Andrea Danti - Fotolia

AI struggles with data silos and executive misconceptions

Bosses don’t seem to understand that artificial intelligence needs good data management and a different understanding of end goals compared with traditional IT

Businesses recognise that they need good-quality data in order to deploy new initiatives powered by artificial intelligence (AI), but many admit their data management needs improving, a study by Comptia has reported.

The report, Emerging business opportunities in AI, based on a survey of 500 organisations, says more than three-quarters of companies recognise that they have data silos.

The survey found that only 18% of small companies feel they have a high degree of data silos. Comptia surmised that many of these companies may be so small that their data is consolidated, but many others may simply be unaware of where their different data sets reside.

The survey reported that mid-sized companies are the most likely to see a high degree of data silos, with 44% recognising this to be the case. “As they have grown, these businesses have accumulated data sets without always having a plan for how everything works together,” the report said. “The largest companies may have the resources to build and execute a formal data management plan, but 35% still believe there is a high degree of data silos.”

According to Comptia, these data silos limit the abilities of AI, especially in the area of discovering patterns or correlation.

A global chief operating officer at an IoT business quoted in the report warned that it is not enough for organisations simply to collect data. “The reality is, there is a lot of data work that has to happen before it can be used,” he said. “In order to be able to use data to its fullest potential, there are different back-end systems that need to be able to access the data. I think the bottom line is that being able to collect data doesn’t mean that it’s going to be readily usable for AI right off the shelf.”

Comptia also found that a lack of organisational capacity to capture the robust data sets needed for AI algorithms and AI workloads has resulted in limited strategic deployments by most companies. It said this has often led executives to identify the areas of their business that are most readily able to adopt AI technology or would stand to gain the most immediate value.

The research suggested that business leaders may not fully appreciate what is involved in an AI project, compared with traditional software development. It argued that traditional software development is a deterministic process, with a project specified to solve a particular problem. But AI is probabilistic, which it said implies a different approach to managing the project’s outcomes.

Read more about data management for AI

  • Many enterprises think their AI projects should start with massive amounts of data. But clean data for machine learning should be their first step toward AI.
  • Looking to AI to automate more of your processes? Don’t overlook the labour that’s still needed to prepare data for training machine learning and AI algorithms.

“Probabilistic outcomes imply an assessment of one or more variables to generate a prediction,” the report noted. “Adding more relevant variables [data points] increases the likelihood of more accurate predictions.”

Comptia noticed that among the people who took part in the study, there was acknowledgement of failures in their organisations to appreciate the intense rigour that AI algorithms require. For some, this has led to the development of misguided solutions to common bottlenecks, said Comptia.

“Several of the study’s respondents reported dealing with bosses who tended to overlook the relevant variables component, mistaking their initial investment in architecture enabling the capture of large data sets as having a ready-made AI solution,” it said.

Read more on Artificial intelligence, automation and robotics

Data Center
Data Management