01
Real core, not a wrapper
A cloud-native ledger, product engine and API layer that runs production banking workloads — accounts, payments, lending, reporting. AI sits on top, not in the middle of the system of record.
Platform / Agentic AI Core Banking
Banking infrastructure for the agentic AI era.
CoreFi is one platform: a real-time core, headless banking APIs and a governed AI control plane. Agents retrieve context, prepare decisions, call banking APIs through policy gates, and post to the ledger only after permissions, audit and the right humans agree.
What changes
From core system to AI operating layer.
CoreFi does not expose banking systems directly to a model. It creates a governed orchestration layer where agents request actions, retrieve context, prepare decisions and escalate exceptions — while the platform enforces permissions, policy and audit. The model proposes; CoreFi controls.
02
Model-agnostic by design
Run governed workflows on ChatGPT, Claude, Gemini, your own hosted models, or a mix. The control plane, audit log and policy gates do not change when you swap the underlying model.
03
Governed, not autonomous
Every agent action passes through role permissions, policy rules, transaction limits and human approval gates before it touches the ledger. Audit captures the model context, the plan and the outcome as one record per case.
How a governed banking agent works
Seven steps from prompt to posted transaction.
Whether the agent is opening an account, deciding a loan, reconciling treasury or investigating an alert, every CoreFi workflow follows the same governed lifecycle. The model proposes; CoreFi enforces; the human approves what the policy says they must.
Sense
The agent receives a banking trigger — a new applicant, an inbound payment, a customer message, a portfolio breach, a scheduled reconciliation — and pulls the relevant context from the core, KYC vendor, document store and ledger through permissioned APIs.
Plan
The agent drafts a structured action plan: which APIs it will call, which limits and rules apply, which evidence supports the decision. It is a machine-readable proposal, not free text.
Check
CoreFi runs the plan through policy gates before any side effect — role permissions, customer consent, transaction and exposure limits, AML and sanctions filters, model-output guardrails, jurisdiction rules. Failed checks stop the workflow.
Act
If checks pass, CoreFi executes the plan through the same APIs a human operator would call — open an account, post a journal entry, release a payment, update a case. Nothing bypasses the core.
Audit
Every step — model context, retrieved data, plan, policy decision, API calls, outcomes — is written to an immutable audit log keyed to the workflow, the customer and the model version. Regulators get one record per case.
Escalate
When a policy gate routes a step to a human, CoreFi prepares the case in the reviewer dashboard with the evidence, the agent's recommendation and the exact action awaiting approval. The human approves, rejects or edits — the workflow resumes from there.
Learn
Outcomes feed back into the workflow: which agent recommendations got accepted, which were overridden, which produced exceptions downstream. Risk and model owners see the signal — the model never gets uncontrolled write access to its own behaviour.
Agent roles
Five agents already wired into the core.
Each role uses the same 7-step lifecycle. The difference is which APIs it can call, which policies apply and which humans approve.
Onboarding Agent
Triages KYC and KYB intake: parses ID documents, extracts beneficial owners, runs sanctions and PEP screens, scores risk indicators and prepares a structured review packet for the queue.
Approval gate: any high-risk classification, edge-case document or sanctions hit routes to a human reviewer.
Lending Agent
Runs credit applications end-to-end: pulls bureau and open-banking data, classifies bank statements, drafts the underwriting memo, suggests a limit and pricing, prepares the offer document.
Approval gate: every credit decision above policy thresholds requires a human underwriter sign-off before the offer is issued.
Treasury Agent
Watches balances, sweeps, FX exposure, payment queues and counterparty limits. Proposes rebalancing actions, flags breaches and prepares end-of-day reconciliation packs.
Approval gate: any movement above defined limits or outside approved counterparties is held for treasurer approval.
Compliance Agent
Investigates AML alerts, transaction-monitoring hits and suspicious-activity cases. Pulls customer history, ranks indicators, drafts the case narrative and prepares the SAR/STR template with evidence.
Approval gate: filing decisions, customer offboarding and high-severity case closures stay with the compliance officer.
Customer Service Agent
Handles inbound questions across chat, email and in-app. Performs read-only actions itself (statements, balances, status) and prepares ticketed write actions for an operator to release.
Approval gate: any monetary action — refund, reversal, fee waiver, limit change — requires a human agent to approve before it is posted.
Banking controls
Permissions, policy gates, audit. Same as any operator.
CoreFi treats the agent like any other operator: it is granted scoped access, it hits policy gates, it leaves an audit trail and it waits for a human when the rule says so.
01
Guardrails
Role-based access, scoped API tokens, transaction and exposure limits, customer-segment rules, jurisdictional restrictions and model-output filters — configurable per workflow, market and customer tier.
02
Approval gates
Workflows declare which steps require a human. CoreFi routes the case to the right reviewer dashboard with full evidence; the decision (approve, reject, edit) is logged and resumes the workflow.
03
Audit trail
One immutable record per workflow: trigger, retrieved data, model and prompt version, plan, policy outcome, API calls, side effects, escalations, human decisions, final state — exportable for internal, external and supervisory review.
The platform around the agent
One platform. Five surfaces an agent can act on.
Core Banking Engine. Real-time ledger, accounts, wallets, products, payments, reporting. The system of record the agent posts into. See the core →
Headless Banking APIs. The same APIs your channels and partners use, scoped per agent, per role and per workflow. See the API layer →
Customer Onboarding. KYC, KYB, AML, documents and consent — the funnel the Onboarding Agent prepares cases inside. See onboarding →
Lending-as-a-Service. Origination, decisioning, servicing, restructuring and collections — the lifecycle the Lending Agent operates in. See LaaS →
Governed AI workflows. The workflow control plane that wires agents, policy gates, audit and approvals across the platform. See AI workflows →
Put a governed banking agent into production.
We will walk you through the 7-step lifecycle on a live workflow, show the audit log on a real case and scope a first deployment together — onboarding, credit or treasury.