Sep 25, 2026 in Research
Authors: Oliver Jaros, Associate and Augie Ilag, Partner
AI is moving from generating answers to taking actions. In commerce, that means agents will increasingly search live inventory, compare options, complete purchases, and manage what happens after checkout on a user’s behalf.
This shift creates a trust problem that today’s commerce stack was not designed to solve. Existing systems can verify that a payment occurred, but they cannot reliably prove that the transaction matched what the user originally asked the agent to do. TrustPixel is building the infrastructure to close that gap.
The Missing Layer Between Intent and Settlement
Today’s commerce systems assume that a person is navigating a website, choosing a product, and approving a purchase. When those decisions are delegated to an AI agent, the payment may look the same, but the path leading to it is fundamentally different.
Consider a user asking an agent to book a refundable flight under $600, or an employee instructing an agent to purchase equipment from an approved supplier. The agent may consider dozens of options before selecting one. A successful payment authorization shows that the payment method worked, but it does not show that the purchase complied with the user’s original instructions.
Merchants, processors, and payment networks currently infer intent after the fact using fraud scores, authorization data, complaints, and disputes. That approach is poorly suited to a world in which delegation becomes a standard part of commerce. The weak point is the “messy middle” between instruction and settlement: what the user requested, what authority the agent received, why it selected a particular option, and whether the final outcome followed the user’s rules.
TrustPixel makes that intent explicit. The company turns a user’s natural-language request into a structured, signed mandate that defines what the agent is authorized to do. The mandate can include limits around price, product category, merchant, payment method, timing, delivery terms, and required approvals.
That mandate is carried through merchant selection, checkout, payment, and post-purchase activity. Rather than trying to reconstruct intent after something goes wrong, TrustPixel preserves a record of what the user wanted, what the agent was permitted to do, and what ultimately happened.
A Team Built for the Problem
Our conviction in TrustPixel starts with the team. CEO Pratyush Tiwari previously co-founded MoneyLion, where he built trust, fraud, payments, and consumer financial infrastructure serving more than 20 million users and processing over $10 billion annually.
Pratyush describes MoneyLion as a laboratory where he learned how financial systems behave and how they break at scale. Those lessons are directly relevant to agent commerce, where the challenge is not simply enabling a transaction but establishing identity, authority, accountability, and trust around it.
He is joined by CTO Jong Char, who previously built KYC, fraud-scoring, and payment systems as a VP of Engineering at MoneyLion. CRO Chris Ashley complements their product and technical experience with more than 15 years in enterprise go-to-market.
The team had already built an MVP before seeking institutional funding, reflecting the urgency and bias toward action we look for in early-stage founders. The market structure for agent commerce is still forming, and the product will almost certainly evolve with it. We believe this team has the experience and adaptability to navigate that uncertainty.
From Marketplace to Infrastructure
TrustPixel is entering the market through its own agent commerce marketplace. The marketplace allows the team to test its infrastructure against live inventory, fragmented merchant catalogs, checkout flows, payment methods, and fulfillment issues rather than designing the product entirely in the abstract.
Within that environment, TrustPixel is building explicit mandates, merchant routing, agent-ready catalog orchestration, and closed-loop transaction tracking. The company’s near-term roadmap includes expanding to millions of SKUs, proving the mandate flow at volume, developing merchant intelligence, and opening its intent APIs to other participants in the commerce stack.
We view the marketplace as the wedge, not the end state. Over time, TrustPixel has the opportunity to provide infrastructure to agents, merchants, processors, and payment networks that need a common way to transport and verify intent.
That infrastructure will need to work across a fragmented ecosystem. We do not expect agent commerce to develop within a single model, marketplace, payment network, or technical protocol. TrustPixel is therefore building a neutral, protocol-agnostic layer that can sit above those systems without requiring merchants to give up control of their inventory, pricing, terms, or customer relationships.
For agent developers, this could provide a common way to access supply and complete purchases. For merchants, it could make inventory available to agent-driven demand while preserving existing commercial relationships. For enterprises, it could enforce budgets, permissions, and approvals before money is spent.
If TrustPixel succeeds, it can become the system of record for what the user wanted, what the agent did, and what actually happened. This would give the company a meaningful control point in a new category of commerce.
Why We Invested
We believe agent-mediated transactions will create an important new layer between user intent, merchant selection, payment execution, and dispute resolution. TrustPixel is one of the few teams building specifically for that “messy middle,” with the founder-market fit and urgency required to shape a market that is still taking form.
The company’s core insight is straightforward: agent commerce should be built around explicit, portable intent rather than systems attempting to infer trust after the fact. We are excited to partner with Pratyush, Jong, Chris, and the TrustPixel team as they build the intent stack for agent commerce.
Disclosures:
For informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. The views expressed herein are those of the author as of the time of writing and may not necessarily represent the views of CMT Digital and its affiliates. Certain information contained in the piece has been obtained from third-party sources, including from portfolio companies of CMT Digital. While taken from sources believed to be reliable, CMT Digital has not independently verified such information.
References to any securities, digital assets, tokens, and/or cryptocurrencies are for illustrative purposes only and do not constitute a recommendation to invest in any such instrument nor do such references constitute an offer to provide investment advisory services. This content is not intended for investors or prospective investors and should not be relied upon when making any investment decision, including a decision to invest in any vehicles managed by CMT Digital. Such offerings are only made via formal offering documents.
Past performance is not indicative of future results. Any projections, estimates, forecasts, and/or opinions expressed in this piece are subject to change without notice.