In the early days of the American railroad system, expansion was chaotic.
Different regions built tracks with different gauges. Safety standards were inconsistent. Interoperability was limited. Trains could not move seamlessly across networks. Accidents were frequent. Capital was fragmented. Growth was real — but unstable.
Only after standardized gauges, coordinated signaling systems, and formal safety rules were adopted did railroads become the backbone of American economic expansion. Standards did not slow progress — they unlocked it.
We are at a similar moment with agentic AI. For the past decade, AI functioned primarily as a tool — predicting, summarizing, optimizing. Now we are entering a phase where AI systems plan, execute, coordinate, and act with increasing autonomy. This shift from tool to actor is operational — not theoretical.
As the former Head of Firmwide Artificial Intelligence at Morgan Stanley and now teaching AI strategy and governance at Columbia Business School, I have spent years building AI infrastructure at scale and advising organizations navigating this transition.
Since starting my own firm focused on the effective and responsible deployment of AI, I have talked to hundreds of business leaders around the globe, and the pattern I see is consistent: the technology is accelerating. Institutional readiness is not.
If we want agentic AI to drive durable economic growth — rather than disruption and instability — we must build the governance equivalent of standardized gauges and safety systems.
The Structural Risks
The risks of agentic AI are not science fiction. They are structural.
Diffused accountability.When AI recommends, humans decide. When AI acts, responsibility blurs. Without defined ownership, "the model decided" becomes a governance gap. Institutions require clear authority to function.
Institutional lag.Technology evolves exponentially; governance evolves incrementally. Agentic systems operate at machine speed. Our oversight frameworks were built for static software, not autonomous digital actors interacting across systems.
Workforce disruption.Agentic AI does more than automate tasks — it can execute multi-step workflows once performed by skilled professionals. Some roles will be augmented. Others redefined. Some displaced.
Railroads reshaped labor markets as profoundly as transportation. New industries emerged. Others faded. The difference between prosperity and unrest was preparation.
Governments, corporations, and educational institutions must anticipate this shift — not react to it.
Systemic and security risk.Autonomous systems can probe vulnerabilities continuously, adapt misinformation dynamically, and influence markets at speeds beyond human oversight.
And most underestimated — the erosion of human judgment.The risk here is very real. As we hand over more tasks to AI, we risk handing judgment and accountability to machines that don't own the outcomes. This is the pattern no one talks about enough: AI does not seize authority. Leaders cede it — one comfortable delegation at a time.
In a tool paradigm, this tendency is manageable. A recommendation engine gives you a suggestion. You evaluate it against your judgment and experience. You decide.
In an agentic paradigm, it is dangerous. Autonomous systems execute multi-step workflows, interact with other systems, and produce outcomes that may be difficult to reverse by the time a human reviews them. If the leaders responsible for oversight no longer understand the logic chain well enough to challenge it, the "human-in-the-loop" becomes ceremonial. The signature is there. The judgment is not.
This is not a technology failure. It is an institutional one. And it will accelerate as agentic systems grow more capable and more embedded.
The solution is not to slow adoption. It is to build institutional muscle alongside the technology — training leaders not just to use AI, but to maintain the judgment required to govern it. Organizations that fail to do this will retain the appearance of human oversight while hollowing out its substance.
None of these risks are inevitable. But none resolve themselves. Standards built after crisis are always more painful than standards built before it.
Five Actions We Should Take Now
If we want agentic AI to strengthen rather than strain our institutions, deliberate action is required.
1. Mandate Clear Accountability
Organizations deploying agentic systems in high-impact domains should maintain a documented inventory of autonomous systems, assign named executive ownership, and establish board-level oversight. Accountability must be visible and enforceable.
2. Implement Agentic Stress Testing
Before scaling deployment in finance, healthcare, infrastructure, or public services, organizations should conduct structured scenario simulations, engage independent evaluation where systemic risk exists, and disclose governance practices at a high level. Resilience must precede scale.
3. Launch a Coordinated Workforce Transition Strategy
Governments, corporations, and universities must jointly expand reskilling and lifelong learning programs, incentivize retraining over reactive layoffs, embed AI literacy into core education curricula, and support transition pathways for affected workers. Agentic AI will create new industries. But transitions do not manage themselves. Preparation determines whether disruption becomes opportunity.
4. Develop Interoperability and Governance Standards
Industry consortia and policymakers should accelerate technical standards for multi-agent systems, clear audit trails for autonomous workflows, and cross-industry governance frameworks. Standardized gauges allowed railroads to scale nationally. Shared governance standards will allow agentic AI to scale safely. And for what it's worth, this work is hard and will require constituencies from all sides to engage, debate, and act. This can't happen soon enough.
5. Preserve Human Authority in Critical Decisions
In financial stability, healthcare, legal systems, and national security, final authority must remain clearly human. This requires defined sign-off thresholds, escalation protocols, and — critically — leadership training that builds and maintains the institutional knowledge required for informed oversight. Not just the right to override. The capacity to know when to.
Human-in-the-loop is not friction. It is legitimacy.
A Leadership Choice
Railroads became transformative not because tracks were laid quickly — but because standards made the network reliable, scalable, and trusted. Agentic AI presents the same inflection point. We can allow fragmentation and reactive governance to shape the next decade. Or we can deliberately build accountability, workforce readiness, and technical standards that enable confident scaling.
This is not a call to slow innovation. It is a call to strengthen it. Stable systems attract capital. Trusted systems endure. Prepared societies prosper.
Agentic AI will define economic competitiveness in the coming decade. But competitiveness without preparation is fragile. The future will not be determined solely by model capability. It will be determined by whether leaders choose to lay the right track.
We do not need alarmism. We need foresight, and we need to act now.
I welcome you to join me in this call to action.
Our future depends on it.