Firefly RiskOS · Risk and Actuarial Operations

We run the close cycle through to your experts’ sign-off.

Every cycle, we prepare the data, coordinate runs on your approved models, reconcile the results, and assemble the review pack.

RiskOS · Risk and Actuarial Operations, as a line of five stages. Prepare: deterministic code applies your approved rules. Run: the work is done in your systems. Reconcile: an AI agent works the case within the limits you approve. Review: your team decides. Deliver: deterministic code applies your approved rules.
Unit of work · A completed close or model cycleScroll to explore

The problem

The ECL or reserve figure is the easy part.

Then finance and the reviewers ask what went in, what moved, and who signed it, and the answers sit in email.

Who it servesCRO offices, chief actuaries, controllers, and model-management teams

What the service includes

We do the work within the limits of your policy.

Your service schedule lists the tasks your operation includes.

01

Data preparation and controls

We collect the agreed exposure, policy, claims, and finance inputs, run quality checks, and freeze a versioned snapshot.

02

SICR and IFRS 9 staging

Under your criteria for a significant increase in credit risk (SICR), we allocate stages by rule and list each transfer.

03

ECL and CECL runs

We coordinate expected credit loss (ECL) or CECL runs on your PD, LGD, and EAD models and approved scenarios.

04

Reconciliation, movements, and overlays

Once the runs finish, we reconcile results to source and ledger, explain movements by stage, and prepare management overlay proposals for your approvers.

05

Reserving triangles and IBNR

For insurers, we build reserving triangles and prepare incurred-but-not-reported (IBNR) runs on your methods.

06

Model monitoring and IFRS 17 data

We run the back-testing you approve and prepare inputs for your IFRS 17 and Solvency II engines.

How the service works

Five stages, and who does each one.

Each stage runs on the method that fits the cost of an error.

  1. 01

    Prepare

    Approved rules

    We validate and freeze the input set under your approved data-quality rules.

    InputSource data and requirements

    OutputA controlled data snapshot

  2. 02

    Run

    Your systems

    Your approved engine executes the model; we coordinate the run and record its context.

    InputApproved model and parameters

    OutputVersioned model outputs

  3. 03

    Reconcile

    AI agent

    Between runs, an AI agent traces each difference and drafts the explanation.

    InputOutputs and comparisons

    OutputAn analytical review package

  4. 04

    Review

    Your team decides

    Your experts challenge the package and sign it off or ask for a rerun.

    InputPackage and questions

    OutputAn approved result or a rerun

  5. 05

    Deliver

    Approved rules

    We pass the approved figures to finance and reporting under your rules.

    InputApproved figures and evidence

    OutputA completed cycle with its evidence

What you receive

The finished work and the evidence behind it.

Unit of workA completed close or model cycle

  • Controlled inputs and run context
  • Stage allocation, with transfers and exceptions
  • Reconciliation and movement analysis
  • Expert sign-off history and cycle evidence

Authority

The decisions stay with your institution.

You decide

  • Model approval
  • Assumptions and overlays
  • Actuarial judgments
  • Independent validation
  • Final sign-off

We do

  • Data preparation
  • Run coordination
  • Reconciliation
  • Evidence packs
  • Result delivery

Systems and technology

We work in your systems, on open-source modules.

Connected to your systems

Risk and actuarial engines, data warehouses, loan and policy records, claims systems, ledgers, and model inventories.

Built on Firefly OpenPlatform

Your teams can read the code of each module.

Measures

We measure the work you receive.

  • Cycle completion against calendar
  • Input and reconciliation exceptions
  • Reproducibility of run evidence
  • Review rework and unresolved findings
  • Model-monitoring breaches reported on schedule

Scope, pricing, and growth

Start small and expand on evidence.

01

A focused starting point

One recurring cycle on an engine you already use.

02

How the service is priced

Per close cycle, portfolio cycle, or model-monitoring service, with a recurring commitment.

03

Where the service can grow

Add portfolios, reporting cycles, monitoring services, or ReconcileOS.

Questions

What buyers ask before they scope.

Does RiskOS replace our actuarial or risk engine?

No. We coordinate runs on the engines and models you approve.

Is this independent model validation?

No. Independent validation is separate work. Running your models and documenting the runs does not validate them.

Can one service cover banking and insurance?

Yes, as two services. A bank’s IFRS 9 close and an insurer’s reserving cycle each have their own experts, inputs, and acceptance criteria.

Your first operation

Bring us one operation.

Tell us the process, the volume, and what “done” looks like. We will come back with a scope, a deployment plan, and a proposal.