This keynote explores how to prepare for a future where artificial intelligence becomes abundantly available and increasingly capable.

IQ Test Results - September 2024: In late-2024, the smartest AI models were scoring around 80-90 IQ, placing them solidly in the average human range with significant limitations.

AIME25 Benchmark: On advanced mathematics competitions designed for gifted high school students, AI models now achieve perfect or near-perfect scores that exceed most human capability.

GPQA Benchmark: When tested on PhD-level questions across specialized domains, top AI models score 84-95%, matching the performance of actual domain experts with advanced degrees.

Vinod Khosla: Every workflow and software system has been designed around the constraint of limited, expensive human time. AI removes that constraint entirely.

Moving from AI anxiety to AI Agility and beyond requires progressing through five levels: building awareness, developing literacy, practicing agility, creating strategic workflows, and ultimately conducting cross-domain agent teams.

AI Evolution Timeline: AI is progressing through distinct phases at an accelerating pace. We moved from perception AI like speech recognition and medical imaging, through generative AI for content creation, to today's agentic AI that can autonomously execute complex tasks. The next leap to spatial AI with self-driving cars and general robotics is already beginning, fundamentally transforming how AI interacts with the physical world.

OpenAI Road Towards AGI: OpenAI's roadmap shows we're currently at Level 3, where AI agents can autonomously execute tasks and maintain context under supervision. We've already passed chatbots and reasoners. The path forward leads through innovators that generate new ideas and proofs, ultimately reaching organizations where AI can coordinate entire teams and run complex operations at scale.

Agency Economy:We're entering another major economic revolution. Just as agriculture gave way to industrial production and then knowledge work, we're now transitioning to the agency economy where human agency multiplies through agentic AI. This isn't about replacing humans but about amplifying what humans can accomplish when freed from routine constraints.

Andrew Ng:The real breakthroughs in AI over the next few years won't come just from bigger models. They'll come from better workflows. This shifts the competitive advantage from those who have the best technology to those who design the best systems for humans and AI to work together effectively.

Human Skills Framework: Success in the agency economy requires developing skills across four critical dimensions. On one axis, you need both self-determination and human-to-human collaboration abilities. On the other, you must balance uniquely human potential like creativity and emotional intelligence with human-machine collaboration skills like AI agility and computational thinking. The question isn't whether to develop these skills, but how intentionally you're building them in your organization right now.

Overcapacity and Skills Shortage: Organizations face a paradox. AI is creating massive workforce overcapacity, with 40-50% of companies expecting to have too many employees for traditional tasks. Yet simultaneously, most organizations face critical shortages of people who can actually work effectively with AI. The gap between these two realities represents both the challenge and the opportunity of the next few years.

Geoffrey Hinton: If you work as a radiologist, you're like the coyote that's already over the edge of the cliff but hasn't yet looked down. Geoffrey Hinton said this back in 2016, warning that certain professions were already disrupted even if they didn't realize it yet. The lesson isn't about radiologists specifically but about recognizing when your profession has fundamentally changed before it becomes obvious to everyone.

Jevons Paradox: In 2016, Hinton predicted that within five years, deep learning would surpass radiologists at interpreting medical images, suggesting we should stop training new radiologists immediately. This captures Jevons' Paradox, where improving efficiency doesn't reduce demand for a resource but transforms how we use it. The real question isn't whether AI will match human experts, but how quickly we adapt our training and workforce development to this new reality.

Radiology's Last Exam: Board-certified radiologists still lead AI models in diagnostic accuracy at around 84%, but the gap is closing fast. Current AI models like Claude Opus 4.1 and GPT-5 are already performing at levels comparable to radiology trainees, scoring between 23-45% accuracy. This isn't just about one profession, it's a preview of how AI capabilities will rapidly approach and potentially exceed expert human performance across knowledge-intensive fields.

Agentic Workflow Design: The future of competitive advantage lies in agentic workflow design. This is where organizations move beyond using AI as a tool and start architecting entire systems where AI agents autonomously handle complex, multi-step processes while humans focus on the work that truly requires human judgment, creativity, and ethical oversight.

AI-First Companies: Leading organizations across every sector are already declaring themselves AI-first companies. From tech giants like Microsoft, Meta, and Salesforce to financial institutions like Goldman Sachs and JPMorgan Chase, from e-commerce platforms like Shopify to AI infrastructure companies like Anthropic, the message is clear. Being AI-first isn't about the technology you use, it's about fundamentally redesigning how work gets done when intelligence becomes abundant and accessible.

Goldman Sachs CEO: Goldman Sachs CEO David Solomon captured the shift perfectly when he said the last 5% of human work now matters more than ever because the rest has become a commodity. When AI can handle 95% of the routine analytical and drafting work, the premium shifts entirely to the uniquely human elements like judgment, relationship building, ethical reasoning, and strategic insight that AI cannot replicate.

10 Phases to Goldman Sachs S-1 Workflow: Goldman Sachs has broken down their S-1 filing creation process into ten distinct phases, each requiring different levels of human agency. The workflow starts with full human control for mandate and framing, drops to autonomous AI for data assembly, oscillates between delegated AI agency and shared human-AI collaboration for modeling and drafting, then rises back to augmented and full human agency for investor-facing activities and post-mortem learning. This mapping reveals where humans add irreplaceable value versus where AI can operate independently.

Human Agency Scale Framework: The Human Agency Scale provides a framework for deciding which tasks should use which level of human-AI collaboration. At one end, autonomous AI agency handles high-volume, well-structured data tasks with zero human touch. Delegated AI agency lets AI execute within defined boundaries while humans approve at milestones. Shared agency creates true collaboration where human and AI iterate together. Augmented human agency keeps humans in the driver's seat while AI expands their capabilities. Full human agency reserves the highest stakes, trust-critical, and nuance-heavy decisions for humans alone with AI offering only passive support.

Goldman Sachs S-1 Workflow - Human Agency Scale Chart: When you map Goldman's S-1 workflow phases against the Human Agency Scale, you see a clear pattern. The workflow begins at maximum human agency for mandate and framing, drops sharply to autonomous AI for data assembly and risk compliance, rises through shared agency for narrative drafting, then climbs back to full human agency for the final investor-facing phases. This isn't random, it reflects a sophisticated understanding of where human judgment creates irreplaceable value versus where AI efficiency should dominate.

Goldman Sachs S-1 Workflow - Human Agency Distribution: In Goldman's redesigned S-1 workflow, roughly 60% of the work now happens at the autonomous AI agency level, 15% uses delegated AI agency, 10% involves shared agency, 10% requires augmented human agency, and only 5% demands full human agency. This distribution represents a fundamental inversion from traditional workflows where humans touched nearly everything. The organization that can successfully execute this shift gains massive efficiency advantages while maintaining quality and judgment where it matters most.

S-1 Prospectus Creation Evolution: As AI maturity progresses from AI Agility through Agentic Workflow Design to Agent Orchestration, the nature of tasks, talent, and teams fundamentally transforms. In the AI Agility phase, GenAI produces first drafts while humans manually stitch outputs together and bankers learn prompt discipline. In the Agentic Workflow Design phase, agentic workflows handle entire processes like modeling and drafting while roles shift to AI Workflow Designers and Compliance Interpreters who guide the system. In the Agent Orchestration phase, multi-agent systems auto-build and file the entire S-1 while core talent becomes Agency Stewards who oversee ethics and alignment, and teams become distributed orchestration pods managing continuous AI pipelines for high-stakes judgment calls.

Warren Berger with "A More Beautiful Question": Warren Berger's work on inquiry reminds us that the quality of our questions determines the quality of our outcomes. In an era of abundant intelligence, the competitive advantage shifts from finding answers to asking better questions. The organizations that master the art of beautiful questions will be the ones that unlock intelligence-driven opportunities others can't even see yet.

Lacking Abundant-Intelligence Use Cases: Most organizations are still focused on incremental AI improvements within existing workflows, the green box on the left. They're working on core ideas that were already feasible before abundant intelligence arrived. But the biggest opportunity sits in the upper right, in the intelligence frontier zone where we haven't yet imagined what becomes possible when intelligence is no longer the constraint. We're suffering from a dangerous lack of ideas about what to do with unlimited access to intelligence.

Lacking Abundant-Intelligence Use Cases - Populated: When we start populating the intelligence frontier with actual ideas, we see the massive white space of opportunity. These are intelligence-driven opportunities that only make sense when you can access reasoning capabilities at scale, problems that were unsolvable simply because we couldn't hire enough talented people to work on them. This is where 10x improvements hide, not in making existing work 10% faster but in tackling entirely new categories of problems.

Intelligence Frontier Questions: The intelligence frontier demands we ask fundamentally different questions. What's already working that we could 10x if we had access to abundant intelligence? What problems feel impossible right now simply because we can't hire enough talent to solve them? What new kinds of work become affordable when intelligence is no longer scarce? And most importantly, what are the things only humans can do, and how do we use abundant intelligence to multiply them? These questions shift the conversation from efficiency to expansion.

Generative AI Value-Creation Pyramid: Organizations should build AI value systematically through four levels. Start at the base with individual improvements that enhance productivity and build foundational skills through applied practice. Move up to collective intelligence that strengthens team alignment and integrates AI as a team member with proper governance. Then advance to transformation and growth where you drive measurable business outcomes by reimagining core processes for abundant intelligence. At the peak, pursue visionary innovation that creates entirely new markets, products, and services by expanding the horizon of possibilities with human agency. Each level builds upon the previous capabilities.

AI Agility represents the foundational capability every person in your organization needs. This is about moving beyond anxiety and building the practical skills to work effectively with AI tools. AI Agility transforms individuals from observers into participants, giving them the confidence and competence to integrate AI into their daily work without waiting for permission or perfect solutions.

AI Agility Challenge v4 is a comprehensive four-week program that takes people from AI foundations through advanced techniques. Week one covers prompting fundamentals, onboarding AI for best work, building your AI toolkit, understanding AI-first versus human-first mindsets, and knowing when to use AI for fast work versus slow work. Week two focuses on effective prompt design, elevating AI outputs, complex-task strategies, building trust in AI systems, and developing healthy AI habits. Week three advances to meta-prompting, capturing value with AI, integrating AI into workflows, understanding responsible AI, and leveraging GenAI's data analytics capabilities. Week four covers SOTA prompt design techniques, building your AI advisory board, getting creative juices flowing with AI, staying relevant in the agency economy, and aiming higher with AI for lasting resilience.

The journey to the agency economy starts with a single step. Scan the QR code on the left to access today's slides and continue exploring the frameworks and insights we've covered. Scan the code on the right to connect on LinkedIn and stay engaged as we navigate this transformation together. The future belongs to those who prepare for abundant intelligence now, not those who wait until the transformation is complete. Your next move matters more than you think.