How OpenAI Can Avoid Google’s Mistake
Happy Friday, Fintech Takers!
After a fun and productive trip to Salt Lake, I am back in Montana, where it is cold and dreary and utterly Fall, which is my wife’s favorite season of the year (and a season that we don’t usually have for long in these parts).
Next week, Kiah and I will be in Washington, D.C., where we are thrilled to be participating in the first-ever Fintech Takes Coworking Day on Thursday, October 9th.
The idea (which was dreamed up by Rachit Khaitan) is to pick a city with a lot of banking and fintech folks, reserve some space at a centrally located coworking location, and allow people to work in the same space for a day.
In the age of remote work, this is exactly the type of thing we need to be doing more of. Not conferences or trade shows (though those have their place), but rather creating space for people to get work done *and* to experience the benefits of synchronicity, which only happens when people are together in the same physical location.
At the end of the day, we will celebrate with an evening reception, hosted by my friends at Canapi Ventures. We still have some available space for folks to attend that reception, so if you’ll be in D.C. next Thursday and are interested in attending, apply here.
And if you are interested in participating or organizing future Fintech Takes Coworking Days, you should apply to join the Fintech Takes Network. That’s where this concept was born, and that’s where banking and fintech nerds are connecting with each other (digitally and IRL)!
- Alex
How OpenAI Can Avoid Google’s MistakeCopy anchor linkCopied
Today’s essay is about OpenAI and agentic commerce. And because this is a very popular topic right now, and because this essay is going to get a little theoretical towards the end, I want to start with some facts.
ChatGPT is one of the most popular websites and apps in the world. In November of 2023 (a year after OpenAI released ChatGPT), it wasn’t even in the top 100 websites. Now it’s more popular than Twitter, Reddit, WhatsApp, and Wikipedia. And it’s quickly gaining on the big four: Instagram, Facebook, YouTube, and Google.
What’s most notable about this, as this excellent article from SemiAnalysis points out, is how quickly ChatGPT has ascended to this rarified air. It is significantly younger, as a product, than any of the others in the top 10:
(Editor’s Note — Bing! Lol. Embedded distribution for the win!)
This is rather astounding on first glance, but it makes sense when you look at its year-over-year growth rate, compared to its peers:
The result of this growth is that OpenAI, in addition to being arguably the most important technology infrastructure provider in the market today (or at least one of them), also has a MASSIVE consumer business.
In February, ChatGPT had 400 million weekly active users. By August of this year, it was 700 million. And while I haven’t seen any estimates that are more recent, 800+ million is not at all unreasonable.
What’s most interesting (and relevant for our purposes) is that the vast majority of those users are free users, rather than paying subscribers. OpenAI hasn’t publicly shared stats on this, but estimates I have seen put the percentage of free users at 97-99%.
And that, obviously, raises an important question: How is OpenAI going to monetize all of those free ChatGPT users?
OpenAI’s Commerce OpportunityCopy anchor linkCopied
It has been clear for quite some time that OpenAI was going to pursue opportunities at the intersection of AI and commerce.
In May, the company hired Fidji Simo to be its new CEO for its Applications business. Simo was formerly the CEO of Instacart and, before that, a long-time executive at Meta (including the head of Facebook). It also hired Kevin Weil, a former executive at Twitter and Instagram, to be its Chief Product Officer, and it was recently reported that the company is looking to hire a head of monetization (including advertising), reporting to Simo.
What was less clear, until recently, was exactly how OpenAI would try to realize these opportunities.
OpenAI’s newest product — GPT-5 — gave us our first clue, with the shift from a discrete set of different models (optimized for different speed/performance trade-offs) to a single, unified system that uses an intelligent router to determine, in real-time, which model to use for each inquiry, based on the conversation type, complexity, tool needs, and the explicit intent of the user.
The SemiAnalysis article I linked to above explains why this is such a notable shift:
The router serves multiple purposes on both the cost and performance side. On the cost side, routing users to mini versions of each model allows OpenAI to service users with lower costs. On the performance side, it will enable many users to use thinking aka CoT (Chain of Thought) reasoning for the first time. Over 99% of the free users have yet to interact with a thinking model like o3, and for the average user, ChatGPT just got a huge upgrade. The number of free users exposed to thinking models went up 7x in the first day and the number of paying users up nearly 3.5x.
This is interesting! Even though AI power users like myself were a bit disappointed by GPT-5, it represents a huge capability upgrade for free ChatGPT users (who will be benefiting from models using Chain of Thought reasoning for certain queries), while, at the same time, allowing OpenAI to dynamically downshift to less expensive models when they are sufficient for other queries.
This dynamic capability, enabled by the GPT-5 router, introduces a massive change in the way that computing power (and thus cost) is allocated for different consumer search queries.
In the old world (described wonderfully in Ben Thompson’s work on aggregation theory), the marginal cost for internet-enabled companies — like Google — to serve one additional user was basically zero. Combined with the right monetization model (advertising), this zero-marginal-cost attribute allowed these companies to generate unprecedentedly enormous profits.
However, they also flattened the results for the end user. While different search queries were worth more or less to Google (based on how much commercial intent they signaled to advertisers), the results for users were always the same. A search for “Why is the sky blue?”
resulted in roughly the same experience for the user as a search for “What is the best DUI lawyer near me?”
Thanks to routing, the free version of ChatGPT can dynamically answer a harder or more valuable question with a better answer. The SemiAnalysis article elaborates:
As you may know, [“What is the best DUI lawyer near me?”] is an extremely valuable question. Today on search, this is one of the higher cost per click keywords, and it is plastered with ads. In a world of dynamic supply, ChatGPT can not only answer this question, it could realize this is a very valuable question and answer this question at the level of a human. It could throw $50 dollars of compute if there is a belief of high conversion, because that transaction is worth $1000s of dollars.
The router makes this possible. ChatGPT 5 could decide to allocate $50 to the query, create a plan, gather information about the incident, research local lawyers, consider who is likely to answer fastest, consider your budget, and then contact multiple lawyers on your behalf. It could even Agentically reach out to lawyers on behalf of the free user knowing that the conversion ratio of this query is even higher.
The implications of this are both easy to imagine and deeply disruptive to the status quo in search and commerce. SemiAnalysis goes on:
You can see the future. Imagine a world where you ask for new dinner recipes for the week, and ChatGPT gives you multiple options and orders the cart for you to check out. The fee would be paid on the completion of the purchase, and Search is completely cut out of the picture. Everything that can be researched or planned in an AI app could be purchased
The SemiAnalysis article goes on to explain that in order to actually make this future a reality, OpenAI would need to do a lot of plumbing work, integrating merchants’ systems with ChatGPT and figuring out a way to make payments within the chatbot interface as seamless as possible.
If only OpenAI could find some partners that could help it with those things!
Ohh hey, Stripe, Etsy, and Shopify! I didn’t see you there!Copy anchor linkCopied
Last week, we got some very interesting news:
[OpenAI and Stripe] unveiled an Instant Checkout feature in ChatGPT, powered by a new commerce protocol they co-developed. The feature is launching first with U.S.-based Etsy sellers and will soon extend to more than a million Shopify merchants, including buzzy brands like Glossier, Skims, Spanx, and Vuori. The protocol sits on top of an open standard for connecting AI models to business systems, developed by Anthropic, called MCP — but focuses specifically on commerce and payments. Stripe brings fraud prevention, global payment rails, and a vast merchant network, making the new Agentic Commerce Protocol (ACP) usable by millions of businesses right out of the gate.
From what I understand, here’s how it works:
- The user asks ChatGPT something like, “Help me pick a nice coffee mug under $25.” The model searches or ranks product listings and returns a few that support Instant Checkout (via ACP-compatible merchants like Etsy and Shopify).
- The items from ACP-compatible merchants will have a “Buy” button. When the user clicks it, ChatGPT issues a request to the merchant’s ACP endpoint to start the checkout.
- If the user has payment info stored in Stripe Link, the transaction is authorized (via encrypted tokens) for the specific merchant and amount. If not, they enter payment info (and are given the option to save it for future use). The ACP ensures this is done securely and only for the intended merchant/amount.
- The merchant’s backend receives the order, processes it using its existing systems (inventory, shipping, returns), and reports status (confirmation, tracking, refunds). ChatGPT may display status or let users check order progress.
- OpenAI collects a commission from the merchant (not from the user).
So, to sum up, ChatGPT now allows free users to access sophisticated Chain of Thought reasoning for valuable, research-heavy queries (AKA commercial queries) *and* it now offers a protocol that allows any merchant to feed it product, pricing, and inventory information and facilitate transactions to purchase those products directly within the chat interface.
OK! Now we’re cooking with gas!
However, there’s one final question that is important to answer: How will ChatGPT determine which products to recommend?
No one knows yet, but as the Fortune article linked above points out, OpenAI will have some tough questions to answer on this front:
ACP effectively positions ChatGPT as a new arbiter of product recommendations. Instead of shoppers browsing Google search results, scrolling Amazon’s “customers also bought,” or consulting reviews on Wirecutter, the assistant itself will increasingly decide what to surface. That shift raises thorny questions: How will ChatGPT determine which product to recommend? Will it offer a menu of options or streamline to a single choice? And down the line, will OpenAI accept money from vendors to boost their placement — turning conversational commerce into a pay-to-play channel?
The answer to that last question is likely yes. OpenAI has to figure out a path to profitability (or, at least, less staggering losses), and the most obvious path is to sprinkle advertising on top of its massive free user base. And according to reporting, the company is working to hire an executive to lead its monetization efforts, which will include advertising.
But will OpenAI implement advertising the same way that Google and all of its other commerce enablement competitors have?
I don’t think so.
How OpenAI Can Avoid Google’s MistakeCopy anchor linkCopied
OpenAI’s CEO, Sam Altman, has never been a big fan of ads. Here is a quote from a fireside chat with him from last year:
I will disclose, as a personal bias, I hate ads. I think ads were important to give the early internet a business model. I’m not totally against them, but ads plus AI are uniquely unsettling to me. I kind of think of ads as a last resort as a business model.
More recently (and likely thanks to the influence of Fidji Simo), Altman has taken a slightly softer stance:
I am not totally against it… if you compare us to social media or web search where you can kinda tell that you are being monetized… we would hate to ever modify anything in the stream of an LLM… maybe if you click on something in there that is going to be there we’d show anyway, we’ll get a bit of transaction revenue and it’s a flat thing for everything, maybe that could work. It’s clearly possible to be a good ad-driven company, but there are obviously issues to it.
Still, he seems to be reluctant. And if we parse that second quote a little bit, I think we can understand why (emphasis mine):
if you compare us to social media or web search where you can kinda tell that you are being monetized… we would hate to ever modify anything in the stream of an LLM
What Altman seems to object to is an advertising model that prioritizes the interests of the highest-paying advertisers over the experience of users.
He acknowledges the importance of that model in paying for the build-out of the early internet, but he appears very nervous about making the same mistake that Google did: squandering a massive early lead in consumer adoption by slowly degrading the experience of those consumers using the product.
He needs an alternative approach. One that does a better job of aligning the interests of users and advertisers.
If you think about how Google attempts to achieve this alignment, you can easily understand why Altman hates ads so much. The design of the search engine makes the implicit argument that users can trust it because it separates ads from organic search results. However, as Google has grown (and become more focused on shareholder returns), it has increasingly blurred the lines between ads and organic content and polluted the results page:
Or, as Charlie Warzel at the Atlantic observed a few years ago:
Unlike its streamlined, efficient former self, Google Search is now bloated and overmonetized. It’s harder now to find answers that feel authoritative or uncompromised … Using Google once felt like magic, and now it’s more like rifling through junk mail, dodging scams and generic mailers.
Sam Altman wants the experience of using ChatGPT to continue feeling magical. He seems to understand that companies always, eventually, become their business model. He needs an approach to digital advertising that structurally protects the product from the commercial pressures that OpenAI will one day be subject to.
Here’s my idea: Quality-weighted Ads
Instead of letting the biggest wallet win the ad placement, OpenAI could price and rank ads by how happy users are after they buy.
This is an approach that OpenAI is uniquely well-suited to implement. Google’s search engine has no visibility into its users’ post-purchase experiences. By contrast, with purchases made inside an AI chatbot like ChatGPT, the assistant already knows enough context to follow up at the right moment. “Did it arrive on time?” “Was it as described?” “Would you buy again?” One or two taps, optional free-text, and done. No public reviews or review bombing. Just a quiet, first-party signal from a real, verified buyer.
OpenAI could roll those answers up, by category, with sane normalization (i.e., sofas aren’t socks, budget brands aren’t luxury brands, etc.) It could use recency and sample size to keep the numbers honest, and shrink noisy data toward the category average so nobody games the system with a handful of five-star plants.
The company could then plug that quality signal straight into its ad market. Every advertiser would still set a bid. OpenAI would convert that bid into an “effective” bid by applying a small multiplier based on post-purchase quality. Great experiences would get a discount. Mediocre experiences would pay a surcharge. Truly bad experiences would get throttled or, if they keep failing, removed.
Money would still matter. But money couldn’t consistently outrank a better experience.
Importantly, the ranking would also continue to care about relevance in the moment. If you’re asking for trail-running shoes, a mattress company shouldn’t be able to buy its way to the top, no matter how beloved it is. Quality would help merchants win when they belong; it wouldn’t buy them relevance they don’t have.
For users, the upside would be obvious. Fewer junky pitches. More products that actually deliver on their promises. A sense that giving quick feedback makes the system better for you and everybody else. For advertisers, the incentive would be equally clear. Fix your shipping SLAs, improve support, stop over-promising in your creative, and your ad spend goes further.
Obviously, the integrity of the system would be critical. Only verified buyers would contribute to the quality signal. Strong anomaly detection would spot sudden bursts of support from suspicious places. Clear appeals and fast re-estimates would allow merchants to get out of the dog house if they could demonstrate that they addressed the problems.
This is how OpenAI could avoid Google’s mistake.
Don’t make ads the product that slowly bends the user experience around it. Make ads compete to serve the user experience that already exists.
Let ChatGPT discover what happened after the click — and price the next impression accordingly.
INDUSTRY ASK
I’m putting together a webinar next month about the best way to protect consumers and lenders featuring TruStage and the Financial Health Network (come Nov. 6th, which is somehow both very soon and very far away).
I’d like to add someone from a credit union or bank who is knowledgable about portfolio management and collections to round it out. Know someone? Hit reply and make a recommendation!
MORE QUESTIONS TO PONDER TOGETHER
Big news for the endlessly curious (yes, you): I’m collecting your fintech questions on a rolling basis.
What’s keeping you up at night? What great mysteries in financial services beg to be unraveled? Think of it this way, if a stranger is a friend you just haven't met yet, your question is a Fintech Takes conversation waiting to happen.
One that could headline a Friday newsletter or be answered in an upcoming Fintech Office Hours event.
Drop your question here, whenever inspiration strikes!
MONEY20/20 SPOTLIGHT
It’s officially Money 20/20 season, which means I’ll be highlighting a handful of (rotating) sessions, meetups, and happenings in every newsletter.
Some I’ll be at, some just look too good to miss. You’re welcome!
🏀 Fintech Takes The Court @ Money20/20 | 10/26 | 10am–2pm PT
Sunday morning: we’re playing basketball, baby!
Join us for a high-energy 3x3 pickup game, hosted by Fintech Takes & SOLO. All are invited - female, male, young, old, seasoned, out of shape. Bring your A-game (or just your sneakers) and come play or hang courtside with fellow fintech enthusiasts. RSVP here.
💰 Deepfakes, Real Risk: Fighting Fraud in an Age of Synthetic Identity | 10/26 | 3:00pm–3:30pm PT
I’m thrilled to be moderating this discussion at Money20/20, featuring the head of fraud at Varo and the co-founders and CEOs of SentiLink and Oscilar.
☕ Nova Credit Coffee + Conversation | 10/27 | 8:15am–10:30am PT
Start the AM with lending leaders unpacking the real-world journey of cash flow analytics (where to begin, how to apply it, and what it takes to make it work). Breakfast, networking, and discussion included! RSVP here.
🍸 MX Happy Hour Panel: Data into Action | 10/27 | 3:30pm–6pm PT
Small panel conversation featuring Jane Barratt (Chief Advocacy Offer, MX) and yours truly (among others!): 3:30pm–4pm panel, followed by drinks and hors d’oeuvres at The Grand Lux Cafe, Venetian (5-6). RSVP here.
🍸 Fundbox After Hours | 10/27 | 7:30pm–9:30pm PT
Come for the conversation on the future of embedded finance and small business lending. Stay for the one-on-one conversations (over drinks and appetizers, of course!)
FINTECH TAKES: BUILDERS SUMMIT
As you may know, Fintech Takes is hosting our first-ever in-person event on November 12th and 13th in the mountains outside Bozeman, Montana.
The Fintech Takes: Builders Summit is the industry event that I’ve always wanted, but have never quite been able to find. We are bringing together experienced founders and operators from banking and fintech — the folks who are actually building products in our industry — and giving them the content and networking opportunities they need to find (and understand) the next big problem they are going to tackle.
If that sounds like something you’d be interested in participating in, apply to attend or hit reply to this email to get more information on sponsorship opportunities. We still have room, but it is going fast!
Thanks for the read! Let me know what you thought by replying back to this email.
— Alex
