AI-poweredfraud & compliance automation
Catch more fraud. Close investigations faster.
Smart transaction monitoring platform, fully customizable.
SaaS or On-Premise.
Why was this payment blocked?
- Basket above customer average (no match)0
Threshold 60 crossed · case opened
Block and review, score 92 (threshold 60). Contributing rules: disposable email provider +40, datacenter IP egress +30, 6 cards on one device in 24h +22.
A sample card payment. Switch the rules off one by one: the score, the outcome and the reasoning on record are rewritten as you go. In the product, every decision’s response carries this detail. Tap a rule to switch it off. In this scenario, review starts at 30 and block-and-review at 60.
Why does one rule change take weeks?
Fraud patterns change in a week. If changing a rule waits on a software release, that week is a loss.
-
Rules live in the codebase.
The risk team spots a pattern, explains it to an engineer, and the request joins the backlog to wait for a release. The gap between noticing and stopping is measured in weeks.
-
The score is a black box.
The vendor returns a number. When a customer asks why they were declined, or an auditor asks for the reasoning, that number is all you have.
-
Your data leaves.
Transaction and customer records land in someone else’s cloud. Under most regulatory regimes, that alone is a separate compliance exercise. Krino runs on your own servers; nothing leaves.
Same question, two answers.
| Topic | Typical approach | Krino |
|---|---|---|
| Decision rationale | A score comes back; the detail stays hidden | Which rule, how many points, one by one |
| Publish approval | Whoever changed it publishes it | Second approval: the requester cannot approve their own request |
| Record of a rule change | Versions kept, the difference visible to nobody | Two versions side by side, in sentences, not JSON |
| Testing before release | Hope | Shadow mode against live traffic, or against last month’s transactions |
| Where data lives | The vendor’s cloud, often another jurisdiction | Your own servers; nothing leaves |
| Changing a rule | Engineer plus a release, days to weeks | Risk team, on screen, immediately |
| Data model | The vendor’s schema; your fields don’t fit | You define your own tables and fields |
| Investigation workflow | Spreadsheets, email, a separate ticket tool | In the product, with an audit trail |
| Device intelligence | Separate product, separate contract | Included |
| Sanctions screening | Separate product, usually priced per query | Included, matched locally |
Everything an auditor asks is on record.
Who changed a rule, when and why; what it was tested against before release; which decision it moved and how. Krino writes all of it into the product itself, not into a separate system.
Releasing a hold takes two people.
The analyst who proposes releasing a held transaction cannot approve the proposal. It goes to a second approver, and both steps stay in the case history with their timestamps.
A published version never changes.
Edits happen on a draft; the live version keeps its number. A faulty draft is stopped before release, with the reasons listed.
The answer to “which rules were in v3?” is the same months later.
Tested on real traffic before release.
A new version runs in shadow mode next to live traffic, or is replayed against last month’s transactions. It decides nothing and opens no case; it only records what it would have said.
Two endpoints. One afternoon.
You send the transaction, you get the decision. The reasoning comes back in the same response.
-
First, send the event.
A payment, a login attempt, a signup: one API call, in your own data model with your own field names.
POST /api/decisions { "object_type": "card_payment", "scenario_id": "019f8e68-…", "trigger_object": { "object_id": "tx-88214", "updated_at": "2026-08-06 10:00:00", "basket_total": 12500.00, "customer_no": "4471", "client_ip": "203.0.113.45" } } -
Krino enriches it and runs the rules.
Email, phone, IP and device signals come from lists held inside your infrastructure; nothing is queried outside. The scenario’s rules are evaluated together and each adds its score. Around 34 ms.
-
The decision comes back with its reasons.
Approve, review, block-and-review or decline, with the contribution of every rule in the same response.
Ask for the rule detail and the response also carries where the time went: reading data, counting velocity windows, running the rules. When you would rather not wait on the decision, send events separately and receive the outcome over a signed webhook. Details in the documentation.
{ "decision": { "outcome": "BlockAndReview", "score": 92 }, "rule_executions": [ { "rule_name": "Disposable email", "outcome": "hit", "score": 40 }, { "rule_name": "Datacenter IP", "outcome": "hit", "score": 30 } ], "execution_time_ms": 34 }
What if I moved the threshold?
Krino re-reads past decisions at new thresholds and shows how many would have landed differently. Not a forecast: a replay of decisions already made. A rule that fires constantly but is never decisive is adding noise to every score; the same screen shows that too.
Nor do you wait after publishing: a screen streams decisions as they are recorded, and the first minute tells you whether the thresholds held and whether the volume landing in the queue is a number your team can carry.
From detection to filing, in one product.
- Rule engineYour risk team builds rules on screen: compose the conditions, set the score, publish. Velocity rules, lists, fuzzy matching and threshold-based outcomes included.
- Device intelligenceBrowser fingerprinting, datacenter and VPN egress detection, disposable email providers, phone line type. All offline, no extra contract.
- Sanctions screeningName screening against consolidated sanctions and PEP lists. Tolerates accent, spelling and transliteration differences; refreshed on a schedule.
- Case managementDecisions about the same customer collect into one file, route to a queue, get annotated, and record when they were reported to a regulator.
- AI assistantAsk about decisions, rules and cases in plain language and get the answer as a table and a chart. Runs on your own model provider: it reads, it does not write.
One plan,everything included.
No per-transaction fee, no module-by-module upsell. It installs on your own infrastructure; we scope the licence to your organisation.
- Rule and scenario engine
- Device intelligence
- Sanctions and PEP screening
- Case management
- Shadow test runs
- Scheduled runs
- Webhooks and API
- Version upgrades
- Unlimited rules, scenarios and users
How much are you losing a month?
Put your own numbers in. Nothing is converted; whatever currency you enter is the currency you get back.
$1,071,000 prevented loss per year
$89,250 prevented loss per month
Krino is one plan. No per-transaction fee, and it does not grow with your volume. Get a quote
See it onyour data.
The best demo is the one built from last month’s transactions.
Leave your details and we’ll get back to you within two working days to set it up on an anonymised sample.