Inefficient data processing is holding back AI projects


In the Fivetran / Vanson Bourne study, 87% of respondents believe that their organization’s data scientists are not used to their full potential.

While artificial intelligence is seen as an opportunity by a majority of organizations, several data-related obstacles are holding back their initiatives, as revealed by a study conducted by Fivetran and Vanson Bourne.

AdvertisingArtificial intelligence (AI) technologies today appear to be an essential lever for digital transformation, as confirmed by a recent study carried out by Vanson Bourne for the publisher Fivetran. However, persistent obstacles prevent the deployment of large-scale AI projects.

A majority (87%) of the 550 IT and data science professionals polled in the survey see AI as vital to their survival, and companies seem ready to invest, planning to spend an average of 13% of their turnover about in the next three to five years, against 8% at present. However, respondents are not yet ready to fully trust AI tools. Thus, 86% of employees would have trouble delegating business decision-making to such tools, without any human intervention. Moreover, nine times out of ten the companies surveyed continue to rely on manual data processing for their decisions. Of the five countries represented in the study, including France, only 14% of respondents believe they have advanced maturity in AI. For more than four out of ten respondents, their organization has significant room for improvement in its use of AI. French respondents stand out on this point, with only 14% sharing this opinion, while a majority (79%) believe they have a small margin for improvement.

Poor data quality has a cost

Looking in more detail, the study reveals the presence of persistent obstacles to AI projects, particularly on the complex issue of data. For example, 71% of respondents struggle to access all the data needed to run AI programs, workloads and models. They are also more than seven in ten (73%) consider the steps of extracting, loading and transforming data as a challenge. These problems around data have a cost: companies estimate that they lose an average of 5% of their turnover due to AI models fed by poor quality data, which therefore do not offer performance at the expected level. .

Another issue stems from these data quality issues: the misuse of the skills of data scientists, who spend on average 70% of their time preparing data, compared to less than a third for designing models. 87% of respondents believe that their organization’s data scientists are not used to their full potential, with only 13% indicating that their data science teams can work with all data from operational systems. This gap between skills and the tasks actually performed is all the more problematic as data scientists remain rare profiles on the market, recruitment representing moreover the first obstacle to AI projects for 39% of respondents.

Share this article

We would love to thank the author of this write-up for this outstanding content

Inefficient data processing is holding back AI projects

You can find our social media profiles , as well as other pages that are related to them.