Banks, lenders and insurers make the same kinds of decisions thousands of times a day, each one built from data that comes out of a dozen systems. Helixor applies your written policy to that data the same way every time, explains every outcome, and only changes when you approve a new pack version.
These are examples, not a complete list. A decision usually fits when it has written criteria, pulls from several systems, happens often, and someone may later ask why it went the way it did. Pick one to jump to it.
An auto or homeowners application touches a lot of systems before anyone decides on it. The rating engine prices it. A motor vehicle report shows violations and suspensions. A loss history report shows prior claims. A credit-based insurance score comes back, where the state allows one. On the property side there is replacement cost, roof age and material, wildfire and flood scores, and distance to the nearest fire station. Most carriers hold this together with rules spread across the rating engine, the policy admin system and an underwriting manual, so two similar applications can get different answers depending on the path they took.
Helixor doesn't replace your rating engine or your third-party scores. It takes their outputs as typed facts and applies your underwriting guidelines as one versioned pack. Every application gets one of the actions you declared: accept, accept with conditions, refer to an underwriter, or decline. Hard stops are fatal rules, so a great score elsewhere can't paper over them. Files in the gray zone go to an underwriter, or to hosted reasoning, which is allowed to say it isn't sure. As outcomes come back (claims, mid-term cancellations, underwriter overrides) they feed a proposed new version of the pack, and that version is tested against past applications before anyone promotes it.
{
"application_id": "HO-2026-118452",
"receipt": "hx_proof_3f9a1c07b2e84d5a6c1f0e92",
"pack": "acme.homeowners_uw v14",
"action": "refer_underwriter",
"rules": ["UW-ROOF-AGE-03", "UW-WILDFIRE-SCORE-02"],
"input_sha256": "9c1e...a7f4",
"sources": {"loss_history": "LH-77120", "wildfire_score": "vendor-x 2026-09-12"},
"remedy": ["roof_age_years <= 20 or inspection_passed"],
"logged_at": "2026-09-28T14:02:11Z"
}When a claim comes in, the first notice of loss arrives alongside the policy, its coverages and deductibles, prior claims, photos or a repair estimate, and often a fraud score. Somebody has to decide whether it can be paid quickly, needs an adjuster, or should go to the special investigations unit. On a busy day that call depends a lot on who picks it up.
Helixor routes every claim on the first pass using your written criteria. Small, clean claims get fast-tracked. Coverage questions go to an adjuster. Claims with fraud indicators go to the SIU with the specific indicators listed, so the investigator starts from a reason instead of a hunch. Helixor doesn't send denial letters. When coverage looks absent, the claim goes to a person for coverage review with the reason attached.
A consumer or small business application pulls a bureau report, internal account history, income or cash-flow data, and an identity and fraud check. A scorecard or model turns that into a score. Then policy takes over: cutoffs by product, debt-to-income caps, exposure limits, and the exceptions a credit officer is allowed to approve.
Your model keeps producing the score. Helixor applies the policy on top of it, in-process, with the same pack version across web, branch and partner channels. When a criterion fails, the remedy names it and says what it would take to pass, which is exactly what an adverse-action notice needs.
A property insurer can only take on so much risk in any one area before a single hurricane or wildfire could hurt it badly. Cat models estimate probable maximum loss by zone, reinsurance treaties cap what's covered, and regulators and rating agencies watch capital. Most carriers manage this with monthly spreadsheets and blunt moratoriums, so they stop writing good business in one zip code while an overexposed one next door keeps filling up.
The accumulation limits, treaty terms and appetite rules go into the pack, and the goal is expected premium at an acceptable loss ratio. The solver works out how much capacity each zone and product should get this month and returns a plan you can hand to distribution. When a new application arrives, the underwriting pack checks it against the remaining capacity in its zone. After a rate change or a new cat model version, re-checking the whole book runs as one batch on a GPU, which is what the engine is optimized for. The pack holds the hard rules and the goal. The solver searches for the best plan that satisfies every hard rule, working from the same typed state the decision core uses. If no plan can satisfy them all, it tells you so instead of quietly breaking one. The solvers are built for GPUs, and that's where large plans solve fastest. Solvers are in Preview today.
{
"plan_id": "CAP-2026-10",
"receipt": "hx_proof_5c0e8a2f71b9d43e60a1f7c2",
"pack": "acme.property_capacity v4",
"action": "plan_issued",
"binding_constraints": ["PML-ZONE-FL-SE-100YR", "TREATY-CAT-XOL-LAYER2"],
"input_sha256": "40ab...e913",
"logged_at": "2026-09-28T12:00:00Z"
}| Decision | What Helixor decides |
|---|---|
| Fraud step-up | Pass, challenge or hold a login or transaction, with the reason attached. |
| Payment release | Release, hold for review or return a payment. |
| Collections routing | Choose the treatment path and the team. |
| Renewal and non-renewal | Apply renewal guidelines to the updated loss and driver history. |
| Workload | Pattern |
|---|---|
| Fraud, limits and payments on the hot path | In-process or loopback sidecar |
| Underwriting workbench, quoting and claims intake | Shared decision service, so every channel runs the same guideline version |
| Gray-zone files and learning | Hybrid escalation, sending only selected, redacted evidence |
This is not legal advice. Fair lending, unfair discrimination and adverse-action rules vary by product and by state. Helixor records exactly which criteria decided each case, but your compliance team decides which criteria you're allowed to use.
Go deeper on helixor.dev: Industry guide ↗ • Deployment patterns ↗ • Core concepts ↗ • Receipts ↗
Bring a written policy and a handful of past cases. We'll turn it into a pack and run it next to your current process so you can compare.