How Global RADAR is delivered
Global RADAR licenses the AI Alert Clearing Agent. The platform screens approximately 10 million names a month across its client base. We do not employ the analyst bench. There are three ways to buy it.
- Platform. You license the engine. It screens, matches, ranks and disposes of low risk alerts automatically, with a recorded reason code and a rationale for every disposal. Your own analysts review what escalates.
- Supervised Clearing. The platform, plus the clearing policy framework, the sampling methodology and a monthly audit pack. Your analysts still make the residual judgements, inside a supervision model your examiner can read.
- Managed Clearing. An accredited delivery partner performs the residual review, running on our platform. You can contract for it through Global RADAR under a single agreement, with the partner behind us, or directly with the partner.
In all three, the clearing policy is yours to approve, every disposal carries an exportable decision record, and the regulatory determination stays with your compliance officer.
How alert clearing works
Illustrative ratios showing how volume falls across the four stages. An alert is a screened name that hit something. A concern is a single list entry that name matched against, and one alert can carry many. John Smith is one alert and may carry fifty concerns. Clearing is priced per concern, so the concern count, not the alert count, is the number that sizes the work. Your own ratios are established during the pilot and reported to you monthly.
The six steps behind those numbers.
The control stays yours. The engine carries the volume, your compliance officer makes every decision that counts, and every step above writes to an audit trail you can export.
What always reaches a human
The decision logic is deliberately asymmetric. It is built to escalate rather than to clear.
- Any uncertainty, any missing critical data, any conflicting information is escalated. That is a rule written into the agent, not a tendency we hope for.
- One possible true match escalates the whole case. The case moves to Customer Action and can only be updated by you.
- An exact name match is never cleared on the name alone. It is only disposed of where a date of birth, address, nationality, passport or IMO positively establishes a different party.
- Vessels are not cleared without IMO evidence, and entity type is never assumed where the documents do not make it clear.
- Aliases are always checked, because sanctioned parties routinely use them when buying financial products.
The failure modes fail safe. If the agent returns anything that does not parse, or the service is unreachable, the case stays in the alert queue for a human. It never falls through to cleared.
Read the full escalation and governance policy
How you supervise the work
Automating the work does not automate away the obligation, and we do not ask you to take the Agent's disposals on trust. Four things make this service supervisable:
- A decision record for every alert. Every cleared alert carries the reason it was cleared, the data it was cleared against, the model score and the timestamp. Nothing is disposed of silently.
- Independent sampling. You draw your own sample of cleared alerts, at a rate you set, and re-review them yourself. We do not choose the sample.
- Our own quality assurance. A second Global RADAR reviewer re-checks a standing proportion of cleared alerts, and the result is reported to you rather than kept internally.
- An exportable audit trail. The full record is exportable on demand, in a form you can hand to an examiner or an internal auditor without translation.
If a sampled decision is wrong, it is our finding to explain and our process to correct, and the correction is recorded in the same trail.
Validated on a live test run, not a slide
A 1,000 name validation run was completed in a controlled non-production environment in April 2026, measured against the same population processed without the AI Alert Clearing Agent:
- 47.4% reduction in manual workload. Fewer cases reached a human reviewer.
- 98.2% reduction in raw, uncategorised alerts. Alerts arrive categorised rather than raw.
- Clearance rate improved from 94.3% to 97.0%.
- Escalated cases arrive in a targeted Investigate state carrying a written rationale, not in a generic alert queue.
The service runs today for specialty insurers and banks operating to Lloyd's of London market standards, delivered alongside partners including NTT DATA, Alchemy London Market and Buckhill.
See the full methodology, guardrails and audit design
How clearing is priced
Clearing is priced on volume, not on seats. A quote is built from up to five components, and which of them apply depends on whether you take Platform, Supervised Clearing or Managed Clearing.
- Platform fee. A fixed monthly fee for the environment, the configuration and the audit trail.
- Per name screened. A unit rate applied to every name put through the engine, whether or not it raises an alert.
- Per name under continuous monitoring. A monthly unit rate for names re-screened against list updates, rather than screened once at onboarding.
- Per concern serviced. A unit rate for each list match that needs human disposition. One alert can carry many concerns, and the concern is the billable unit, not the alert. This applies under Managed Clearing, where an accredited delivery partner does that work.
- Volume bands and a minimum. Unit rates step down as monthly volume rises, against an agreed minimum volume or minimum monthly fee.
Growth is priced in automatically rather than renegotiated at renewal. When your book grows the unit rate steps down and the bill follows the work rather than the headcount. Volume bands, reporting cadence and support hours are set out in your licence schedule. Rates are not published, because bands are set against your actual volume and list scope. Talk to us about which components apply to your book.
Start with a pilot on your own alerts
The fastest way to evaluate a clearing engine is to give it real work and audit the result.
Run the Agent on your own alert queue for 30 to 90 days, scoped to your volume. You draw your own sample and you see the disposal rate, the decision record and the turnaround on your own data before anything is signed. Enterprise books run a 90 day pilot at $45,000, creditable against year one. You keep the output either way.
Start a pilot on your own alerts
Read the full clearing methodology and model governance
Background reading on the obligations: UK sanctions and OFSI duties for insurers, OFAC obligations for insurers writing US risk, bordereaux screening under delegated authority, EU sanctions and AML obligations, FinCEN and the Bank Secrecy Act, and adverse media under FCA expectations.
Sanctions Clearing FAQ
The AI Alert Clearing Agent ingests every alert from your screening system, scores it on genuine risk, disposes of low risk matches with a recorded reason code, and puts what remains at the top of the queue for a human decision. Global RADAR licenses the engine. Under Platform and Supervised Clearing your own analysts review what escalates. Under Managed Clearing an accredited delivery partner does, running on our platform. Global RADAR does not employ the analyst bench.
Yes. The platform is built for mid size and global financial institutions running tens of thousands of alerts per month. Every clearing decision is documented, reproducible and aligned to your existing risk policy, and the same workflow scales from a single book of business to a group wide programme.
Yes, and it is worth being precise about who does what. Global RADAR supplies the platform that covers screening, matching, risk ranking, autoclearing, escalation routing, list management for OFAC, EU, UN and HMT, and the exportable audit trail. The residual human judgement is performed by your own analysts under Platform and Supervised Clearing, or by an accredited delivery partner under Managed Clearing. Global RADAR does not employ the analyst bench, and you retain regulatory responsibility for every decision.
Managed Clearing is delivered by an accredited partner running on the Global RADAR platform. You can contract for it through Global RADAR under a single agreement, with the partner behind us, or directly with the partner. From us you get the engine, the clearing policy framework, the sampling methodology, a decision record for every alert, monthly reporting on false positive rates and alert volumes, and quarterly tuning recommendations. Turnaround times, coverage hours and analyst staffing are set by the delivery partner in the service schedule.
A 1,000 name validation run was completed in a controlled non-production environment in April 2026. It produced a 47.4% reduction in manual workload, and a 98.2% reduction in raw, uncategorised alerts. The clearance rate improved from 94.3% to 97.0%. Your own ratios are established during the pilot, which runs 30 to 90 days scoped to your volume, and reported to you monthly.