Rethinking Take Rates in the AI Agent Economy

Take rates are the hidden prize of commerce. Every merchant pays them, whether they notice or not. Credit card networks skim them. Marketplaces demand them. SaaS platforms bundle them into subscriptions and services. The gas station in town with the lowest prices is cash only because they don’t want to pay them. Now, it’s very likely that LLMs like ChatGPT and Perplexity will move into the same territory. It’s a $67 billion commerce opportunity that’s hiding in plain sight.

As AI agents begin to handle more of the buying process, take rates will remain the primary exchange of value in digital commerce. Let’s take a look at why this is the case.

Ecommerce is one of the few markets measured in trillions. Global B2B e-commerce is projected to reach $36.2 trillion by 2026. Global B2C e-commerce will climb to $5.5 trillion by 2027. These numbers explain why so many technology companies’ business models focus on transactions rather than subscriptions or advertising.

Case in point: Apple Pay generated about $9.4 billion in 2025, or 3.4% of Apple’s total revenue (while modest for Apple, this is a number still larger than many standalone companies). To put this in perspective, Shopify makes three-quarters of its revenue from payments. Toast earns about 70% of its revenue from merchant services.

And now, with OpenAI introducing Instant Checkout in ChatGPT, the playbook for take rates in AI commerce is already being written. Currently, Instant Checkout is monetizing with affiliate fees. But, when AI agents start facilitating more transactions, why wouldn’t take rates be the top way to monetize?

Marketplaces: The Original Take-Rate Engines

Amazon, eBay, and Apple’s App Store built their businesses by taking a cut of every sale. Fees of 10 to 30 percent became standard. In return, merchants gained access to customers, trust, and logistics. While this arrangement worked, it left merchants dependent on platforms that steadily eroded their margins.

Companies like Shopify and Toast updated the playbook. While they provided storefront tools and point-of-sale systems, the software was never the real revenue driver. Payments were. Instead, hardware and software were the value exchange for merchants, with the subsidy to capture the payments take rate. For Toast and Shopify, the real business model is and always has been tied to the money moving through the system.

Credit card companies wrote the earliest version of this story. Interchange fees, invisible to buyers but costly for sellers, built global financial giants. Apple extended that model through Apple Pay. Payments remain the most consistent place to extract value in commerce.

The AI Shift: From Ads to Transactions

Now, all eyes are on AI agents and LLMs. We’ve seen how quickly AI is changing behaviors in search. It took Google 11 years to hit 365 billion annual searches, while it took ChatGPT only two years.

But search and answer engine business models are not apples-to-apples. For years, Google defined digital monetization with advertising. A buyer searched, sellers bid, and Google collected its spread.

AI agents change this pattern since they don’t deliver a list of links. They deliver a decision, along with all of the information a buyer needs to make it. Whether a buyer is B2B or B2C, they ask a question while the agent guides them directly to a purchase. The moment of monetization is the transaction itself. That means referral fees and take rates will replace sponsored results. It is the most direct path to product revenue at scale for LLMs. This business model is already taking shape.

Tools like OpenAI’s Instant Checkout and Perplexity’s Buy With Pro make it possible to move from “What are the best running shoes under $100?” to a completed purchase without ever leaving the chat. It’s an early but clear signal that agent-driven transactions will monetize through take rates, not ads, while preserving the brand’s position as merchant of record.

Commerce Platforms as Systems of Work

Here’s where the commerce platform fits in. AI agents can’t transact without reliable inputs. They need structured catalogs, accurate pricing, promotions and bundle options, inventory visibility, and compliance rules.

When the storefront is no longer the center of the experience, the commerce platform will have to act as a system of work — an orchestration layer that ensures data flows cleanly between buyers, sellers, and agents. It coordinates the transaction so that agents can complete it without friction, and so sellers can maintain visibility and control.

The value of the commerce platform becomes less about design templates and more about ensuring every agent, marketplace, or social commerce platform can work with accurate, trustworthy data. The risk of not using a commerce platform prepared for this future is clear. Merchants will have little leverage over their visibility and results in LLMs. Commerce platforms should give merchants the ability to connect directly with agents, maintain ownership of their data, and ensure they’re “found” by the right buyers at the right time.

Preparing for Agent Buyers

Apple, Shopify, and Toast have already demonstrated the fastest way to scale into the billions: take a small cut of every transaction, make it up in volume. AI platforms are highly likely to follow the same route. The economics are too compelling to ignore.

That makes the evolution of commerce platforms urgent. The storefront era is ending. The system of work era is here. Platforms that can coordinate agent-driven transactions, keep data accurate, and preserve merchant control will define the next chapter of digital commerce.

Take rates will continue to shape the flow of money. The real question is whether merchants participate in shaping the system, or whether they remain invisible to an entirely new ecosystem of machine buyers.