Great insights, Sangeet! As always, you are opening our minds to new ideas. For me, this helps me tell a better story of why AI pilots are so important, as long as they are shorter time frames and we document what worked and didn't so that we can learn for the next ones. Unfortunately, other people are telling organizations to stop running pilots because they waste time. You are helping us prove why that is not true.
I like the framing. Maybe I need to understand better but almost seems like we from human judgement to large experiments back to judgment?! That is what is higher abstraction ?
The learning architecture point makes me think one layer is easy to underweight: organizational memory. If exploration becomes cheap, companies can run more experiments than they can metabolize. Unless each test updates a reusable record of what changed, where it held, and why the next commitment changed, abundance can produce repeated exploration rather than compounding learning. The advantage may depend not only on searching and testing cheaply, but on making each cycle leave the next one smarter.
Yes totally agree. The learning architecture is something like a company specific harness that helps encapsulate all of this. Your product evals, experiments, and context all evolve from one another so you need to manage and steer that.
We are seeing this at my company today - as models improve, it becomes essential to manage the context it operates in and there is real taste in how you curate that.
Very interesting perspective. One thing to note is democratisation strategic thinking. All of this might not need to happen at an organizational level but can happen at Individual level much more than before. Channeling that will also become an important element of strategy
Thank you, Sangeet. I agree with your point that reducing one form of scarcity allows companies to experiment with factors that were previously constrained by the economics of the industry.
But I wonder whether scarcity is ever really removed—or whether it simply moves.
Streaming removed the scarcity of the prime-time slot and made consumption flexible across time and place. But people did not gain more hours in the day. The same attention now competes with messaging, social media, reading, gaming, work, and an expanding number of other services and activities such as family friends, sport, etc. In many cases, that seems to have pushed consumption toward shorter, faster, and more fragmented and hybrid interactions.
So perhaps the scarce resource is no longer distribution capacity or even time, but sustained human attention.
That makes me wonder: when an industry removes its traditional scarcity, how can we tell whether it has genuinely created abundance—or merely shifted competition toward a new attention shared bottleneck that is harder to measure and potentially even more valuable?
And if more of these abundant services eventually converge on advertising or hybrid subscription models, could real attention itself become the new “prime-time slot” that everyone is bidding for?
The biggest call out here is the irreversible decisioning. If you are using AI to determine human capital deployment or reshuffle combined with shifting ICP’s and a product roadmap to match. How can your triangulate this across departments and introduce a new way of making decisions that doesn’t paralyze the org?
Still the best! Keep them coming! It's been a while...
Great insights, Sangeet! As always, you are opening our minds to new ideas. For me, this helps me tell a better story of why AI pilots are so important, as long as they are shorter time frames and we document what worked and didn't so that we can learn for the next ones. Unfortunately, other people are telling organizations to stop running pilots because they waste time. You are helping us prove why that is not true.
I like the framing. Maybe I need to understand better but almost seems like we from human judgement to large experiments back to judgment?! That is what is higher abstraction ?
The learning architecture point makes me think one layer is easy to underweight: organizational memory. If exploration becomes cheap, companies can run more experiments than they can metabolize. Unless each test updates a reusable record of what changed, where it held, and why the next commitment changed, abundance can produce repeated exploration rather than compounding learning. The advantage may depend not only on searching and testing cheaply, but on making each cycle leave the next one smarter.
Yes totally agree. The learning architecture is something like a company specific harness that helps encapsulate all of this. Your product evals, experiments, and context all evolve from one another so you need to manage and steer that.
We are seeing this at my company today - as models improve, it becomes essential to manage the context it operates in and there is real taste in how you curate that.
Very interesting perspective. One thing to note is democratisation strategic thinking. All of this might not need to happen at an organizational level but can happen at Individual level much more than before. Channeling that will also become an important element of strategy
Thank you, Sangeet. I agree with your point that reducing one form of scarcity allows companies to experiment with factors that were previously constrained by the economics of the industry.
But I wonder whether scarcity is ever really removed—or whether it simply moves.
Streaming removed the scarcity of the prime-time slot and made consumption flexible across time and place. But people did not gain more hours in the day. The same attention now competes with messaging, social media, reading, gaming, work, and an expanding number of other services and activities such as family friends, sport, etc. In many cases, that seems to have pushed consumption toward shorter, faster, and more fragmented and hybrid interactions.
So perhaps the scarce resource is no longer distribution capacity or even time, but sustained human attention.
That makes me wonder: when an industry removes its traditional scarcity, how can we tell whether it has genuinely created abundance—or merely shifted competition toward a new attention shared bottleneck that is harder to measure and potentially even more valuable?
And if more of these abundant services eventually converge on advertising or hybrid subscription models, could real attention itself become the new “prime-time slot” that everyone is bidding for?
Scarcity always moves.
The biggest call out here is the irreversible decisioning. If you are using AI to determine human capital deployment or reshuffle combined with shifting ICP’s and a product roadmap to match. How can your triangulate this across departments and introduce a new way of making decisions that doesn’t paralyze the org?
A symphony if ideas in an otherwise very noisy environment! Thanks for sharing these impactful thoughts with clarity