Essay Topic Why is There Massive Revenue Opportunity for PEAI?
What Are the Impressive Facts That Most People Don't Know Here?
Toyota led with a gross profit of around $3,000 per vehicle, while the rest of the industry averaged around $1,500 per vehicle for decades. Tesla surpassed them all, reaching over $10,000 per vehicle by 2022.
How Do These Relate to AI?
Tesla operates under a concept known as 'management by AI.' This approach minimizes traditional middle management by relying heavily on data-driven decision-making.
AI tools handle many of the roles previously assigned to middle managers, such as coordinating projects, analyzing performance metrics, and allocating resources, resulting in faster decision-making.
What are the main learnings?
Middle Management will largely cease to exist. Tesla has largely eliminated this not only for cost savings but speed to revenue growth.
Resource allocation is better handled by machine. Humans are inherently inferior decision makers vs. machine1.
How do these relate to revenue growth?
Framework
A company has two ways to grow revenue: increase prices or sell more units. Both require making game theory optimal2 bets at high velocity.
Eventual Outcome = Current Advantages X Quality of Learning Bets X Speed Implementing Quality Learnings
Pace of Improvement = OODA3 Loop Cycles per Period X Improvement per OODA Loop
Price Increase Example
Tesla understood that to increase the average selling price it needed to sell an SUV. An SUV is heavy which reduces the EV range. Batteries are also heavy. To deliver an SUV with range the weight of the battery needed to be dropped. With that comes tradeoffs of battery longevity etc.
The challenge GM, VW, etc. face is this requires immense “cross-functional collaboration” - this is code for lots of human communication overhead. In addition, most of these companies do their planning quarters if not years ahead which require making assumptions with lots of margin of safety.
In contrast, Tesla was able to effectively hand all R&D asset allocation decisions to the engineers themselves as well as allow them to self organize their teams. They could do this because they used “AI” to help make these decisions and “re-underwrite” several times a day.
One of the best ways to look at this is through the lens of the OODA Loop. A company needs to Observe (aka collect data), Orient (stitch together insights), Decide, and Act.
In this scenario the human needs to do two things. First it needs to synthesize what is going on. This requires deep understanding coupled with multiple domain mental models. Second, there needs to be an accurate decision of what to do given this information.
Tesla in contrast can place several bets several times a day, reorient teams around them, and quickly burn-down the high value uncertainty. Toyota forever was viewed as the leader of this and could not keep up.
Companies that can place R&D and product bets faster and more accurately via “management by AI” will be able to pull away from the competition.
First Principles: What Can We Infer as the Current Bottlenecks of Value?
Speed of bet re-underwriting
Speed of uncertainty burn-down
% of resources required on information collection and decision making vs. direct uncertainty burn down
What is my personal experience?
In a past life I worked directly with the CEO of a company with a market cap of ~$250B. Our margins were good, but we were hitting the ceiling of our category in terms of revenue (we owned the market). We had spent the past several years purchasing 10+ relevant companies to expand the category reach. However, this had hit the point of diminishing returns as well.
We had exhausted all low hanging fruit and needed to place bigger bets of what the customer was going to need in the future. My formula for this is the cost of the problem to the customer (direct cost + opportunity cost), % of the problem we can solve via our solutions, % of that problem solved we can capture as take rate.
The challenge was our cycle time was slow (even though by software standards we were considered the absolute leader). The implication is we could not afford to take large risks. Bottom line this had negative effects for our customers4 and our investors5.
Bottom Line: Where will companies pull ahead?
My hypotheses: Companies where R&D requires lots of different domain expertise (aka cross functional communication overhead) are likely to see immense upside on product quality due to management by AI.
Humans currently outperform machines in areas requiring complex contextual understanding (e.g., navigating unpredictable driving situations), moral decision-making (e.g., ethical dilemmas in medical triage), creativity (e.g., generating abstract art or inventing new product concepts), empathy (e.g., negotiating or providing emotional support), handling ambiguity (e.g., resolving conflicts between stakeholders with unclear objectives), and physical adaptability (e.g., performing manual repairs in unstructured environments).
The bottlenecks for AI include limited real-world experience, lack of ethical frameworks, absence of subjective experience, and challenges in recognizing emotional cues or adapting physically. However, rapid advancements in neural networks, reinforcement learning, generative models, and sensory technologies are quickly reducing these gaps, allowing AI to handle increasingly complex, empathetic, and creative tasks with greater proficiency.
Game theory optimal (GTO) refers to a strategy that maximizes a player's expected payoff by anticipating all possible actions and counteractions, making it unbeatable in a given scenario if executed perfectly. It assumes rational players and seeks equilibrium where no one can improve their outcome by unilaterally changing their strategy.
The OODA loop—Observe, Orient, Decide, Act—is a decision-making framework designed to help navigate complex, dynamic situations. Observe gathers data from the environment; Orient processes that data by integrating previous experiences and context; Decide selects the best course of action; and Act implements that decision. This cycle is continuous, allowing for real-time adjustments as new information becomes available. Originally developed for military strategy, it’s now applied in business and other fields to maintain agility and outmaneuver competitors by quickly cycling through decision-making faster than others.
Missed out on more of their problems being solved
Missed out on higher discounted cashflow