Data and analytics leaders say they are unable to meet business leaders’ high expectations for artificial intelligence and machine learning initiatives because they are understaffed and underequipped, according to a new report titled Build A Winning Data Analytics Offense: C-Level Strategies for an ML-Fueled Revenue Engine.
In a survey of 100 U.S. chief data officers and chief data analytics officers conducted by Wakefield Research on behalf of Domino Data Lab, 95% said that company leadership expected investments in AI and ML applications to pay off with growth in revenues. One-third (33%) expect a revenue increase amounting to a double-digit percentage.
According to the study, just 19% of CDOs and CDAOs surveyed said they had the resources necessary to meet their bosses’ expectations, with 29.4% saying there was a “meaningful shortage” in the staff, funding and technological resources they needed to drive revenue growth using AI and ML.
A shortage of tech skills was identified as a major concern, with 87% of respondents saying their inability to recruit and backfill data science roles was hindering their organization’s ability to innovate in the field.
Likewise, 81% of respondents reported that their current tools lacked the ability to fully measure the impact that their AI/ML initiatives had on revenue, leaving data teams “flying blind” with their applications.
- Why CDOs and CDAOs want more purchasing power
- Moving from “defensive” to “offensive” applications
- The risk of under-equipping data teams
- How business leaders can close this gap
- Survey methodology
Why CDOs and CDAOs want more purchasing power
Budgets — and more precisely, those in charge of budgets — were identified as one of the biggest sticking points for CDOs and CDAOs.
Nearly two-thirds (64%) of respondents reported that their company’s IT department controlled the majority of spending decisions around data platforms, with data and analytics teams only having a say in around 56% of purchases.
CDOs and CDAOs alluded to competing priorities between data and analytics teams and the IT department when it came to tech spending: 99% said it was difficult to convince IT to focus budgets on data science, ML and AI initiatives as opposed to traditional IT areas like security, interoperability and governance.
Data leaders suggested that the lack of purchasing control had an effect on staffing and hiring, with 99% of CDOs and CDAOs reporting that not being able to provide data and analytics teams with their tools of choice had a negative impact on their ability to hire, retain and upskill tech talent.
Moving from ‘defensive’ to ‘offensive’ applications
CDOs and CDAOs feel even more pressure to wrangle control of their organization’s AI/ML initiatives now that business leaders want to make more innovative use of their data, the study found.
Two-thirds (67%) of respondents said their strategy was moving from a “defensive” posture that centered around data management, governance, compliance and business intelligence modernization to a more “offensive” strategy that aimed to drive new business value through innovative AI and ML applications.
As such, 98% of data leaders agreed that the speed at which organizations could develop, operationalize and improve AI/ML applications would “determine who survives and who thrives amid persistent economic challenges.”
Because of this, another 67% of CDOs and CDAOs felt that it was “time to take the reins from IT” to prevent their organization from falling behind, with Domino Data Lab concluding that IT departments “[do] not have the remit to drive AI/ML innovation.”
The risks of under-equipping data teams
Besides falling behind rivals and missing out on new, data-driven revenue streams, ill-equipped data teams face more immediate risks: 46% of surveyed CDOs and CDAOs admitted they did not have the governance tools required to prevent data teams from introducing risks into the organization, while 44% felt that a failure to properly govern their AI/ML applications could result in revenue losses of $50 million or more.
“Today’s vast and quickly-evolving regulatory landscape, paired with the high-stakes of many enterprise data science initiatives, means that a lack of trustworthy AI could cost tens of millions,” said the report.
Kjell Carlsson, head of data science strategy and evangelism at Domino Data Lab, said the findings were “sobering” and warned against pressuring data leaders to do more with less.
“Leaders are struggling with the persistent challenges of hiring and retaining data science talent, getting IT to prioritize investment in AI/ML over traditional priorities like data management, and weak capabilities for managing and governing AI/ML models,” Carlsson said. “CDAO and CDO roles are already notorious for their rapid turnover, and this widening gap between expectations and the ability to deliver does not bode well for their life expectancy.”
How business leaders can close this gap
Carlsson urged business leaders to invest in their organizations’ ability to scale the development and deployment of new AI/ML-based applications across more parts of the business.
Additionally, in order to attract and retain talent, organizations should invest in supplying data scientists with the “broad range of different tools” they are trained on, as opposed to just a handful of proprietary tools dictated by the IT department.
SEE: Report: The ROI of upskilling and other employee learning programs (TechRepublic)
“To accelerate time to value and impact, they need to invest in MLOps platforms that span the end-to-end ML model life cycle from development to deployment, monitoring and retraining,” Carlsson said. “To accomplish this, CDAOs and CDOs need to build alignment and a close working partnership with IT. If that’s not possible, they have no choice but to implement these platforms themselves.”
The Domino Data Lab survey was conducted by Wakefield Research among 100 chief data officers and chief data analytics officers at U.S. companies with more than $1 billion annual revenue between Dec. 5 and 18, 2022, using an email invitation and an online survey. According to Domino Data Lab, the margin of error for the study was more or less 9.8%.
Read next: Hiring kit: Data architect (TechRepublic Premium)