Data Deception: How Trust in AI Is Undermined by Data Quality
Data Quality heavily influence trust in Enterprise AI projects; 52% of respondents indicating that their organizations have processes in place to ensure data projects are built using quality, trusted data.
When it’s successful, Enterprise AI is enabling organizations of all sizes to embed AI methodology into their operations, suitably augmenting human capabilities of learning, perception and emotion with automated intelligence systems that can process data at a scale far beyond our own capabilities.
Now, at the very core of many organizational data strategies, Enterprise AI is essential for businesses wanting to guarantee their future success. Yet a report from IBM suggests the rate of AI adoption in 2019 lagged behind the level of interest, with an estimated 20% uptake. Further research based on a survey of 400 data professionals across the US and Europe has revealed that trust in Enterprise AI projects is being undermined by data quality, with just 52% of respondents indicating that their organizations have processes in place to ensure data projects are built using quality, trusted data.
When asked whether their organization had processes in place to ensure data science, machine learning and AI are leveraged responsibly and ethically, 57% of these same organizations stated that their business did not have the processes in place to ensure the responsible and ethical leveraging of data science, machine learning and AI technologies.
Trust in AI presents significant challenges
As with any new technological implementation, a period of trial and error is often experienced in terms of how a technology is to be best applied and adjusted for use within a business.
In AI, trust is a hurdle that demonstrably presents significant challenges. However, this lack of trust is not wholly due to a lack of understanding or education. In many cases, managers and employees may have had previous negative experiences with such technologies and are perhaps reluctant to place their faith in them again so soon.
Arguably, the only way to overcome this barrier is to implement AI with transparency, instilling a culture of openness, responsibility and quality at the forefront. With over half of organizations stating that they have appropriate processes in place to ensure data projects are built using quality, trusted data, organizations are experiencing their issues from the ground up.
Trust in AI projects will continue to raise significant challenges if we cannot solve fundamental issues such as data quality, or other, more complex problems associated with responsible AI. Building internal trust within organizations will ultimately provide the foundation for external trust.
However, this has to start with trust in the data itself that is being used across all AI systems, so that organizations are to build sustainable Enterprise AI solutions that deliver AI business value – not business risk.
Underpinning this lack of navigation and process is the situation many businesses are now finding themselves in with regards to data governance. The modern enterprise, in many cases, is ruled by shadow IT, whereby various departments have invested and relied on all kinds of different technologies over the years. In turn, these same departments continue to access and use data in their own ways. For organizations looking to implement new, innovative AI, machine learning and data tools, how can trust be instilled in the workforce when outdated processes and attitudes towards data storage and use still dominate?
Data access disparity
The research also reveals that the perception of the impact that AI would have on individual roles differs greatly from C-suite level roles to non-management. So much so, these findings suggest that AI projects also tend to struggle with inclusivity alongside an outdated data and technology culture, and naturally, this does not help with building a universal trust for AI in the workplace. For example, Managers and C-suite executives have been found to be significantly more likely to respond that AI would “completely” change their company than non-managers.
Despite the fact that non-managers in non-technical roles – including business professionals in marketing, risk, operations, and more – should see a positive impact of AI in their jobs, only 11 percent of these individuals believe that AI would “completely” change their role – a much lower percentage than the other more senior roles. This suggests that the problem with building trust in data roots deeper than just data quality, but by transparency – specifically, transparency with data implementation and plans.
Establishing a culture of inclusive AI will help to overcome the trust barrier currently experienced by many organizations. By implementing AI following an inclusive approach, a better outcome is usually achieved. This is because the process naturally involved more people, a better diversification of skills and points of view, and different use cases to build ideas upon.
Effectively, all businesses looking to implement enterprise AI tools must not restrict it to specific teams or roles, but instead equip and empower all to make day-to-day decisions and larger process changes that are navigated using data.
Data for all builds trust
It goes without saying: AI is and will continue to impact individual roles, enterprises and industries, yet there are clearly some questions around trust, responsibility and inclusivity that need addressing before we can achieve the benefits of Enterprise AI as well optimal results.
Enterprise AI is an organization asset, and when embedded properly into the core of a business, it is a methodology that will quickly deliver real business value. The difference between the businesses that struggle to get started on their Enterprise AI journeys and the ones that thrive will often come down to fundamental issues like data quality and trust.
If businesses can bring all people together, from business experts, to analysts and data scientists, they can take a collaborative approach that is conducive to building trust and transparency. And if they can prioritize responsible AI as well as sound governance practices, Enterprise AI can be accelerated to help grow businesses in a way that will assure all stakeholders, adapting to new requirements and creating new solutions in response to market demand.