#Paris #France #AI Governance #Sopra Steria #decision-making

Understanding the Need for Governance in Automated Decision-Making in AI

In a rapidly evolving digital landscape, artificial intelligence (AI) has become a transformative force across various sectors. However, with its integration into decision-making processes comes the pressing need for accountability and governance. In the latest issue of Sopra Steria Next's CIO Compass, the spotlight is cast on the stark reality that less than 20% of AI-executed decisions have an identified owner.

The Governance Gap in AI

Despite significant investments and efforts made by large organizations over the last decade in managing data governance, they have largely overlooked the governance of automated decisions. Traditional approaches have focused on data management; however, as AI progresses beyond merely making recommendations, the urgency for a robust decision-making framework becomes evident. The report titled 'AI is Nothing Without Control and Purpose' outlines a nine-dimensional framework that emphasizes the need for organizations to govern not only their data but also the automated decisions stemming from it.

Key Figures from the Report

• - Less than 20% of decisions influenced or executed by AI have a clearly defined business owner.
• - Under 25% of AI-driven decisions in practice are traceable and replayable from end to end.
• - The impact of a faulty automated decision can affect a staggering 10 to 100 times more operations, customers, or revenues compared to an ordinary IT failure.

What is AI Sovereignty?

AI sovereignty is multifaceted; it encompasses more than just awareness of data storage locations. It also involves the ability to demonstrate who controls, interprets, and arbitrates these automated decisions at any given time. The reliance on vendors has transitioned from infrastructure management to cognitive resources, a shift that is difficult to reverse. Sopra Steria Next urges CIOs to clarify a five-layer 'sovereignty stack' which includes data, models, platforms, APIs, and decisions, each demanding accountability.

Reassessing Accountability

Valérie Thiblet, Partner in Data & AI at Sopra Steria Next, highlights the critical new question regarding decisions made by AI without anyone being held accountable. With automated systems, the lack of real-time oversight translates into governance failures. Automation that cannot be halted at a moment's notice lacks the governance necessary for industrialization.

The Challenges Ahead for IT Departments

The document particularly identifies several critical challenges that IT leaders now face:
1. Shadow AI: This phenomenon can undermine traceability and cost control before management even realizes the scale of the issue.
2. Model Governance: This area is still seen as ancillary rather than essential, leading to potential oversight.
3. The AI Act Compliance: Organizations may not suffer from outright non-compliance due to negligence, but from unintended breaches caused by inadequate preparation.

As automation becomes more entrenched in decision-making, the importance of governance around these automated decisions cannot be underestimated. The demands of accountability, traceability, and oversight will require organizations to rethink their approaches, ensuring that AI systems operate with defined ownership and control.

Conclusion

The conversation about AI governance is not just timely; it is imperative for businesses aiming to harness the full potential of AI responsibly. By addressing the gaps identified in Sopra Steria Next's report and implementing a clear governance framework, organizations can not only mitigate risks but also enhance their decision-making processes, thereby driving value responsibly in the AI-driven era.

For further insights and details, refer to the full report by Sopra Steria Next.