Co-authored by R “Ray” Wang, Vala Afshar, and Dr. David Bray

Executive Summary: When the co-designer of the Internet’s foundational protocols tells you we’ve encountered “a new life form” and when seasoned Senior National Intelligence Service Executive suggests we stop calling it Artificial Intelligence and start calling it “Alien Interactions” you’re not hearing isolated observations. You’re witnessing the convergence of critical insights that boards and CEOs must internalize: the shift from deterministic code to probabilistic agents executing at machine speed, the erosion of critical thinking when humans optimize purely for efficiency, and the urgent need to define accountability when autonomous systems operate without human oversight. In DisrupTV Episode 439, a framework emerged for leading through this transition: both Dr. Vint Cerf and Dr. David Bray believe successful organizations must move from “Human in the Loop” to “AI in the Group” while doubling down on what Cheryl Strauss Einhorn calls “The Human Edge,” the irreplaceable capacity for judgment, values-driven decision-making, and consequence awareness that no algorithm possesses.

Editor’s Note: This article represents the second OODA Loop collaboration among R “Ray” Wang, CEO of Constellation Research; Vala Afshar, Chief Digital Evangelist at Salesforce; and Dr. David Bray, Chair of the Accelerator and Distinguished Fellow at the Stimson Center.

From Deterministic Protocols to Probabilistic Agents: The Intent Problem

When Vint Cerf co-designed the TCP/IP protocols that became the foundation of the Internet, intent was straightforward. The network ran itself, but humans made the decisions. As Cerf explained, “It’s true that we had people in the loop, but the networking part was all automated. That was the whole point. So the network ran itself.” Today, we are entering what Cerf calls “the agentic world of AI,” where agents run autonomously and interact with each other at machine speed.

The shift from deterministic code to probabilistic AI models introduces a fundamental challenge: non-determinism at scale. Cerf emphasized a critical concern: “The big problem I worry about is agents talking to each other using natural language because humans use natural language and we misunderstand each other. We don’t need agents to misunderstand each other and execute at the speed of light compared to human speed.”

For boards and CEOs, this introduces a new category of operational risk. When agents misunderstand their tasks, the resulting errors propagate instantly across systems, supply chains, and customer interactions. Cerf stressed the necessity of unforgeable audit trails: “We have to know who is responsible for the agent’s actions, who is accountable or what entity is accountable. How do we identify the agents so that if something goes wrong, we know where to turn? How do we make sure that the audit trail is unforgeable and acceptable in a court of law as evidence?”

Dr. David Bray connected this to organizational readiness, noting that most enterprises haven’t established the governance structures to answer these questions. “We haven’t fully figured out almost like the right of way when it comes to humans and agents when it comes to doing tasks in the workplace,” Bray observed. He offered a vivid historical analogy: “Think about the photos of Chicago or New York in 1910, where they literally had cars on the same road as horses, on the same roads as trolleys, on the same road as people. We’re sort of in a repeat of 1910. We haven’t invented stoplights yet. We haven’t even figured out stop signs or right of way or sidewalks.”

Ray Wang pressed on the practical implications: when an agent messes up, who do you sue? Bray’s response reframed the question: “I want to actually add the phrase AI in the group. It’s not that this is artificial distinction of AI or human. It’s actually they’re a collective. If an organization does something bad, right now we sue the organization, even if it was an individual in the organization. We’re going to have to figure out whose flag is this agent flying when it is doing something? Whose organization is it flying the flag of? That organization should assume accountability if it goes off the rail.”

Cerf reinforced this point: “That leads to the notion about registration of agents or registrations of entities that create and support agents.”

The 1910 Moment Requires New Management Techniques

Bray emphasized that we’re in a transitional period where traditional management approaches don’t apply. “Let’s face it. When do humans always do what they’re told? We have this illusion that just because I told a human to do something, they’re going to go forth and do it and they’re going to understand it. I think it really is how quickly do you sense when something is being done that was not intended? It could be not intended because it was a misunderstanding. It could be because someone’s abusing it.”

This connects to what Bray calls “good governance,” which he defines as “how we avoid anarchy. We’ve got to have anarchy protection, not just for humans, but for agents, but allow for the fact that humans and agents will not always do what they’re told.”

The practical implication: organizations need real-time sensing mechanisms to detect when agents deviate from intended behavior, whether through misunderstanding or misuse. This requires new monitoring systems, new escalation protocols, and new accountability frameworks that don’t exist in most enterprises today.

Vala Afshar highlighted a reality already operational at Salesforce: “In the last 12 months, nearly 4 million conversations on our service and support site has been completed to satisfaction of a customer with zero human in the loop.” The scale of autonomous operations is accelerating. Afshar cited Waymo’s trajectory: “600, 700 Waymos surpassed 45,000 Lyft drivers in California, in San Francisco specifically, last year.” The numbers are staggering: 200 million autonomous miles, half a billion rides weekly, adding a million additional miles per week.

But autonomous scale without human oversight creates what Cerf called “the recourse problem.” Bray shared a cautionary example: “I’ve had experiences with other calls with other entities where you keep on saying, speak to a human, speak to a human, speak to a human, and they never let you out of the loop. There needs to be a way to, if you’re not getting what you want, to first try and signal it to the agent, and if the agent still can’t provide, escalate.”

Cerf introduced a term that should become standard in board governance discussions: “When we’re dealing with these non-deterministic systems, establishing a mode for recourse in a variety of circumstances might be a very high benefit and maybe even a necessity.”

Cheryl Strauss Einhorn framed the central challenge: “So much of our concern has been that this is a tool that asks us to share our cognitive load, which means to give it some of our actual thinking. What’s so interesting is it knows nothing about us and it doesn’t care about consequences, but we do.”

The risk for modern organizations is what Einhorn calls “cognitive atrophy.” “If you find yourself optimizing purely for efficiency and you start asking AI questions without first asking yourself, without testing back and pressing against the machine, you’re actually turning off your thinking. And over time, it’s going to be harder to decide that you’re going to sit in the discomfort of thinking to actually figure out your own mind.”

Bray connected this to a generational challenge emerging in the workforce. “We’re starting to see Gen Z at graduation. There is an AI backlash that is present in Gen Z, especially because they’re concerned about the future of work. What we’re seeing with Gen Z is in school, when teachers gave them hard problems they were supposed to work on, they turned to the AI and they didn’t wrestle with them. When they took their midterms and their finals they didn’t do well. They are graduating not having had the experience of knowing what good looks like.”

Cerf offered historical perspective: “I seem to recall it was either Plato or Archimedes or somebody who said that writing was a bad thing because people would stop remembering anything.” The challenge isn’t new, but the speed and scale are unprecedented.

The Human Edge: Why Cognitive Atrophy Is a Strategic Risk

Bray introduced a powerful reframing of AI’s role in organizations: agents can serve as organizational mirrors that reveal management failures humans are reluctant to acknowledge. “If a manager gives an assignment to an employee and the employee doesn’t complete it, the usual assumption is it was the fault of the employee. Well, it might actually be the manager didn’t explain the assignment correctly. It might be that the manager gave it to the wrong employee or they didn’t actually have the time to actually upskill them.”

He continued: “If you had an AI in the background observing that manager employee interaction, the AI might come out and say, dear manager, are you sure you explained this appropriately to the employee or did they have the right skills or is this the right person? And maybe they’re overburdened. That’s where it really is, AI in the group. They’ll be able to do autonomous things just like how we’re able to do autonomous things, but collectively how can we be better together.”

This represents a fundamental shift: AI not as replacement for human judgment, but as a feedback mechanism that surfaces organizational dysfunction humans are incentivized to ignore. The strength of AI, as Bray noted, is that “it doesn’t have the same ego as a human. It might actually help identify when a manager gives an assignment to an employee and the employee doesn’t complete it.”

Einhorn’s framework for maintaining the human edge centers on what she calls “special sauce,” the unique combination of values, priorities, and judgment criteria that distinguish one decision-maker from another. “Each of us have a special sauce. It is the way we make decisions. Most of us don’t really have awareness of what that is, but it tends to combine some of our core values.”

The critical insight: if you don’t communicate your special sauce to AI, it will give you generic answers optimized for popularity, not for your specific strategic context. Einhorn illustrated: “If you’re looking to change jobs, it doesn’t know when you’re asking it about jobs, are you doing it because you’re exhausted or are you doing it because you’re looking for something that’s more meaningful? Either answer, only one of these things is going to be right for you.”

Cerf raised a concern that should alarm every board: the potential loss of auditable decision-making records as software evolves. “There are two aspects to the question. The first one is just losing access to digital information is a serious issue because a lot of it requires software to correctly render or interpret or execute. None of our digital technology has that longevity. The critical thing that’s different is that in order to interpret preserved bits, you need software to understand them.”

Bray introduced a more immediate threat: the proliferation of synthetically produced information that undermines decision quality. “There are some stats that say by 2030, more than 40% of the information on the planet will have been synthetically produced by an AI. Not saying it’s all bad, but that’s going to create massive questions for CEOs and boards, which is if I’m using that information to make a decision, I might actually be making it on wrong information. There’s actually evidence that in some of the more recent elections, some of the polling relied on online polling that may have actually been completed by a bot.”

Cerf’s response captured the paradox: “I don’t want to overstate this, but it feels like we’ve encountered a new life form and we’re trying to figure out how it thinks.”

Bray offered a reframing that boards should adopt: “When we did work with the People-Centered Internet, even back in 2017, 2018, I would sometimes say maybe instead of calling it artificial intelligence, we should call it alien interactions. Because that way we will not try to anthropomorphize the machine.”

Bray introduced a framework for navigating information environments where synthetic data proliferates: triangulation. “A healthy response for societies in these times is increasingly don’t trust the first thing you see unless you triangulate it. We should be teaching kids in school, how do you triangulate from multiple sources and try to assess what is real and what’s not? That’s what the CIA does.”

He elaborated with an intelligence analogy: “Everything debuts with a walk-in. That walk-in might be trying to fool you for intentional deception. That walk-in might be genuine, but they’ve been told a lie to sell to you. Or they may actually be real. How do you sense as that walk-in comes in? Are they really trying to defect? Or are they really just trying to spin you a lie? It’s the same tradecraft.”

Cerf suggested that AI itself could become part of the solution: “Critical thinking has become a skill which is even more needed than ever. Some people go to multiple sources, multiple agents, in order to get different answers and compare them. I can imagine another agent that helps you do the comparison. The seeds of solution to the problems we’re posing may lie in the system that we’re developing.”

Strategic Recommendations for Boards and CEOs

Organizations need to recognize the new reality: AI proficiency is now a hiring requirement at all levels. But proficiency without judgment is a liability. Prioritize candidates who demonstrate both technical fluency and the critical thinking skills to challenge AI outputs.

Given this converging risks, corporate Boards and CEOs should at a minimum:

1. Establish Agent Registration and Accountability Frameworks

Every agent deployed by your organization should be registered to a specific accountable entity with cryptographically validated audit trails. Define whose flag each agent is flying and ensure that entity assumes accountability when agents go off the rails.

2. Build Recourse Mechanisms Into Every Autonomous System

As you move toward autonomous operations, ensure there is always a clear, immediate path for human escalation. Do not trap customers, employees, or partners in agentic loops with no exit.

3. Combat Cognitive Atrophy Through Structured Critical Thinking

Teach your teams to ask AI for disconfirming data and to reframe prompts to avoid inherent human bias. Make triangulation from multiple sources standard practice, not an exception.

4. Communicate Your Special Sauce to AI Systems

Before deploying AI for high-stakes decisions, invest time investigating your organization’s core values, priorities, and decision-making criteria. If you don’t tell AI why you are solving a problem and what constraints matter most, it will give you generic answers.

5. Deploy AI as Organizational Mirror

Use AI to surface management failures and organizational dysfunction that humans are incentivized to ignore. AI can identify when managers fail to explain assignments clearly, assign work to the wrong people, or overburden employees.

The transition to an agentic economy is not just a technical upgrade. It is a cultural and structural revolution. As Cerf observed, “It feels like we’ve encountered a new life form and we’re trying to figure out how it thinks.” Organizations that treat this as a software deployment will fail. Organizations that recognize this as a fundamental reimagining of work, accountability, and human-machine collaboration will shape the future.

Bray’s reframing captures the mindset shift required: “AI in the group. It’s not that this is artificial distinction of AI or human. It’s actually they’re a collective.” The question is not whether humans or agents will do the work. The question is how to structure that collective so that human judgment remains the critical path, so that accountability is clear when things go wrong, and so that the special sauce of human values guides autonomous execution.

About the authors:

R “Ray” Wang is the CEO of Silicon Valley-based Constellation Research Inc. He co-hosts DisrupTV, a weekly enterprise tech and leadership webcast that averages 50,000 views per episode and blogs at www.raywang.org. His ground-breaking best-selling book on digital transformation, Disrupting Digital Business, was published by Harvard Business Review Press in 2015. Ray’s new book about Digital Giants and the future of business, titled, Everybody Wants to Rule The World was released in July 2021. Wang is well-quoted and frequently interviewed by media outlets such as the Wall Street Journal, Fox Business, CNBC, Yahoo Finance, Cheddar, and Bloomberg.

Vala Afshar is currently chief digital evangelist at Salesforce, where he advises customers on the rise of agentic AI and the future of advanced technologies. Previously, he has served as VP of engineering, chief customer officer and CMO. Recognized as a leading industry thought leader, Vala boasts over a million followers on X and LinkedIn, holds multiple US patents, writes a weekly column for ZDNET and has hosted the popular enterprise podcast DisrupTV for over a decade. He is co-author (with Henry King) of “Autonomous: Why the fittest businesses embrace AI-first strategies and digital labor” (Wiley, 2025) and “Boundless: A new mindset for unlimited business success” (Wiley, 2023).

Dr. David A. Bray is a Distinguished Fellow and Chair of the Accelerator with the Alfred Lee Loomis Innovation Council at the non-partisan Henry L. Stimson Center. He is also a CEO and transformation leader for different “under the radar” tech and data ventures seeking to get started in novel situations. He is Principal at LeadDoAdapt Ventures, Inc. and has served in a variety of leadership roles in turbulent environments.  He previously served as a non-partisan Senior National Intelligence Service Executive, as Chief Information Officer of the Federal Communications Commission, and IT Chief for the Bioterrorism Preparedness and Response Program. Business Insider named him one of the top “24 Americans Changing the World” and he has received both the Joint Civilian Service Commendation Award and the National Intelligence Exceptional Achievement Medal. David accepted a leadership role in December 2019 to direct the successful bipartisan Commission on the Geopolitical Impacts of New Technologies and Data that included Senator Mark Warner, Senator Rob Portman, Rep. Suzan DelBene, and Rep. Michael McCaul. From 2017 to the start of 2020, David also served as Executive Director for the People-Centered Internet coalition Chaired by Internet co-originator Vint Cerf. Business Insider named him one of the top “24 Americans Who Are Changing the World” and he was named a Young Global Leader by the World Economic Forum. For twelve different startups, he has served as President, CEO, Chief Strategy Officer, and Strategic Advisor roles. The U.S. Congress invited him to serve as an expert witness on AI in September 2025.