

Internal control. Banking automation. Auditable from the first day, not from the audit.
Ophan gives banks and asset managers a complete internal control system: one reference set where risks, processes, controls, evidence, residual risk and remediation finally sit together.
Alongside it, we automate the banking workflows that should no longer be manual: choosing the level of automation each one actually warrants, and building the control into it rather than around it.
Two different problems. One operating principle: the machine proposes, a person decides and signs, and every step leaves its own evidence.
Control that is scattered. Work that should not still be manual.
Fragmentation
Spreadsheets, mailboxes, shared folders. Each entity keeps its own version of the truth, and consolidating a single view of controls takes weeks.
Traceability
Who ran which control, on what sample, with what evidence: the answer is assembled when the question is asked, not before it.
Coverage
Risks with no control, controls attached to no risk, directives linked to nothing. The gap surfaces during the inspection rather than ahead of it.
Manual load
Documents read by hand, reconciliations done twice, information rebuilt from systems that already hold it. It rarely appears as a line in the budget, which is exactly why it survives every cost programme.
The two are connected. Whatever is automated becomes something that has to be controlled, and most of what is deployed today was designed to work, not to be inspected.
Start with the control system. Or start with the workflow.
Ophana complete internal control system
One reference set connecting risks, processes, controls, evidence, residual risk, incidents and remediation. Built for institutions that have outgrown spreadsheets, shared folders and a GRC tool nobody opens.
The platform is built. A pilot applies it to your institution, on your own controls.
Explore Ophan →Banking automationremove the manual work, not the control
We take one workflow, establish what it costs today, and determine the level of automation it actually warrants. Sometimes that is a rule. Sometimes an integration. Sometimes machine assistance. Sometimes an agent.
The objective is not maximum automation. It is the most useful automation that remains explainable to your auditor.
Bring us a workflow →The machine proposes. A person decides and signs.
Nothing here removes a signature. Five rules govern the design: of Ophan, and of every workflow we automate. They are not negotiable in it.
Independence
Actor, controller and auditor stay three separate roles, whether they are held by people, by software or by agents. They are never collapsed into one.
No unilateral authority over an irreversible act
Four eyes on anything that moves money, alters a critical record or files with a regulator: a person as the second pair, and a second independent agent taking that seat only where the record supports it.
Evidence produced as the work runs
The record is written while the work executes, not reconstructed when audit arrives. Every relevant step captures what was asked, what was retrieved, what was proposed, which rules applied and who decided.
The human is a failure mode too
An override, an exception, a manual decision: each is an event. It is recorded, attributed and reviewed like any other.
Data stays inside the agreed perimeter
Inside your environment, or Swiss-hosted under your contract. Client information is not treated as training material. Your compliance function asks this first; we answer it before being asked.
No authority certifies an internal control system, and we claim no approval. What we can say is this: the design has been walked through with supervisory representatives in more than one jurisdiction, and what came back is the reason we are building it rather than writing about it.
Controls before capability. A process that cannot be explained to your auditor is not properly automated: it is a liability that has not been priced yet.
Four stages. One chain of evidence.
Generic GRC answers this with its own model and asks the bank to fit it. Ophan starts from the banking practice: fifteen years of internal control, put into a system. The framework, its execution, the residual risk and the remediation live in one place: no separate truth for Risk, Internal Control, Audit and the business.
Define
A three-level risk taxonomy, the process catalogue, directives and procedures, risk appetite, regulatory watch.
Execute
Controls dated, assigned and sampled. Reminders on each location's calendar. Evidence attached to every execution.
Assess
Testing campaigns, self-assessment by process owners, thematic reviews run across entities.
Remediate
Action points, audit findings and incidents tracked to closure, then fed back into the next assessment.
Three screens, three uses.
Built for the questions Risk, Audit and IT will ask.
Swiss hosting
Dedicated Swiss infrastructure. Control data partitioned by institution.
Complete audit logs
Every change and every rights assignment is recorded. Nothing is erased: removed items are deactivated and reversible.
Granular rights and roles
Creation, modification and deletion defined role by role: QA, risk management, regulatory watch, administration.
Migration of what exists
File import with field mapping. Several hundred existing controls carried across in a single operation.
System connectors
Test results produced by your monitoring tools are collected automatically and assessed against your own control definitions.
Contractual continuity
A source-code escrow clause is open to negotiation, so the control system survives whatever happens to us.
Machine assistance inside Ophan
Ophan proposes a first set of risks and controls from a described process, flags regulatory changes that may be relevant and translates them into implementation options, generates the automation code for a repetitive control test, and surfaces the gaps: a risk with no control, a control with no evidence.
Every proposal goes through a specialist. Validating, rejecting or adapting a proposal, judging whether a regulatory change is relevant, weighing inherent risk against the context of the process, closing an action point on evidence: those stay human. Assistance is offered where it frees specialist time, never as a matter of principle.
How an Ophan pilot works
Four steps, in orderInstitution profile
Entities, departments, location calendars, risk scales and appetite. Users, roles and permissions imported.
Reference sets
Your risk taxonomy and your directives aligned to the reference base, then your existing controls imported.
Pilot processes
Two or three critical processes modelled end to end with their actors and steps, controls attached, responsibilities assigned.
Going live
A real control campaign, real evidence, real reporting, teams trained. The system is running.
A typical pilot runs six weeks on one entity, two to three processes, up to around twenty-five named users. Those figures describe the usual shape, not a package: yours are set during scoping, against your perimeter.
At the end of the pilot you either move to licence with your reference set already loaded, or you stop, and you leave with your risk and control map, built and yours to keep. The pilot fee is credited against the first licence year.
A smaller institution is the better starting point, not the harder one: a readable perimeter, decisions taken quickly, a result you can show.
Automate what should no longer be manual.
Banks still carry a great deal of operational work through email, spreadsheets, document review, manual reconciliation and the recurring preparation of information that already exists somewhere else. The burden is not spread evenly, and three places carry most of it.
Client lifecycle
Onboarding and periodic review: volume, heavy manual reading, and a fully auditable outcome required at the end of it. Retrieval, extraction, completeness checks and a drafted proposal: the file arrives prepared, and is opened, decided and signed by the person who owns it.
Compliance and transaction monitoring
The alert is not the problem; the ratio is. Assistance works on triage and on the preparation of the case, so specialist time goes to the alerts that deserve it and the reasoning behind each disposition is recorded rather than remembered.
Audit and regulatory reporting
Recurring cycles rebuilt by hand from systems that already hold the answer. Where the evidence is produced as the work runs, the report becomes a view of it rather than a document reassembled every quarter.
The same approach applies elsewhere: front office and client file preparation, operations and reconciliations, finance and management reporting, legal and governance, risk and internal control. The method does not change with the department; only the level of automation does.
The point is not to automate everything. It is to find the workflows where automation creates enough value to justify the control burden it brings with it. We would rather build one useful workflow than present ten hypothetical ones.
Use the simplest technology that can reliably do the job.
Rules and deterministic automation
A threshold, a reconciliation, a scheduled check, a calculation. Predictable, explainable, trivially testable, and still the largest single source of recovered time in most institutions. If a rule can do it, a model should not.
Reading what is unstructured
Documents, correspondence, statements, contracts. Extraction with the source passage attached to every field, so a reviewer can check what was read against where it came from.
Assisted judgement
Retrieval, comparison against your own directives, a drafted recommendation with its reasoning exposed. The file arrives prepared. The person still decides.
Agentic workflows
Several steps, several systems, an objective rather than a fixed script: gather the file, query the sources, flag the discrepancies, prepare the next action. The value is higher on long processes, and so is the control requirement, which is why this level is reached on evidence, not by default.
How an engagement starts
The RAIL methodWe run this on a method we call RAIL:
Reframe the workflow, its economics, its failure points and the controls it requires;
Activate one governed solution with real users and real operating conditions;
Integrate it beside the existing process and reconcile every outcome before anything critical is replaced;
Learn and scale: transfer the capability to your team and expand only where the evidence supports it.
Discoveryaround two weeks
Workflow selection, value case, process mapping, control design, technical approach, decision framework. Fixed fee.
Buildeight to twelve weeks
A working solution with real users, documentation and operational evidence, reconciled against the existing process. Fixed fee, named team.
Scale decision
Expand, adapt or stop: on the result, not on commitment to a technology. You own the outputs, the documentation and the models.
Data architecture, security, resilience testing and core integration are delivered with specialist partners, under our coordination and our contract.

Technology is only useful if it survives contact with the institution.
The hard part of banking automation is rarely the model. It is the process, the controls, the people, the legacy systems, the regulator, and how much change an organisation can realistically absorb in a year.
Our partners have led transactions on financial institutions and have run a listed technology services group through a full operational and financial restructuring. We know what legacy actually costs, what a regulator actually asks, and what an internal team can actually take on.
Ophan is not a framework we drew and then looked for someone to apply it to. It is fifteen years of internal control practice in banking, built into software because the spreadsheet version of it stopped scaling.
That is why we start with one entity or one workflow rather than a transformation roadmap, and why every engagement contains an explicit decision point where the answer can still be stop.
Show us what you need to control, or what you want to stop doing by hand.
If it is the control system, we review your current setup and define a pilot perimeter that fits your institution. If it is a workflow, bring us one that is manual, repetitive, document-heavy or impossible to scale, and we will tell you whether it should be automated, how far, and what controls it requires.
Either way, the first conversation produces something written down.