Achilles Heel in Artificial Intelligence is Data Silos. Here's the why of this analysis.

Sometimes when you install tools and software, it doesn't mean artificial intelligence will appear. It requires preparation and, most importantly data. According to a new study, many firms are failing to gather and comprehend the appropriate data needed to create AI and machine learning algorithms. According to the research of 150 data executives commissioned by Capital One and Forrester Consulting, organizations struggle with data silos, explainability, and openness. They claim that machine learning deployments and results were hindered by internal, cross-organizational, and external data silos. The majority of respondents, 57%, think that data silos exist between data scientists and practitioners, and 38% agree that these silos must be broken down within the company and with partners. Working with vast, diverse, complex data sets is difficult, according to 36% of respondents.

The Achilles Heel of AI may very well be data, according to analysts in the field. Ajay Mohan, principal, and AI and analytics head at Capgemini Americas, claims that there is a lack of data literacy that is stifling progress. Achieving this level of literacy, according to him, entails "understanding the value of data and knowing how to modify and use it to create value." He notes that a common problem for many businesses is that they "often lack the proper personnel, such as data scientists, data engineers, or technology-oriented subject matter experts to look at the business difficulties and the potential for data to unlock solutions to these challenges."

Additionally, quantifying business value or ROI for another data-driven activity is frequently challenging. In addition, Mohan notes that many end users are lacking in this fundamental skill. "Challenges using data from diverse legacy systems and sources that can make it cost-prohibitive to construct really significant AI applications" should be included in the list.

Lack of data literacy leads to data silos, which impede AI. Even if firms had the means to be data literate, operational silos—business functions, geographic teams, or other business lines functioning independently from their intra-company peers—remain a significant issue for many larger enterprises, according to Mohan. For instance, "many large consumer goods corporations may operate dozens or hundreds of brands globally, each employing their individual marketing teams and IT frameworks."

Additionally, there is "inertia within firms" that needs to be overcome in order to move toward data literacy, connection, and human abilities. According to Krishna Tammana, chief technology officer at Gupshup, "They are either investing and not realizing the value or not investing enough time and money to make it successful in data management systems." High-quality data, which many firms lack, is a must for effective AI. However, giving AI systems complete access to data could be problematic since it could introduce bias and incorrect information. According to Peter Gordon, worldwide head of AI product at Hogarth Worldwide, "We evaluate every adoption of new AI discoveries through the prism of ethical responsibility." "We carefully consider how to prevent abuse and damage that can be brought about by algorithms that have an inherent bias in the training data. This is more of a matter of due diligence than a problem, but it will understandably slow down adoption.

There are some use cases where data has been successfully tapped into, so not everyone is losing out. According to Tammana, "There have been some specific use cases with the correct quality of data that are performing well." Customer engagement via conversational AI is one area where we see it gaining a lot of traction. There is much room for improvement in the existing environment for consumer involvement. We have observed improved conversions, retention, and brand recall using AI-powered tailored dialogues.

More sophisticated types of AI will become possible if we have a complete understanding of the data required to ensure better accuracy in output. The next transformation, according to Gordon, is generative AI, which uses data to create automatically brand-new headlines, photos, movies, and music. "The exponential rate of maturation is amazing to witness. And we approach this with enthusiasm but prudence to make sure it is prepared for widespread deployment.