01
Loss at machine speed
Autonomous systems can initiate irreversible transfers faster than human oversight can intervene.
FINTECHLAB / AGENTIC FINANCE RESEARCH
Agentic AI in Finance
We research, build and stress-test autonomous financial systems.
RESEARCH · BUILD · SIMULATE · ATTACK · MEASURE · DEFEND
Why now
When AI can move money, the risk is no longer informational.
It is financial. Machine-speed. And poorly governed.
02
Authority without identity
Financial rails were built for people and institutions - not for non-human actors with delegated power.
03
Agent vs agent conflict
Offense and defense increasingly operate as autonomous systems. Institutions need research on both sides.
FinTechLab studies this transition before autonomous finance becomes infrastructure.
Programme map
Agentic finance ecosystem.
Eight connected acts. Tap a section to jump.
TAP A SECTION TO JUMP
The transition
From AI assistance to financial autonomy.
We study what happens when AI stops advising and starts acting on money.
SWIPE
Main research question
What happens when AI is authorized to act on money?
Traditional
- Analyze
- Recommend
- Human decides
Agentic
- Observe
- Reason
- Plan
- Act
- Verify
How we work
We study both sides: how autonomous financial systems work - and how they fail.
Authorized adversarial research in controlled environments. ATTACK means red-team evaluation, not offensive capability transfer.
SWIPE
01 / RESEARCH
Study emerging agent architectures, financial workflows and systemic risks.
Research architecture
An autonomous financial system.
From human intent to money movement - with authority, risk controls and audit as first-class layers.
Priority fronts
Where FinTechLab focuses first.
Six priority fronts. Full programme remains available below.
- Agent identity & ownership
- Model / provider identity
- Credentials & provenance
- Persistent agent identity
- Trusted execution
Know Your Agent - KYA
KYC, KYB and KYT are not enough. Autonomous finance introduces KYA - Know Your Agent.
- Financial mandates
- Permission boundaries
- Transaction & time limits
- Category & asset restrictions
- Approval thresholds & revocation
- Delegation between agents
Core question
What exactly should an AI agent be allowed to do with money?
- Owner & model provider
- Permissions & autonomy level
- Financial history
- Behavioral deviation
- Tool access & external agents
- Compromise indicators
Agent Risk Score
A continuous risk assessment for autonomous financial actors.
- Autonomous scam agents
- Agent-assisted social engineering
- Compromised financial agents
- Malicious merchant agents
- Defense agents & intervention
Financial Action Risk
Intent × Identity × Behavior × Agent × Context × Transaction → ALLOW / CHALLENGE / LIMIT / DELAY / BLOCK / ESCALATE
- Evidence & transaction analysis agents
- Relationship / graph agents
- Typology & risk agents
- Case agents + human investigator
- Continuous KYC
- Liquidity & cash positioning
- FX & hedging
- Working capital
- Intercompany transfers
- Yield & counterparty exposure
Autonomous cycle
Sense → Forecast → Decide → Execute → Verify → Audit
Threat surface
How autonomous financial systems fail.
We study financial action risk, agent-versus-agent conflict, and authorized red-team failure modes - in controlled environments.
Should this financial action happen at all? Intent, identity, behavior, agent, context and transaction decide ALLOW / CHALLENGE / LIMIT / DELAY / BLOCK / ESCALATE.
- ALLOW
- CHALLENGE
- LIMIT
- DELAY
- BLOCK
- ESCALATE
Understand how financial agents fail before attackers do.
Horizon
What we expect to emerge.
Research hypotheses and emerging directions - not claims that every product already exists.
- Agent-enabled banking
- Financial MCP interfaces
- Agentic fraud investigation
- AI payment credentials
- Agent wallets
- Agentic treasury pilots
- Autonomous finance operations
Work with the lab
We do not study agents only when they work.
We study them when they fail.
Autonomous finance is becoming infrastructure.
We want partners who will study it before that happens.
Who this is for
Banks & PSPs
Agent identity, delegated authority, loss prevention, runtime controls.
Security & risk
Red-team research, agent risk scoring, financial action firewalls.
Regulators & policy
Observability, replay, liability boundaries, governance models.
Universities & partners
Joint research, simulation environments, publication tracks.
legal@ftlab.space