In this era of human-guided AI, AI can play an important role in helping leaders prevent moral drift and preserve what matters most.
Sorapop Udamsri, Canva
Last month, Mrinank Sharma, the head of AI Safety at Anthropic, penned a resignation letter read ‘round the world. “The world is in peril,” he warned, “And not just from AI, or bioweapons, but from a whole series of interconnected crises unfolding in this very moment.” In particular, he spoke to how difficult it is for values to govern decision-making and actions. “I’ve seen this within myself, within the organization, where we constantly face pressures to set aside what matters most, and throughout the broader society, too.” As if to underscore the broader societal issues, Anthropic was recently blacklisted by the U.S. Government for not conceding to AI use for mass surveillance and fully autonomous weapons. Anthropic CEO, Dario Amodei, said that the company would rather not work with the Pentagon than agree to AI uses that may “undermine, rather than defend, democratic values.”
While it’s true, as Sharma quotes in his letter, that “Not-knowing is most intimate,” it’s also true that in the everyday world of egos and organizations, of values and profits, trade-offs and social harms, leaders must continually decide and distill a course of action from the cloud of possibility. Leadership that cares for the bigger picture and partners with the knowingness of AI is best positioned to implement strategies and decision rules that face the peril and honor what matters most.
The challenge for values to govern decision making is real, both in ourselves as individuals and in organizations. Years of research into cognitive bias, behavioral ethics and motivated blindness have shown that even when people believe in a set of values, their attention can be constrained by focus on profits, targets and rivalries that justify decisions to the contrary. Organizational decision making can be even more constrained by cultural patterns of “bottom-line” thinking, hitting key performance indicators (KPIs), and beating the competition. Even when individuals might sense an ethical boundary getting crossed or unintended consequences, they may not want to risk going against the reward and incentive systems of their organizations (much less countries), from which they draw belonging, respect and opportunities for achievement and advancement.
So, the related challenge for leaders is to be able to sense, make sense and act out of a wider net of concerns than their ego’s self-interest. That challenge is generally met through the maturing process, by which we grow through stages of development and worldviews, well mapped in Ken Wilber’s Integral Theory. From the start of ego formation around the age of two, that process unfolds into a sense of self that keeps expanding to include relationships with individuals and groups, beliefs, abstract concepts, rationality and logic, empathy, strategic and systems thinking, a sense of the flex and flow of life and ultimately what Wilber calls causal or nondual consciousness.
This journey from an ego that feels apart from others to being wholly connected and a part of the wholeness of life may be well mapped, but it is not equally traveled. Most leaders today operate at about a halfway point in this journey from a rational worldview, where their decisions are based in facts, truth, and calculations of what serves their interest. In recent years, we’ve seen a regression from even this stage, paralleling the erosion of democracy toward authoritarianism, where “facts” are made up and truth is whatever the leader says it is. But generally, most leaders and organizations operate from a rational worldview, oriented toward bottom-line, KPI-driven thinking.
Unfortunately, the rational worldview (much less any retreat from it) is not up to the complexity of today’s interconnected crises. Indeed, it has produced these crises and AI amplification of such thinking makes them worse. That’s why leadership development training for the past 30 years has aimed to propel leaders into more strategic, systems thinking, more empathetic connection with others, and more courage and resilience—what Dotlich, Cairo and Rhinesmith summed up as Head, Heart and Guts. While I prefer the term “hara” to “guts,” either term underscores that the journey from rationality to nondual consciousness goes beyond head and into the mind and body as an integrated whole. It involves physical training, accelerated by meditation and contemplative practices, amplified by connection with nature and mind-expanding experiences, and illuminated by Zen seeing through the ego to a radically boundless sense of Self. When leaders function from this integrated sense of wholeness, which is the essence of Zen Leadership, they sincerely take care of the bigger picture—for example, consider unintended consequences or social harms related to their decisions—not so they can be known as “good” or altruistic leaders, but because that bigger picture is part of how they experience themselves. They’re taking care of their whole selfless Self.
When the knowingness of AI is used from this more holistic consciousness something extraordinary becomes possible: namely, holistic leadership strategies and decision guidelines that can face the peril, reduce risk, and avert crises. Human leadership in this era of human-guided AI plays the crucial role, as McKinsey puts it, of setting the “why” and “what” of work, that is, the right context and aspiration, inspiring ownership, and demonstrating judgment by aligning choices with clear values. When that leadership functions from more holistic consciousness, AI becomes a more holistic thought partner. It can show, for example, how to use AI itself to avoid the moral drift pointed to in Sharma’s letter. For the leader committed to caring for the bigger picture—creating enduring value for society and mitigating harms—here are five strategies and decision rules that point the way to holistic AI partnering.*
1. Design for value creation and harm mitigation at the same time
Recognizing that all good things can go too far or have unintended consequences, anticipate and design guardrails around likely harms. Microsoft’s Satya Nadella advocates addressing potential harms simultaneously with value creation—a principle put into practice in the development of Microsoft’s 365 CoPilot. AI itself can be a partner in identifying potential harms and suggesting ways to manage them.
Decision Rule: Only launch an offering once you can name the top benefits you want to create as well as the top unintended consequences and how you’ll monitor for both.
2. Engineer AI safety in, don’t bolt it on at the end
Akin to the lessons from the quality and safety movements of the past 50 years, the responsible use of AI cannot be properly governed as an afterthought. Governance itself can be framed and engineered as an essential element of building trust or a story of enduring value, rather than as an anti-competitive burden. Frameworks to support this process, for example, NIST’s AI Risk Management Framework, lay out ways to govern, map, measure and manage risk at every stage in the lifecycle of an AI offering.
Decision Rule: Don’t frame governance as an unwanted add-on, rather embrace it and build it in from the beginning to safeguard the value you want to create in the bigger picture, longer term.
3. Use AI to avert moral drift
AI is not limited by the baud rate of human attention, nor is it trying to make its mark in the organization. As such, it can avoid much of the bias and blindness that leads to moral drift. Steeped in the story of value that the organization is trying to create, AI can provide decision guidance or a real-time telemetry layer of monitoring that ensures that value is not being overly compromised by purely bottom-line thinking or a specific KPI.
Decision Rule: Let AI speak or function in service of the overall value the organization is trying to create.
4. Evaluate AI offerings in terms of human outcomes
From a rational stage of development, AI offerings might be evaluated in terms of technical benchmarks, productivity enhancements, or bottom-line earnings. But from a more holistic perspective, how those offerings impact human lives and society at large is even more important. Amodei’s courageous stance is a shining example of this perspective.
Decision Rule: When evaluating a new AI offering, prioritize how people will use it to reach their goals and what will be the impact of such use on them and those around them.
5. Expand from designing an offering to designing social systems
One of the lessons learned from social media is that the value or harm of an offering cannot be divorced from how people actually use it. As such, systems-thinking leaders like Nadella advocate for a social systems engineering approach to AI offerings. Like simulating a game, such systems engineering can model the AI offering in use, experimenting with different agents, permissions, memory and guardrails for safety.
Decision Rule: Treat the real-world implementation of an AI offering as a systems engineering matter that calls for modeling and safeguards for use.
The uniquely human role of leadership in the age of human-guided AI is to set context and aspirations, enroll others and align choices with clear values. But as Sharma points out, it’s easy for leadership decisions to drift from values and what matters most. Kudos to Anthropic for both highlighting the issue and nevertheless standing firm in their values at considerable cost. Such courage is formative in this embryonic era of AI, as is the human consciousness with which AI is used. As this sampling of decision rules shows, AI is perfectly capable of functioning from a more holistic consciousness that cares for the bigger picture if it is so primed. Moreover, it can assist human leaders in avoiding unintended consequences and moral drift. It would be wonderful if every leader made decisions from nondual consciousness. But short of that, any leader who is sincere about creating enduring value in the world can be assisted by AI in facing the peril and better serving the whole.
