Hyperlab is the innovation lab of Consult Hyperion, consulting by Fime. We work at the edges of technology in payments, transit, CBDC, and digital identity. The lab brings together experts across disciplines — data science, test and certification, DevOps, mobile and web development, architecture, and subject-matter expertise in digital identity, EMV, CBDC, stablecoins, payments, agentic systems, transit, crypto, and cybersecurity. We also contribute to international working groups and standards bodies — including FIDO, W3C, and others — to foster interoperability and trust at industry scale.
Our purpose is practical: explore new trends, evaluate market potential, test assumptions and hypotheses, and help the industry build interoperability and trust before scale becomes irreversible.
Our focus on agentic commerce
AI agents are starting to discover merchants, build carts, and pay on behalf of users. That opens real opportunity, and raises open questions about how trust will travel with those journeys.
Hyperlab is exploring that question through two complementary building blocks:
ATLAS (Agent Trust Layer and Assurance Standard) is an open protocol proposal for agent observability at scale: a shared rulebook so independent assessors, agents, merchants, wallets, and payment parties can exchange compact, portable trust signals.
FACT (Framework for Agentic Commerce Trust) is Fime’s working implementation of ATLAS, with operational services and a Hyperlab sandbox where partners can run proofs of concept and evaluate capabilities such as FACT Verify (can we trust this agent action now?) and FACT Evidence (can we prove what was assessed later?).
In short: ATLAS is an open standard we are contributing to, while FACT is how we make the idea operational, explored and demonstrated in Hyperlab, delivered with Consult Hyperion and Fime.
A research angle we are exploring
Agentic commerce is emerging on the buy side. Networks and banks are enabling agents to initiate payments, and important complementary work is already underway:
- KYA / KYA-OS helps establish who the agent is and its delegated scope.
- AP2 helps bind the checkout: consent, mandate, cart integrity, and payment instrument.
- Verifiable Intent (VI) helps cryptographically prove the user’s captured intent and constraints, including limits such as budget.
Our working assumption is not that these approaches are incomplete or incorrect. It is that an additional research question may still be useful to the industry: alongside identity, mandate, and bound intent, could independent observability of agent behavior along the path from initial request to cart strengthen trust — especially for disputes, oversight, and liability design?
Early industry commentary on agentic disputes and fraud — for example FraudBeat — suggests that when consumers delegate purchasing to an AI, the hard cases may involve interpretation and outcome (“I authorized the agent — not that specific result”), and that evidence practices for those cases are still maturing. Reported early signals of higher dispute rates on some agent-initiated flows are, for us, a prompt for research rather than a settled diagnosis.
Under that assumption, the hypotheses we are testing focus on the most painful points where a lack of observability can lead to bad outcomes, wrong or unsafe purchases, unfair selection, compliance risk, and disputes that are hard to adjudicate:
Observability before AP2 / VI (immediate flows). In immediate agentic flows, AP2 and Verifiable Intent typically start after the shopping agent has already interpreted the user’s request and selected merchants and items. What happens earlier — between the user and the agent, and between the agent and merchants during discovery, ranking, and cart building — is often not captured and not independently analyzed. Our hypothesis: that pre-binding phase is where intent can drift and bad actions can take shape, and independent observability there may reduce the risk of binding an already-flawed cart.
KYA verification across repositories. KYA checks often need to happen before a transaction hits a payment rail — and before the rail is even chosen. For merchants that accept several methods, that can mean many lookups per agent visit: complex and heavy. For networks, the pain is different: without a shared minimal signal, Network B may not know that the same agent was already blacklisted or flagged on Network A. Our hypothesis: a lightweight cross-repository trust view may reduce that burden and surface elevated risk earlier.
Compliance verification at transaction time. Many rules assume a human chose the merchant, the product, and the payment instrument. An agent can steer payment-method choice in ways that change cost, routing, or legal treatment. One EU illustration is IFR (Interchange Fee Regulation) and fair payment-method choice: if an agent steers toward a higher-cost path the user did not knowingly select, fairness and scheme economics can be hard to justify. Similar patterns apply to network rulebooks, restricted goods, age limits, and geo/sanctions. Our hypothesis: independent checks at transaction time may help surface compliance risk before authorization.
Dispute and liability evidence when outcomes go wrong. When a purchase is disputed — “I authorized the agent, not that result” — parties need clearer evidence of what the agent did. Our hypothesis: an independent, redeemable assessment record may help schemes, issuers, and merchants reason about agent-mediated disputes with less reliance on unverifiable claims — and may also give payment networks a clearer basis to design and publish liability policies for agent-initiated transactions.
ATLAS and FACT are our contribution to that research path, proposed as complementary to KYA, AP2, VI, and related work, not as a substitute:
- ATLAS is an open protocol for agent observability at scale, aiming to produce compact, portable trust signals the ecosystem can evaluate.
- FACT consists of operational services to test the model in practice: FACT Verify and FACT Evidence, delivered via a licensed Trust Assessor, designed so sensitive conversation need not travel on the payment rail.
Our broader hypothesis is simple: trust infrastructure may be a condition for agentic commerce to scale as a realistic future, and that hypothesis is best tested openly, with partners, in a sandbox, against real use cases.
See it in action
The sandbox is where we turn assumptions into evidence. It can run several agentic commerce flows across emerging protocols —including but not limited to UCP, AP2, VI, KYA-OS, x402, and ACP — with simulators or real components: shopping agent, wallet, token service provider, merchant, payment networks, issuers, and merchant processors. Simulated merchants support multiple integration styles (A2A, MCP, REST) to cover the main patterns seen in emerging implementations. On top of that stack, the sandbox includes the FACT trust layer implementing the ATLAS protocol. Components are plug-and-play: swap a simulator for a real partner component (or the reverse) without rebuilding the journey.
Question for the ecosystem
If agent-led purchases landed on your rails tomorrow, as a merchant, what evidence would you need before accepting an agent-built cart; as a domestic payment scheme, what trust signals would you need to set rules, monitor risk, and design liability for your participants?
How we can help
Consult Hyperion (consulting by Fime) supports organizations across the Agentic Commerce Advisory Framework — from market understanding to large-scale adoption:
- Industry landscape review: Establish a shared executive and technical understanding of the Agentic Commerce landscape, its actors and standards, and implications for payment networks.
- Scenario planning: Explore strategic scenarios and positioning options: under different plausible futures, what should we decide and when? (strategic posture, triggers, no-regret moves, Board-ready roadmap).
- Readiness assessment: Ascertain organizational and technical readiness and maturity, identify gaps, define strategic options, and prepare a KYA design brief.
- KYA Framework / Governance: Design a custom framework for agent governance, delegation, registry, metadata, certification, liability, and dispute resolution — and design the pilot.
- Technical Blueprint – PoC: Define and validate implementation architecture, APIs, sandbox flows, trust services (including FACT/ATLAS), and participant integration.
- Industrialization Support: Support transition to full commercial deployment: rulebook updates, participant onboarding, certification operations, and ecosystem scale-up.
Each phase can stand alone or combine into an integrated program. We would rather test assumptions early than assume consensus later. If you are exploring agentic payments, KYA, or trust infrastructure, we would be glad to compare notes. The ATLAS working draft is public: ATLAS Protocol Specification (0.1 draft).
Hyperlab | Consult Hyperion, consulting by Fime
Discover more in our agentic AI commerce blog series:
Chapter I: Agentic AI and payments: when AI gets a wallet and a will of its own.
Chapter II: Agentic commerce: when your wallet gets a brain.
Chapter III: Agentic commerce: issue on Llamas.
Chapter IV: Rethinking security in the age of Agentic AI.
Chapter V: From emotion to algorithms: why Agentic Commerce needs a new trust layer.
Chapter VI: Closing the trust gap in agentic commerce.
Chapter VII: Trust framework: building verifiable trust for autonomous transactions.
Chapter VIII: From KYA to continuous trust: governing agentic commerce in production with FACT.
Chapter IX: KYA is not enough: introducing FACT, the runtime trust layer for agentic commerce
Chapter X: The agent that cheated at the exam
