Jay Mok is a payments product leader whose work at PayPal spans card-processing strategy and agent authorization for payments. His work connects the commercial demands of accepting payments—conversion, processing costs, and fraud prevention—with the question of how software can spend on a person’s behalf under clearly defined permission.
From card processing to agent permissions
PayPal’s Evolving Payments ebook identifies Mok as director of product management and credits him and Jon Krauss with its card-processing chapter. Their argument treats payment processing as a set of connected business decisions: improving conversion, managing processing costs, using fraud tools, coordinating processing through orchestration, and developing card-network relationships. A successful payment strategy must balance the ability to complete purchases with the economics and risks of accepting them. Their account gives orchestration and network relationships a place alongside conversion and fraud management, rather than reducing processing to a single transaction fee.
Mok’s work on agentic payments brings another decision into that picture: what a buyer has actually permitted software to do. At the time of Your Agent Just Authorized What?!, which he co-presented with Ben Coumes, Mok was a product manager in agentic payments at PayPal. His contribution centers on matching authorization controls to the consequences of an action and the relationships between the parties involved.
Consent, scope, and the consequences of spending
Human consent, scoped permission, and later proof. Mok frames authorization around three distinct questions: whether the human authorized an action, whether it is allowed within the present scope, and whether that permission can be demonstrated afterward. For payments, the answers can involve human authentication, an expiry time, a spending amount, a specified merchant, and purchasing instructions. Establishing who the user is does not, by itself, settle what the agent may buy. Evidence of permission also matters when a transaction is disputed.
Trust changes the evidence required. Mok distinguishes actions by their stakes and by whether the participants operate within a known ecosystem. Reversible coding changes can often be governed through tool permissions, investigated through logs, and undone. Money movement raises the stakes, but participants within a shared payment ecosystem can rely on its provider to enforce permissions and retain transaction records. His office-building analogy makes the relationship concrete: people admitted through common entrance controls have a shared basis for trust; strangers meeting outside need another way to establish it.
A payment vault is only part of the system. Mok’s travel-data example considers a provider selling occupancy information or reviews to buyer agents. PayPal and Braintree infrastructure can hold payment credentials, while OAuth scopes govern access to them. The partner-operated ecosystem combines credential storage with scoped access and payment mandates; transaction logs support subsequent investigation. The distinction matters because storing a means of payment and granting permission to use it are separate responsibilities. It also places agent payments in a broader commercial setting: agents may purchase useful information as well as physical goods.
Across these contributions, Mok approaches payments as a product-design problem involving both commercial performance and accountable action. Card processing asks how a purchase can succeed economically and safely. Agentic payments adds the need to preserve the buyer’s instructions as software takes a more active role in making that purchase.
Jay Mok and Ben Coumes of PayPal build an authorization ladder from reversible coding work to autonomous payments, showing how consent, scope, and evidence must strengthen as actions become harder to undo and counterparties become less familiar.
Agent authorization must answer three separate questions: whether the human consented, whether the action remains inside its current scope, and whether that authorization can be proved later.
Use lightweight permissions and ordinary logs when actions are reversible and participants share a trusted boundary; stronger proof should earn its complexity by addressing higher consequences or unfamiliar counterparties.
A shared vault plus OAuth scopes can support machine payments inside a known ecosystem because the common operator enforces mandates and retains transaction history.
For autonomous transactions among unfamiliar parties, layered signatures and selective disclosure separate credentialing, human instructions, agent action, checkout verification, and payment verification.
PayPal’s approval-token example reverses the usual order: the human approves constrained instructions first, then the agent searches and transacts within the amount, expiry, and merchant limits.
The ladder extends beyond payments to any consequential, hard-to-reverse agent action, including medical orders, electronic signatures, and securities trading.
