Business-centric artificial intelligence: A perspective on AI governance
Business-centric artificial intelligence: A perspective on AI governance
By Atos staff
From the well-acclaimed achievements of the artificial neural networks in breaking former compute barriers to the ubiquitous presence of virtual assistants in contemporary households, artificial intelligence (AI) research has long since transcended the lab’s confinements. Enlightened applications of machine consciousness can contribute significantly to the cause of human civilization. AI can deliver a multitude of benefits for businesses and societies by aiding decisions that are more inclusive, informed and fair. But realizing the true potential of the technology hinges on maintaining a strict vigil over how AI-based products continue to be developed, consumed and governed, curbing possibilities of their amoral use.
Over the last decade, synthetic intelligence has managed to eliminate the complexities of choices, streamlining business administration to unlock extraordinary precision, speed and convenience. As the business adoption of artificial intelligence inflates, it is increasingly portrayed as a tool for maximizing returns. The trend is echoed by McKinsey’s Global AI Survey that reports more than a 25% year-over-year increase in AI deployment. While 58% of the respondents acknowledge having embedded AI functions in their business processes, 44% have directly attributed the measurable reduction in operating cost to the technology.
However, a study published by Deloitte in the World Government Summit 2019 studied the opinion of business executives knowledgeable in AI to conclude that 32% of the respondents ranked ethical issues as one of the top three risks involving the technology. Moreover, Cornell University found 82% of the Americans believe that AI calls for meticulous governance. As institutions choose to drive revenue and prudence by leveraging AI, it is evident that as fundamentally social entities, they will be under mounting pressure to improve transparency and accountability around AI use cases by dismantling the inscrutable black box of algorithms wielding the delegated authority to assume life-changing calls.
In an interview with distinguished thought leaders, two prominent voices in the artificial intelligence domain examine how companies can navigate their AI strategies to balance core business interests against pertinent social concerns around prejudice, cultural sensitivity, data security and privacy. Can the quadrants of AI governance be drawn by incorporating the lessons learned while setting similar ground rules for application, cloud and data management regimes earlier?
Reimagining business innovation by factoring governance modalities
With the onset of Industry 4.0, innovation has become the motive force of economic progress and AI as a means for scaling digital transformation challenges. But the governance angle makes it imperative that AI-centric innovation by enterprises should no longer be unidimensional, resting solely on technology considerations. For Tatianna Flores, Head of the AI Lab, Atos North America, the question of innovation using AI is about governance, as it is about technology. “It has been the AI Lab’s mission to act as a moderator for its customers that conceives innovation solely from either technology or governance perspectives, affecting time-to-value. Dispelling pessimism, we encourage businesses to be creative and imaginative about the AI use cases while investigating the practicality and feasibility of such ideas and addressing concerns that might arise from their applications,” she says.
Further citing the example of how the discourse on big data has shifted from volume to the actual quality and integrity of data owing to the evolution of data governance frameworks, Bozhidar Hristov, Senior Analyst, Technology Business Research, Inc. (TBR) feels that better AI governance standards may depict organizational maturity and in turn fuel innovation in highly regulated sectors like financial services and healthcare. “Sound governance frameworks act as a yardstick and allows businesses to innovate without overstepping the limits of regulatory and liability constraints,” agrees Tatianna Flores.
Forging new service roles, partnerships and a culture conducive to AI governance
Traditionally, AI has been perceived to be a subset of the IT realm. But as AI models evolve, the interpretability of their outputs are relying more on the understanding of the business side regarding various market events, situations and anomalies. This progressive convergence, says Tatianna Flores “argues for multi-departmental governance of AI spanning not only IT developers, engineers, system administrators and business subject matter experts, but also legal and audit policymakers who can look at the bigger picture through transparency and ethical prism to ensure orderability and viability of solutions.”
According to her, this is where service integrators with multidimensional expertise will be vital in resolving complexities associated with AI models by ensuring that all the validations and regulatory considerations are taken care of up-front, and a clear implementation roadmap is laid for businesses within the boundaries of governance realities.
Hristov agrees, expanding this notion on expertise to future planning. “So in the future, professionals with hybrid-skill sets incorporating technology, business and policy aspects of AI will be pivotal in managing AI governance positions. We may witness new roles like the Chief Ethics Officer or the expansion of existing ones like that of the Chief Compliance Officer.”
However, “trust in AI cannot be built overnight, and it will require a balanced ecosystem of interconnected dependencies and accountabilities, fueling the required cultural shifts and intents towards better AI governance.” says Bozhidar Hristov. New generation AI business leaders will have to drive stakeholder commitments and customer loyalties by capitalizing on change management and their cultural inclination toward governance more than their posturing on digital readiness and technology ascendency. Establishing overarching centers of excellence (CoEs) focused on AI and employing Governance Ambassadors may go a long way in permeating that trust and steering the cultural repositioning necessary for the seamless adoption of the technology.
Approaching governance as a function of AI being ethical, transparent and explainable
As the AI models become more complex and use comprehensive datasets, finding broader applications into areas closely associated with citizens’ lives, there is a need to widen the institutional base of AI governance, making it everyone’s responsibility. The conformity of an AI product to sound governance principles should be measured tri-laterally, the first of which is whether it is ethical. Does its operation reflect the same values and ideals enshrined in the vision and mission statement of the enterprise that owns it? Tatianna Flores says, “AI models used by a recruiter for screening candidates and determining who gets interviewed and hired must operate on the same principles that the organization follows while operating manually. So, in case of equal opportunity employers, the recruitment model must set uniform evaluation standards for all the applicants and essentially avoid all discriminations.”
With the intrusive usage of data by AI models, Bozhidar Hristov feels that being transparent regarding their source and end states is more important today than ever; what are the frameworks employed in its handling? What are the checks and balances adhered to? Are the third-party auditors involved? Can the fairness of the decisions taken be scrutinized? Is it possible to verify if the model is discriminatory? These are some of the questions that businesses indulged in the transparent usage of AI should be able to relate and answer.
According to Tatianna Flores, the third and final benchmark is whether the AI use cases are explainable. Organizations should maintain a trail of breadcrumbs to track the validations that the AI model has used in arriving at decisions that may have severe implications for the people’s futures. Beyond their technical aspects, such accounts must be comprehensible to the stakeholders and customers across the board for the AI model and its applicability to be trustworthy.
Prepare for pervasive AI environments
AI as a technology is decades old and has been in use in areas like financial services. However, it is the customization of business applications and the commercialization of AI models that have spurred AI governance to scale. That AI-based technologies are being embraced by businesses at an unprecedented pace is reported by a Gartner study. If found that while in 2020, companies have five AI models on average, it can be well over 30 within a few years, owing to easy access and cost-effectiveness.
To continue operating AI models and amplifying their benefits while dwelling within the social and regulatory parameters, the companies need to be practical in framing a robust and functional AI governance framework that is responsive to changes. Such mechanisms should be able to handle queries to all stakeholders’ satisfaction, both internal and external. For instance, in the post-COVID-19 world, with the augmenting role of AI in institutional hiring, the AI governance framework should be able to appropriately embed the message for dispelling the recruits’ employment-related anxieties associated with AI-centric labors as part of their onboarding process.
Not only customers but also AI governance bears on the service integrators, specific moral and ethical obligations, transcending business interests. Tatianna Flores says, “many service integrators are putting internal ethics and compliance groups in place to vet the requests from their customers. It is critical to examine and ascertain whether service requests resonate with the code of AI governance and principles set in place by the service integrators for themselves. If requests are found to be too risky, socially detrimental, or against the organization’s fundamental beliefs, they are liable to be dropped.”
With AI use cases proliferating with unmatched haste in all facets of business and community existence, It is paramount for organizations of all sizes and at every stage of digital journeys to focus on building the AI governance charters. The success of AI dwells on a circle of accountability, encompassing both the internal and external environments of an organization. Right AI governance should empower all stakeholders to focus on achieving the key priorities markers while engaging with the technology with confidence and in totality.