Anticipate losses before they become provisions, fines or headlines.
More than 260 AI agents — 130 deterministic agents that compute and 130 generative companions that read, explain and support — cover every dimension of financial and non-financial risks, built on Brazilian regulation and the international frameworks for risk management and controls.
One base of agents, the whole of risk governance
Traditional consulting delivers a diagnosis and a recommendation — and leaves execution to the client. Ypslon.ai diagnoses, implements the control, runs agent-based monitoring and keeps the regulatory base current. The client does not build a risk-control function: it subscribes to one.
The agents rebuild the risks from the transactional and regulatory data the institution already holds and cannot read. Every finding carries the rule, the data and the threshold that produced it.
Nobody reads millions of rows. Managers receive only what departed from the pattern, already with a likely cause, evidence and a suggested action; the board receives appetite versus actual.
Input, execution and decision in an immutable hash chain, verifiable outside Ypslon. The authority decision is always human, named and dated.
The same finding becomes a queue item for the analyst, a decision for the CRO, an amount in currency for the CEO, an appetite exception for the board and a control deficiency for audit.
The same risk intelligence, in the vocabulary of each sector
The entrance is the same for every client. After login, the portal recognises the sector and presents the dashboards, the agents and the language of that business.
Banks, credit unions, finance companies, asset managers and investment vehicles. 260+ agents covering the Central Bank of Brazil (BCB) and CVM framework, from Resolution 4,557 to Joint Resolution 18, with regulatory reporting checked before it leaves.
How it works for your institution →The same backbone — diagnosis, implementation, management and maintenance by agents — applied to companies that operate at scale and need control by exception: shrink, master data and taxes, supply chain, people, consumer and privacy, in the language and regulation of your business.
How it works for your company →Built on Brazilian regulation and international standards
Every agent is born from a rule or a principle: what it checks, the threshold it uses and the evidence it demands come from the framework below — and are cited in the finding itself.
Every dimension of financial and non-financial risks
Ten groups of agents cover a financial institution's entire taxonomy — from capital to climate — and talk to each other: the same fact is seen by credit risks, by accounting and by regulatory reporting.
Credit and provisioning · market and VaR · liquidity (LCR, NSFR) · interest rate in the banking book (IRRBB) · capital and Basel · concentration · contagion · forward-looking (scenarios, monetary transmission, credit cycles by sector) · COSIF trial balance · investment vehicles (FIDC, FII, FIP, FIF)
Operational and continuity · compliance and AML/CFT · conduct and reputation · third parties and outsourcing · cyber · models and AI · legal and litigation · fraud · privacy and LGPD · social, environmental and climate · governance and related parties · tax · IT, technology and data
Financial risks
Operational integrity
Third parties and cyber
Compliance and controls
Model, legal and resilience
Fraud, privacy, ESG and governance
Forward-looking
Tax
COSIF trial balance
Investment vehicles (CVM)
Reporting to the BCB
Orchestration and ingestion
Data, technology and Joint Resolution 18/2025
The quality of information reported to the Central Bank (Joint Resolution 18) is one component of the data and technology risk group — alongside IT, cyber and AI — and earned a chapter of its own because of its deadline and regulatory weight: 73 agents and 25 companions check the CADOC documents, SPED and SICOR across 12 dimensions and 45 requirements, draft the Article 9 communications and protect the submission with a human decision.
It is the most urgent way in; Ypslon.ai is the whole house.
IT governance, continuity, change, obsolescence
quality of reported information, BCBS 239, lineage
exposed surface, vulnerabilities, response, Resolution 4,893
inventory, validation, drift, responsible use — NIST AI RMF and ISO 42001
What we read today, in three minutes
Equities, rates, FX, commodities and indicators in 3 pages, with a practical reading of each block.
Sources cited in the original language; nothing is summarised without the source having been read.
What changed at the BCB, CVM, Anvisa, ANPD and Federal Revenue, with an impact assessment by risk group.
Built to the highest security standards
Processing in São Paulo (sa-east-1), the client's own encryption key, no data leaves the client's account; in the Enclave model, not even Ypslon sees the data.
Login on our own domain with an authenticator; every access, request and decision recorded with author, date and immutable hash.
Deterministic agents: same input, same result. Generative AI only reads and labels; it never computes and never decides.
See the portal with your own data
A diagnosis on your institution's public data in days, not months — and, if it makes sense, a pilot with your internal data in a dedicated account.