Lovely piece. I tend to think that Intelligence advantage is hard to gain by upstarts amirite (assuming you need a lot of historic data, domain specific data etc...)? so the Scenario 2 you seem to consider likely sounds a bit unlikely to me.
Also I'm wondering of the dependency and constraints that lower levels that the unbundlers disintermediate can use to impede an actual OTT rebundling (eg: TOS).
My view is that there are wrapper arbitrage opportunities where if you solve a use case well, you have a limited period during which you can gain traction and data flows and that may help build an intelligent advantage in a vertical.
Great post, Sangeet. Very insightful and well articulated especially the examples of WeChat and Autodesk.
On a much simpler note, would it be fair to assume that the unbundling of value is more like, operators trying to solve niche individual use cases, which eventually don't have massive monetization opportunities because you can only grow so much - like product features that don't have a long runway, standalone.
Then eventually you may scale upto solving a bunch of those niche use cases. Now only if those are well thought through use cases which are adjacent in the value chain of the user problems then you must stitch them well to offer a seemless UX to the user leading to a comprehensive value for the user.
The advantage here lies more with the folks who have the empathy to interpret the customer problem over just expanding the scope of tech.
Eventually, like previous tech waves, winning here would require a deeper understanding of the core customer problems over the tech prowess of the models..
Am I abstracting too much away from the essence of generative AI?
Thanks Ankur. You are largely on point. Except that rebundling is not just stitching together a bunch of use cases but doing so through a control point. Without a control point, you will remain unbundled.
Hi Sangeet. This is partially off-topic, but I couldn't find another way to reach out.
You offered some of your ebooks in exchange for substack referrals. How do you keep track of these, since I believe I've brought three in already, and qualify for the first giveaway.
Thanks, this is managed by Substack and their dashboard attributes 2 referrals to you. It automatically sends you an email when the third referral comes in. The data and workflow are entirely owned by Substack.
Excellent, makes a lot of sense. Stitching together is still close to aggregating but not a good rebundling exercise. Thanks for quick answer, Sangeet.
Excellent synthesis of a complex topic. In your 2X2 of Model Scale vs Customer Discovery, the top right quadrant could evolve over time to be dominated by Incumbents who are smart enough to build high scope/scale model for their industry combined with high customer knowledge/discovery. For example, Schlumberger in Oil/Gas discovery already has $+resources for to build for O&G industry a top-right quadrant gen AI solution (hypothetically). This would be empire strikes back / defends its turf scenario. Thoughts?
I agree that incumbents can deliver well there. The challenge largely remains that any significant fine-tuning will need model engineers who have cut their teeth not in your industry but in foundational models, because that's where most of the innovation has been in the past 5 years. We will see a period of 3-4 years where we will have some talent disconnect on that front till vertical specialization emerges among model engineers. A lot of vertical specialists are more traditional data scientists.
Especially, I find the comparison to app stores interesting. I'm wondering how you see the positioning of Generative AI "app stores" like OpenAI's plugin store or the announced extensions for Bard and Copilot? I can imagine that the dominating foundational model providers aim to keep control of the interface to tools specialized for specific verticals.
OpenAI has already displayed a history of bait and switch with their governance and policy changes. It's very much wild West at the moment. Similar to Facebook back in 2009 where developers just wanted to build apps on Facebook at all cost, till Facebook turned off the faucet. I imagine that's coming here pretty soon as well.
This is one of the best things I’ve read in a month or more. I like it because a) it echoes how I think AI will restructure markets (and echoes how I’m building a side project for a consumer facing platform) b) it is extremely logical and ‘first principles’ based c) expanded my ‘pattern recognition’ in this space.
Thanks! Really appreciate it and those 3 are things I look to bring to every analysis. There's an intrinsic joy to explaining things through first principles that is a reward in itself.
Great post. The closest analogy is to SaaS as it unbundled and then the winners rebundled their offerings to monetise. The example of Adobe and Autodesk that you called out are examples of incumbents who managed to “eat the fish” as they went from on-prem to SaaS and Subscription.
The cycle of unbundling and rebundling has played out with every tech shift and Saas is certainly one of the examples. It's probably the most recent and most well understood one but this has also played out in the shift from mainframe to personal computing as well as in shifts in databases, shift to cloud/saas, shift to mobile, and many others.
The reason I take the mobile and WeChat example above - and not the example of other shifts including Saas - is that mobile in China is the only other shift where the underlying horizontal (Apple App Store) was a consumer product like it is this time (ChatGPT). With Saas, the underlying horizontal (AWS) was not a consumer product.
When the underlying horizontal is a consumer product, people assume that the plays can only be vertical. My point here is also to bring out that the winner will actually be horizontal, to some extent what WeChat did to iOS in China.
Great post, Sangeet, and love the framing of the various stages of unbunding and rebundling. Completely agree that establishing a control point is key, and yet so not well understood by incumbents who seek to "consolidate" the tech stack without first establishing a "control point" to earn the right to rebundle.
Would it be fair to say that rebundling must create more customer value as compared to merely adding up individual workflows and apps? And that 'customer value' needs to be more than cost savings to please the CFO?
How to identify and establish the control point is probably worthy of its own post ;-) We've taken a similar approach with my start-up, Tribyl, in the Revenue intelligence space. One key learning is to find a user group that is already acting as that tacit "control point" for upstream and downstream workflows, and is motivated to "own" that control point. It's easier to drive a wedge by serving the needs of this user group, first. This user group also likely has relationships with other stakeholder groups in the org.
Another implication is how VCs should think about GenAI thesis and funding. Going through the stages you outline will take a lot longer and may require more $$ to build generational AI companies. Most VCs still think in terms of the traditional SaaS playbook.
Lovely piece. I tend to think that Intelligence advantage is hard to gain by upstarts amirite (assuming you need a lot of historic data, domain specific data etc...)? so the Scenario 2 you seem to consider likely sounds a bit unlikely to me.
Also I'm wondering of the dependency and constraints that lower levels that the unbundlers disintermediate can use to impede an actual OTT rebundling (eg: TOS).
btw as always a great piece!
Thanks!
My view is that there are wrapper arbitrage opportunities where if you solve a use case well, you have a limited period during which you can gain traction and data flows and that may help build an intelligent advantage in a vertical.
Great post, Sangeet. Very insightful and well articulated especially the examples of WeChat and Autodesk.
On a much simpler note, would it be fair to assume that the unbundling of value is more like, operators trying to solve niche individual use cases, which eventually don't have massive monetization opportunities because you can only grow so much - like product features that don't have a long runway, standalone.
Then eventually you may scale upto solving a bunch of those niche use cases. Now only if those are well thought through use cases which are adjacent in the value chain of the user problems then you must stitch them well to offer a seemless UX to the user leading to a comprehensive value for the user.
The advantage here lies more with the folks who have the empathy to interpret the customer problem over just expanding the scope of tech.
Eventually, like previous tech waves, winning here would require a deeper understanding of the core customer problems over the tech prowess of the models..
Am I abstracting too much away from the essence of generative AI?
Thanks again for the excellent post..
Thanks Ankur. You are largely on point. Except that rebundling is not just stitching together a bunch of use cases but doing so through a control point. Without a control point, you will remain unbundled.
Hi Sangeet. This is partially off-topic, but I couldn't find another way to reach out.
You offered some of your ebooks in exchange for substack referrals. How do you keep track of these, since I believe I've brought three in already, and qualify for the first giveaway.
Thanks, this is managed by Substack and their dashboard attributes 2 referrals to you. It automatically sends you an email when the third referral comes in. The data and workflow are entirely owned by Substack.
Ok, thanks. Then some of my chums might not have used the referral link.
Excellent, makes a lot of sense. Stitching together is still close to aggregating but not a good rebundling exercise. Thanks for quick answer, Sangeet.
Excellent synthesis of a complex topic. In your 2X2 of Model Scale vs Customer Discovery, the top right quadrant could evolve over time to be dominated by Incumbents who are smart enough to build high scope/scale model for their industry combined with high customer knowledge/discovery. For example, Schlumberger in Oil/Gas discovery already has $+resources for to build for O&G industry a top-right quadrant gen AI solution (hypothetically). This would be empire strikes back / defends its turf scenario. Thoughts?
I agree that incumbents can deliver well there. The challenge largely remains that any significant fine-tuning will need model engineers who have cut their teeth not in your industry but in foundational models, because that's where most of the innovation has been in the past 5 years. We will see a period of 3-4 years where we will have some talent disconnect on that front till vertical specialization emerges among model engineers. A lot of vertical specialists are more traditional data scientists.
Very interesting read. Thanks!
Especially, I find the comparison to app stores interesting. I'm wondering how you see the positioning of Generative AI "app stores" like OpenAI's plugin store or the announced extensions for Bard and Copilot? I can imagine that the dominating foundational model providers aim to keep control of the interface to tools specialized for specific verticals.
OpenAI has already displayed a history of bait and switch with their governance and policy changes. It's very much wild West at the moment. Similar to Facebook back in 2009 where developers just wanted to build apps on Facebook at all cost, till Facebook turned off the faucet. I imagine that's coming here pretty soon as well.
This is one of the best things I’ve read in a month or more. I like it because a) it echoes how I think AI will restructure markets (and echoes how I’m building a side project for a consumer facing platform) b) it is extremely logical and ‘first principles’ based c) expanded my ‘pattern recognition’ in this space.
Much appreciated.
Thanks! Really appreciate it and those 3 are things I look to bring to every analysis. There's an intrinsic joy to explaining things through first principles that is a reward in itself.
Great writeup. Will be interesting to see who dominates. Gen AI seems to be showing up in many different products.
Great post. The closest analogy is to SaaS as it unbundled and then the winners rebundled their offerings to monetise. The example of Adobe and Autodesk that you called out are examples of incumbents who managed to “eat the fish” as they went from on-prem to SaaS and Subscription.
Thanks Ram.
The cycle of unbundling and rebundling has played out with every tech shift and Saas is certainly one of the examples. It's probably the most recent and most well understood one but this has also played out in the shift from mainframe to personal computing as well as in shifts in databases, shift to cloud/saas, shift to mobile, and many others.
The reason I take the mobile and WeChat example above - and not the example of other shifts including Saas - is that mobile in China is the only other shift where the underlying horizontal (Apple App Store) was a consumer product like it is this time (ChatGPT). With Saas, the underlying horizontal (AWS) was not a consumer product.
When the underlying horizontal is a consumer product, people assume that the plays can only be vertical. My point here is also to bring out that the winner will actually be horizontal, to some extent what WeChat did to iOS in China.
This part is a bit confusing.
“Incumbents lacking the model and/or UX chops will be waiting on the sidelines for such opportunities.”
Waiting on the sidelines?
Scouting the market instead of trying to build internally.
Great post, Sangeet, and love the framing of the various stages of unbunding and rebundling. Completely agree that establishing a control point is key, and yet so not well understood by incumbents who seek to "consolidate" the tech stack without first establishing a "control point" to earn the right to rebundle.
Would it be fair to say that rebundling must create more customer value as compared to merely adding up individual workflows and apps? And that 'customer value' needs to be more than cost savings to please the CFO?
How to identify and establish the control point is probably worthy of its own post ;-) We've taken a similar approach with my start-up, Tribyl, in the Revenue intelligence space. One key learning is to find a user group that is already acting as that tacit "control point" for upstream and downstream workflows, and is motivated to "own" that control point. It's easier to drive a wedge by serving the needs of this user group, first. This user group also likely has relationships with other stakeholder groups in the org.
Another implication is how VCs should think about GenAI thesis and funding. Going through the stages you outline will take a lot longer and may require more $$ to build generational AI companies. Most VCs still think in terms of the traditional SaaS playbook.
Would love to chat more!
https://www.ketryx.com/
is this company a good example showing early signs of control hub position?