Born out of necessity
JUSTE began when we saw that newly regulated SMEs were being asked to run a bank-grade AML programme — while most were only doing basic KYC and KYB.
So we built one operating sequence that makes a serious AML function possible without a bank-sized team.
The model provides the intelligence
JUSTE is built control-architecture-first. The AI model is a component inside the system — not the system, and not the authority.
Each agent runs the model best suited to its task, and because the model is selected per agent, any one can be switched to a better-performing model the moment one becomes available — without changing the firm’s policy or who holds authority for the decision. Per-agent model selection has been live since 2025. The intelligence can evolve; the controls remain. How the model provides the intelligence →
Intelligence-first screening
Traditional onboarding collects documents and then decides what the client is. That is backwards: it spends the most effort at the point of least information, and treats a twenty-year client exactly like a stranger.
JUSTE screens first. Sanctions, PEP, adverse media, activity and funds, and territorial risk run before anything is asked of the client. What comes back decides what due diligence is proportionate — and for most clients, that is very little.
Simplified due diligence
An ID document added to a clean, policy-governed screening result is a complete, inspection-ready record. No chase, no checklist, no delay.
Customer due diligence
Evidence requests and questions tied to the specific risk drivers screening surfaced — not a generic pack sent to everyone.
Enhanced due diligence
Deeper questioning, source-of-funds evidence, tracked review and recorded MLRO approval before the client is accepted.
Mobile intake, by whoever is standing there
Compliance work has traditionally been trapped behind the one person licensed to do it, which is why onboarding waits until Tuesday. JUSTE moves intake to the point of contact.
Any qualified team member in scope can open the app, scan a client's ID, and start a governed record. The guardrails travel with them, so the junior at the front desk operates inside exactly the same policy as the MLRO.
Scan the ID
On a phone, at the counter, in the meeting room or at the fair stand. Nothing for the client to install or fill in.
Screening runs
Risk is assessed and classified against your policy before a single further question is put to the client.
The next steps appear
Written out in plain language, so the person at the desk can explain what is needed and why, there and then.
The record exists
Inspection-ready for a low-risk client, and already counted in the firm's audit-readiness score.
Codified risk policy guardrails
Everyone in this market says their AI is human-supervised. In practice that usually means a person can agree with the machine or disagree with it, and nobody records which happened or why.
We built the first working model of executable risk policy governance. Your written AML policy becomes rules that run at the moment of decision. Where the policy requires a different answer than the AI proposed, the guardrail applies it, records both, and locks the field.
The AI recommends. The policy decides.
Each risk type gets an AI recommendation. Your rules sit above it. When they disagree, the rule wins — automatically, every time, for every user.
The override is the evidence
What the AI assessed, what the rule required, which rule and which policy version, and who reviewed it. This is exactly the chain an inspection asks for, produced as a by-product of the work.
Locked, not merely suggested
A field rated by policy rule cannot be quietly changed at the desk. Changing it requires a recorded review, so a guardrail is a real constraint rather than a warning somebody can click past.
Alignment is measured too
Where a case departs materially from policy, JUSTE records it as a material adjustment. A pattern of them tells an MLRO that the policy, or the practice, needs attention.
Illustrative view of a governed risk assessment. The client here would have passed as low risk on the AI's reading alone.
Your whole team can run AML, because the rails decide what they can't do.
This is the change most firms feel in their capacity rather than their compliance. Low and medium-risk administration — which is the overwhelming majority of the book — can be completed by any qualified team member, because the guardrails constrain what is possible and lift out anything that needs a specialist.
Completes the administration end to end. Scan, screen, ID verification, record issued. No queue, no waiting for the compliance officer to have a free afternoon.
Handles customer due diligence with the system directing which evidence to request and which questions to ask. The classification stays governed by policy throughout — they cannot rate their way out of a control.
Enhanced due diligence, source of funds, targeted questioning and a recorded approval or refusal. The file arrives already assembled, with the screening, the rule that triggered it and the evidence so far attached.
The MLRO can be assisted in their team on the platform by an external AML expert with specialist analysis, source-of-funds or source-of-wealth review, or other expert support. But accountability remains with the MLRO at all times.
The same four words, meaning something different.
Everyone claims AI with human oversight. Here is what the phrase means in practice on each side.
Independently noticed, not self-declared.
Participation in the FCA Innovation Pathway does not constitute FCA authorisation or endorsement. Programme and award references are shown for context.
Watch a guardrail overrule the AI on one of your own clients.
It takes about two minutes to see, and it is the fastest way to understand what makes JUSTE different from everything else you have been shown.