Announcing US$ 3.2M pre-seed round · OneVC · Maya · Norte Ventures Read →
Company Brain for Financial Services

Every number has a source. Not everyone may see it.

Core banking, CRM, contracts, tickets and policies each hold a piece of the same customer. Strattum resolves that as one governed context inside your cloud, with permissions applied before the answer and a trace of how it was built.

Better operations under a regulator, with the right context

Compliance applied at the source of every query

The permission filter runs before ranking, and every answer keeps the trace of which data entered it. Sensitive data never leaves your cloud, and the audit stops being a reconstruction.

The same person in every system, and across networks

Customer, account, contract, device and contact resolved as one entity. It is the same resolution that serves support and that exposes the ring of accounts sharing a phone and an IP.

Credit decisions with the whole path attached

Every recommendation arrives with which data was read, which rule applied and what came out. The auditor’s request for a trace stops becoming a project.

Four everyday questions,
and the answer with its source attached.

Each of them today ends in a request to the data team, a query to legal, or a manual reconstruction for the auditor.

Case 01 · Policy and authority

Which rule applies to this customer

The credit policy, the channel’s approval authority and the segment exception arrive as one answer, in the version in force on the date of the application. The query stops depending on who knows the manual.

  • Policy, authority and exception in one context
  • The version quoted is the one in force on the application date
  • The same answer in branch, in digital and on the desk
Case 02 · Fraud

The network behind the account

Account, device and contact resolved as a graph. The pattern that only appears when rows are linked to each other stops requiring a hand-written query.

  • Account, device and contact resolved as a network
  • The shared link appears beside the account that produced it
  • The pattern is found by connection between records
Case 03 · Audit and traceability

The trace the auditor asked for, assembled

Which data entered the decision, which policy applied and what came out, decision by decision. The traceability request stops becoming a multi-week project for the data team.

  • Assembled from each decision’s log
  • Every decision carries data, rule and answer
  • Serves internal audit and on-site examination alike
Case 04 · Portfolio and delinquency

The vintage that broke pattern

Origination, delinquency and channel read together by vintage. When delinquency rises in one specific vintage, the factor arrives tied to the data that exposes it.

  • Origination, delinquency and channel in one context
  • The critical vintage is pointed at as soon as it appears
  • Every number carries the system it came from

Frequently asked questions

How does this meet regulatory and data-protection requirements?

The entire platform runs inside your cloud, so sensitive data never leaves it. Source permissions are preserved and the filter is applied before ranking. Every decision keeps the trace of which data entered, which policy applied and what was answered, which is exactly what an examination asks for. DPA on day one, SOC 2 Type II in progress. Details on /governance.

How does the fraud side work if we already have risk rules?

The layer does not replace your rules, it gives them context. What the Memory Graph adds is the link: the same person under several identities, accounts sharing a device, a phone and an IP, mule networks. Network patterns do not show up in a table row, and that is what rules alone cannot see.

Do we have to replace the core banking system?

No. There are connectors for proprietary cores, FIS, Sopra and TOTVS CRM/Banking, plus Sankhya and Salesforce. Strattum reads from where the data already is, and bespoke connectors ship in days.

How is a decision taken with agent support explained afterwards?

Every recommendation arrives with the path: which data was read, with origin and timestamp, which policy applied in the version in force on that date, and what came out. That is recorded at the moment of the decision, which is the difference between delivering in minutes and opening a project.

How is this different from plugging in Copilot or a custom GPT?

A generic assistant answers from whatever it can search, and in a regulated environment that means confidently citing a repealed policy and reading data that user was not allowed to see. What is missing is not the model, it is the governed context: which version applied, which account belongs to that customer, what each role is entitled to see.

Where does the data live?

Inside your cloud. The entire platform is deployed in your infrastructure: AWS, Azure, GCP, OCI or on-prem. Strattum has no runtime access to the data, and your records, credentials, keys and network access never leave it.

Bring the institution’s context
into every decision.