The AI Alignment Discussion Belongs in Every Company

When 27-year-old AI researcher Jacob Coxon announced his resignation from Anthropic on 8 September, he accused Anthropic and OpenAI of “racing straight to self-improving superintelligence and gambling with our lives.” The warning was stark. So was the response: The Associated Press reported that his posts reached more than 100 million people overnight.

The story quickly moved beyond a single employee’s decision. Current Anthropic researchers publicly echoed parts of Coxon’s concern, lawmakers renewed calls for safeguards and news organizations treated the resignation as evidence of a widening debate inside the AI industry.

Four days later, Anthropic CEO Dario Amodei published "We Must Pace the Frontier," arguing that developers should "slow the pace at which we improve the capabilities of AI models."

One word runs through Amodei's letter and the broader debate: alignment.

In frontier AI research, alignment generally means keeping increasingly capable systems consistent with human intentions and values.

Poor alignment or misalignment creates the risk that a future system becomes capable enough to evade human control, accelerate its own development or cause catastrophic harm. For general models, managing this risk rests with the small number of companies building frontier foundation models, together with the governments that set the rules under which they operate.

While the industry, political and societal debate about that headline-dominating risk is essential, it's not the only critically important conversation.

As businesses accelerate their adoption of AI and begin deploying autonomous agents, leaders need to just as vehemently discuss alignment within their own organizations.

AI Alignment Must Take Place Within Organizations, Too

Nearly 9 in 10 organizations are using AI for at least one business function and 40% of large businesses are already scaling the use of AI agents, according to McKinsey's 2026 "The State of AI: Global Survey."

As companies accelerate deployments of AI, and as agents enter more workflows, these systems will begin to act more autonomously and create new paradigms. As Darden alumnus and GoFundMe Chief Product and Technology Officer Arnie Katz (MBA ’09) said in a recent LinkedIn post: “Saving 2o minutes is good. Changing what your organization believes is possible is a different order of magnitude. The companies that win won't do the old work cheaper. They'll redraw the frontier.”

On this new AI-enabled operational frontier, the same alignment questions facing frontier AI developers become salient within every organization.

To align AI within an organization, leaders should encourage discussion at both the technical AI level and the broader organizational level. They can look to existing, trusted frameworks for coordination. Values, norms, culture and ethics are some of the many alignment mechanisms that coordinate corporate activities and can also be helpful in ensuring AI alignment within organizations.

Three Considerations to Guide Organizational AI Alignment

1. Stakeholder theory offers a useful starting point.

The theory has long been central to Darden’s research and teaching. Professor Ed Freeman is widely regarded as a father of stakeholder theory, while Professor Robert Phillips, one of our co-authors and a Darden Fellow in AI Research, recently advanced ideas on the application of stakeholder theory in organizations with his essay "Stakeholder Theory and a Principle of Fairness." The theory asks managers to consider the connected interests of customers, suppliers, employees, investors, communities and others with a stake in the enterprise. It does not dictate one answer to every workforce decision. It requires leaders to make trade-offs visible and explain whose interests are advanced, whose are burdened and why.

Stakeholder theory posits that shareholder interests like productivity and profit are not at odds with the interests of other stakeholders like employees and customers. Rather, they are aligned, and serving all stakeholders best serves the long-term interests of an organization and its shareholders. A stakeholder theory-guided approach to AI alignment could help AI systems better consider and calibrate their outputs to serve this broader view of successful goal attainment.

2. The discussion around AI alignment belongs to everyone in the organization.

In an ongoing survey of business leaders conducted by Darden’s LaCross Institute for Ethical AI in Business, 53% of respondents named the executive team and 50% named the legal team when asked: "Who's in charge of resolving AI ethics problems?" About 22% said their organizations are relying on a dedicated AI committee to manage such issues.

Ethics cannot be limited to the C-Suite or lawyers when deployment decisions could dramatically impact the entire enterprise. Questions of AI alignment need to be considered by everyone in an organization. Organizations should encourage more buy-in and collaboration from the entire enterprise to effectively align AI.

3. Guardrails and purpose are important to define for every AI deployment.

Research from Darden Professor Roshni Raveendhran, our co-author and Academic Director at the LaCross Institute, has made a compelling case for why organizations should consider the psychological impact of AI in these conversations.

While this list is not exhaustive, helpful guidelines for alignment that have a human-centered focus include:

• Be explicit about what AI is for and whom it serves:

  This allows leaders to define the AI's purpose clearly. Purpose-driven AI is developed for a specified end, serving identified stakeholders, with explicit clarity about trade-offs. It allows organizations to exercise judgment at all levels, from design to deployment to management of AI.

• Design incentive systems for AI and humans that are aligned with organizational values:

  What’s incentivized within an organization is what sticks. It’s as true for AI agents as it is for employees. Teams with no incentive to align with organizational values could, in turn, misalign their AI tools. Values need explicit attention at every stage, including hiring, promotion, team structure and product design and value alignment should not be a siloed responsibility within certain teams.

• Make it psychologically safe and worthwhile for humans to raise concerns and overrule AI:

  Employees should be allowed to experiment with AI tools, speak up about issues they encounter and believe that action will be taken when they speak up about a concern. This can be especially hard with AI as employees may hesitate to raise issues that sound unfounded, and leaders may be uncertain what to do about them. Psychological safety should be front and center, alongside a structure for acting on issues once they are raised.

• Pay attention to how AI is changing people:

  AI is already having profound effects on the behavior and psychology of humans who use it. For example,

  people hold different expectations of AI than of humans, and engaging with AI lowers fear of negative evaluation

  . AI can make it easier for employees to seek help related to hard questions from a machine instead of from a colleague. However, this could reduce opportunities for humans to learn from each other. By monitoring the psychological impact of AI, leaders can take action to promote AI in ways that maximize benefits and minimize harm.

While the Frontier Debate Is Waged, Don't Wait to Manage the Workplace

AI’s impact on organizations is already visible in changing tasks, roles and staffing plans, but its direction is not predetermined. The time for organizational AI alignment has not passed, nor should it wait for the future.

A small group of researchers and executives may determine how quickly superintelligence advances. The broader social consequences of AI will be shaped by millions of management decisions that leaders at every organization can make now. That risk is less open to debate, more widely distributed and, crucially, within the control of individual leaders.