AI success relies on having the right mix of people, processes and technology. One of the main questions organizations are wrestling with is how to know if they are truly ready to benefit from AI at scale while operating within ethical and reasonable boundaries. This has become even more important in a world where Agentic AI is poised for widespread adoption. When an organization isn’t aligned correctly for AI, there are significant real-world consequences as we’ve seen with the recent Air Canada incident.

A recent TechTarget Enterprise Strategy Group Responsible AI study dug deep into many of the issues and concerns organizations have when looking to deliver responsible AI. The survey included 374 professionals at organizations in North America (US and Canada) involved in the strategy, decision-making, selection, deployment, and management of artificial intelligence initiatives and projects.

The paper clearly outlines that organizations need to combine strategies and investments to ensure they are using AI in ways that are ethical, free from bias, and safely contribute to both employee and organizational success.

Lack of Trust = Real World Impacts

The companies surveyed are keenly aware that underperforming in being responsible in AI is already having negative impacts on brand reputation, increased customer skepticism, and abandonment of future projects – all of which undercut the potential of AI.

The top five impacts listed by those surveyed as being either severe or significant include erosion of public trust (55%), distrust from stakeholders (55%), completely scrapped projects (54%), loss of market share (53%,) and reputational damage (55%).

Of note is how organizations see revenue impact due to responsible AI challenges. Of those surveyed, 15% said the loss of customers was severe, and 43% said it was significant. When combined with the loss of market share, we clearly see that not having the essential elements that support responsible AI is having bottom-line impacts.

Addressing AI Fairness Concerns

One area the organizations surveyed are focusing on is being able to measure AI fairness and ethics across multiple dimensions, crucial in targeting areas for improvement and heading off bias and compliance concerns. Interestingly, the top three areas being used to measure AI fairness and ethics – continuous monitoring and evaluation (46%), specific metrics such as demographic parity, equalized odds, individual fairness, etc. (45%) and maintaining compliance with legal and regulatory standards (44%) – all rely heavily on data and can be positively impacted by solutions that automate data classification, policy and governance application and alerts.

Tackling Speed, Compliance and Cost with Automation

As data estates grow and become more complex, it’s essential to make sure technology is supporting AI efforts across multiple vectors, including privacy, reliability, governance, resiliency and eliminating bias. The survey shows that escalating costs (37%), increased regulatory scrutiny (28%) and slower time to market (26%) were all factors that are creating pressure for businesses with responsible AI.

All these areas can be addressed through strong policies and automation. Across the entire data lifecycle (data access and transformation, classification, analysis, and data tiering and re-tiering based on usage and value) well defined policies create the right guardrails, and automation makes it possible to cost effectively and consistently apply and enforce those policies, ultimately increasing the availability of trusted and governed data for AI.

Data Management’s Role in Enabling Responsible AI

Modern data management serves a key role in providing the data foundation through which AI can deliver on its transformational promise while operating responsibly.

Creating a full understanding of the data used by AI requires a strong and scalable approach to data classification, lineage, quality and governance, leveraging metadata to truly understand what data is available and how it should be used. All these elements must work in concert to avoid bias and ensure completeness when creating the data products that will feed AI systems. This is a key driver in how we’ve designed the Pentaho+ platform to help organizations become data-fit and AI ready.

How does this play out in real life? One example is data classification. Say you have strong data quality processes in place. If it is misclassified, even high-quality data can be misused or misinterpreted based on assumptions. Another area is governance. Missing or incomplete data governance policies can allow PII or sensitive information to be unwittingly fed to open models, exposing the organization to potential fines or reputational damage.

Becoming More Responsible with AI Through Data Fitness

As the survey highlights, achieving responsible AI is a complex challenge that touches every aspect of an organization. As companies have looked to manage data at scale to safely and securely deliver what AI demands, they’ve exposed fundamental gaps in their data foundations.

The Pentaho+ Platform delivers battle-tested solutions to create foundational strength for meeting the challenges of an AI-driven world head-on. Our customers see measurable and tangible outcomes, including 3X improved data trust, 7X impactful business results and a 70% increase in productivity. To learn more about how Pentaho+ can help you become more responsible with AI, request a demo.