three data challenge, outranked only by data privacy concerns (35 percent), it is clear that many organizations are poorly set up for effective data management,” said the research firm. Data management and storage are most commonly seen as the infrastructure components that inhibit AI application development. More than a third (35 percent) of respondents see them as a more serious issue than security (23 percent), compute (26 percent) and networking resources (15 percent). What’s more, organizations that are most effectively scaling AI initiatives are less constrained by these data management and storage components. Whereas 28 percent of respondents who reported that AI is widely implemented within their organization perceive storage and data management challenges as their greatest inhibitor; these respondents feel greater pressure from networking or compute resources. This compares to 42 percent of respondents who perceive AI as being limited to a few use cases or projects within their organization. “Organizations that are delivering AI at scale appear to have focused on investing in upgrading the systems and technologies used to store or manage data,” said Weka executives. More than three-quarters (80 percent) of respondents expect an increase during the next 12 months in the volume of data they use to develop their AI models, with nearly half (49 percent) forecasting growth in data volumes of more than 25 percent. All the while, the proportion of organizations using unstructured rich media and text data for AI initiatives has increased notably since 2023. In other words, immature data management toolsets are a worrying backdrop for the increasingly datahungry AI strategies many organizations are embarking upon, the study noted, and outdated data management technologies may prevent organizations from delivering these projects meaningfully. “Leaders are significantly less likely to see storage and data management as their primary inhibitors, presumably because these companies have already prioritized modernizing their data architectures.,” Weka executives concluded. “By building a solid data foundation at the outset, AI leaders have ensured that valuable pilots have a clear path to deliver at scale.” o Top three impediments to organizations moving an AI/ML application from pilot to production environments Source: Uptime Institute; 2022 Source: S&P Global; Weka Organizations find the early data steps of the AI life cycles as challenging as Source: S&P Global; Weka 18% 13% 12% 12% 11% 10% 10% 10% 11% 12% 9% 10% 8% 10% 10% 8% 9% 10% 8% 8% 9% 6% 7% 9% 6% 8% 7% 6% 7% 7% 6% 8% 6% Proportion of respondents that identify AI life cycle stage as “most challenging” Data pre-processing Model build an Criteria Used in Selecting Third-Party Firms for Cybersecurity Source: CompTIA Data quality Rank 1 Rank 2 Rank 3 Skills shortages Availability of AI accelerators Budget limitations Compliance/regulatory requirements Legacy infrastructure cannot support AI/ML applications Employee/internal resistance Illustrating the business case for further investments Potential reputational damage Insufficient vendor tooling Customer resitance 59% 41% Gathering/ sourcing data Preparing data Standardizing data Analyzing data Training a model Testi a mo Access to threat intelligence Specific knowledge in a focused area of cybersecurity Broad knowledge across multiple domains of cybersecurity Clear remadiation policies in event of cybersecurity incident Excellence in core offering where security may be embedded Ability to perform cost/benefit analysis of initiatives Offer cybersecurity insurance 44% 43% 43% 41% 39% 38% 33% Top three impediments to organizations moving an AI/ML application from pilot to production environments Source: Uptime Institute; 2022 Source: S&P Global; Weka Organizations find the early data steps of the AI life cycles as challenging as model building Source: S&P Global; Weka 18% 13% 12% 12% 11% 10% 10% 10% 11% 12% 9% 10% 8% 10% 10% 8% 9% 10% 8% 8% 9% 6% 7% 9% 6% 8% 7% 6% 7% 7% 6% 8% 6% Proportion of respondents that identify AI life cycle stage as “most challenging” Data pre-processing Model build and deployment Data quality Rank 1 Rank 2 Rank 3 Skills shortages Availability of AI accelerators Budget limitations Compliance/regulatory requirements Legacy infrastructure cannot support AI/ML applications Employee/internal resistance Illustrating the business case for further investments Potential reputational damage Insufficient vendor tooling Customer resitance 59% 41% Gathering/ sourcing data Preparing data Standardizing data Analyzing data Training a model Testing a model Deploying a model 12 CHANNELVISION | SEPTEMBER - OCTOBER 2024
RkJQdWJsaXNoZXIy NTg4Njc=