Assay / Solutions / Financial services
Industry · Financial services

Agent Trust for Financial Services

Financial institutions have decades of third-party risk discipline and none of it was built for a counterparty that is an autonomous agent created last week. Assay supplies the missing control point.


The problem

Third-party risk frameworks assume slow-moving vendors

60%
of enterprises are scaling AI agents
4%
have agent governance in place
97%
reported an agent identity incident

Source: 2025 industry surveys of enterprise AI agent adoption.

Vendor due diligence, ongoing monitoring and concentration analysis all assume a counterparty that changes on the scale of quarters. Agents change on the scale of deployments, and a single vendor may expose dozens of them with different capabilities and different reliability.

The controls that exist govern the vendor relationship. They do not reach the individual agent making an individual decision, so the granularity of the control does not match the granularity of the risk.

Regulators increasingly ask who or what made a decision and on what basis. “A partner's agent, authenticated by API key” is an answer that invites the next question.


The solution

Per-agent, per-capability, per-interaction control

Assay moves the control point from the vendor to the agent and from onboarding to the interaction. Every agent carries an identity and a trust score earned from counterparty-verified outcomes, and every delegation can be gated on both.

The financial_services preset tunes the model for the sector's risk posture. Because capabilities score independently, an agent proven at reconciliation is not thereby trusted with settlement instructions.

Every decision produces a record: what was verified, what score it carried, which capabilities were confirmed, and what outcome followed. Controls testing gets a data source instead of a sample of screenshots.

financial_services presetper-agent controlper-capability gatingdurable decision recordsorg isolationrevocation

Why it holds up

What you actually get

Control at the right granularity

Per agent and per capability, not per vendor. The control finally matches the shape of the risk.

Threshold scales with the action

A 0.5 bar for a read-only lookup and 0.75 for anything that moves money — set per call, not per relationship.

Durable decision records

Every gate produces evidence of what was checked and what was relied upon. Useful well beyond the moment of the decision.

Immediate revocation

Revoke an agent and it fails verification everywhere at once, with no propagation window to explain to a regulator.


Live API

Gate a counterparty agent at a high bar

Set the threshold to match the risk of the action, not the vendor.

# 2. Verify it before you delegate. This is the allow/deny call. curl -s https://assay.foundrynet.io/v1/verify \ -H 'X-API-Key: YOUR_ASSAY_KEY' \ -H 'Content-Type: application/json' \ -d '{"agent_id":"agt_4411f105…", "claimed_capabilities":["reconciliation", "exception_handling"], "min_trust_threshold":0.75}' { "verified": true, "trust_score": 0.8, "capabilities_confirmed": ["reconciliation", "exception_handling"], "attestation_count": 8, "last_attestation_age_seconds": 41 } # Gate on BOTH fields. `verified` reflects the trust threshold and account # status only — an unproven capability comes back absent from # capabilities_confirmed, not as verified:false. if verified and "reconciliation" in capabilities_confirmed: dispatch() # Public. Every industry trust model and its tunable parameters. curl -s https://assay.foundrynet.io/v1/admin/trust-presets { "presets": { "manufacturing": { "recency_half_life_days": 30, "failure_penalty": 2.5, // harsher than default 2.0 "counterparty_bonus": 1.5, "verification_ceiling": 0.75, "min_verification_rate": 0.15, "velocity_min_gap_seconds": 300, "velocity_discount": 0.8, "volume_k": 8 }, "pharma": { … }, "financial_services": { … }, "logistics": { … }, "energy": { … }, "default": { … } } }
Read the gate carefully. verified reflects the trust threshold and account status. It does not fall to false because a claimed capability is unproven — an unproven capability simply does not appear in capabilities_confirmed. Gate on both fields.

Questions

Frequently asked

Does this satisfy a specific regulation?

No single control satisfies a regulation, and we will not pretend otherwise. Assay provides identity, evidenced competence and durable decision records that support third-party risk and audit obligations. Mapping to a specific rule is your risk function's call.

Can we run it on-premises?

Enterprise plans include an on-premises option along with custom trust models, SSO and SAML, and private namespaces.

What data leaves our environment?

For a verify call: the agent ID you are asking about and your key. For an attestation: a task type, an outcome, an optional counterparty ID and an optional hash you compute. No transaction data, no customer data.

How do we handle concentration risk across agents?

Discovery lets you enumerate registered agents by capability and trust, so you can see how many independent providers actually clear your bar for a given capability rather than assuming the vendor count is the provider count.


Related

Keep reading