This write-up is great. I shared it with our teams this morning. It's obvious the current clumsiness will be supplanted by more elegant solutions over time. The most interesting thing on my mind is the future of pods to effectively rebuild the digital infrastructure of commerce. What exactly does a team need to look like (skills, size, hierarchy) to properly build a commerce solution and scale and what other system changes are required internally and externally for this to occur? It's easy to build a shopping agent that doesn't really work. Rhetorical question alert: What does it take to make one that really works?
Thank you, Andy! It's an interesting question. I think the first system change will be data standardization -- which is why OpenAI started out by partnering with big marketplaces that standardize across many smaller sellers, and why they're forcing merchants who use their platform to support their Checkout API format & offer checkout via the Agentic Commerce Protocol.
The team skills question is a more nuanced one, particularly as you consider just how broad the consideration set is -- not only do you need good tech, you also need strong brand/merchant/payment processor partnerships, well-trained customer support and systems in place for things like returns/exchanges, and also safety and regulatory considerations (e.g. customs, avoiding prompt injection in product names). Especially for a company like OpenAI who already has a huge user base, being prepared for all of that on day one is a serious feat; rough edges are inevitable, but I'm sure they'll iterate quickly!
Very interesting launch, I will say even though it's an "obvious" monetization model, I worked at another company where it was an incredibly "obvious" monetization model (Pinterest) and yet it was incredibly difficult in practice, because there's a big operational jump from "user has expressed interest in something" and "merchant is ready to ship that item". But with ACP it looks like OpenAI has a pretty good plan and certainly demonstrated the ability to execute on complex endeavors.
Super interesting; I hadn't thought about that with Pinterest, but it makes a lot of sense.
I do think that's part of why the accurate product graph is so critical (but, difficult), and I'm impressed by how well OpenAI seems to be pulling in data from even companies that aren't explicitly partners. As you noted, I think ACP will be helpful because it forces merchants to standardize onto OpenAI's checkout API format, which will likely help to bridge that gap. But there are still plenty of competing standards, so we'll have to see!
This write-up is great. I shared it with our teams this morning. It's obvious the current clumsiness will be supplanted by more elegant solutions over time. The most interesting thing on my mind is the future of pods to effectively rebuild the digital infrastructure of commerce. What exactly does a team need to look like (skills, size, hierarchy) to properly build a commerce solution and scale and what other system changes are required internally and externally for this to occur? It's easy to build a shopping agent that doesn't really work. Rhetorical question alert: What does it take to make one that really works?
Thank you, Andy! It's an interesting question. I think the first system change will be data standardization -- which is why OpenAI started out by partnering with big marketplaces that standardize across many smaller sellers, and why they're forcing merchants who use their platform to support their Checkout API format & offer checkout via the Agentic Commerce Protocol.
The team skills question is a more nuanced one, particularly as you consider just how broad the consideration set is -- not only do you need good tech, you also need strong brand/merchant/payment processor partnerships, well-trained customer support and systems in place for things like returns/exchanges, and also safety and regulatory considerations (e.g. customs, avoiding prompt injection in product names). Especially for a company like OpenAI who already has a huge user base, being prepared for all of that on day one is a serious feat; rough edges are inevitable, but I'm sure they'll iterate quickly!
Very interesting launch, I will say even though it's an "obvious" monetization model, I worked at another company where it was an incredibly "obvious" monetization model (Pinterest) and yet it was incredibly difficult in practice, because there's a big operational jump from "user has expressed interest in something" and "merchant is ready to ship that item". But with ACP it looks like OpenAI has a pretty good plan and certainly demonstrated the ability to execute on complex endeavors.
Super interesting; I hadn't thought about that with Pinterest, but it makes a lot of sense.
I do think that's part of why the accurate product graph is so critical (but, difficult), and I'm impressed by how well OpenAI seems to be pulling in data from even companies that aren't explicitly partners. As you noted, I think ACP will be helpful because it forces merchants to standardize onto OpenAI's checkout API format, which will likely help to bridge that gap. But there are still plenty of competing standards, so we'll have to see!