Sentinel: augmented intelligence for banking, insurance and energy

Faced with a regulatory arsenal that keeps growing (marked by the imminent arrival of the AML Package (Anti-Money Laundering) and the activation of the new powers of the European authority AMLA (Anti-Money Laundering Authority)), the fight against money laundering and terrorist financing (AML-CFT) is reaching an operational breaking point.

Today, the finding is unequivocal: only 3% of the AML alerts generated actually result in a suspicious activity report. Behind that figure lies a major challenge: Compliance departments and IT teams provide powerful tools, but analysts remain trapped in manual processes and siloed interfaces.

To understand how AI can industrialise this intelligence, let’s step into the daily routine of Thomas, an AMLO (Anti-Money Laundering Officer) at a large financial institution, through two alternative realities.

Perspective 1: Thomas’s ordeal, the analyst drowning under siloed tools

8:30 a.m. Thomas sits down at his desk. A new level-3 alert is flashing on his screen: Atypical transactional behaviour: Suspected money laundering. For Thomas, a race through regulatory and technical obstacles begins. The bank has equipped him with the best solutions on the market, yet every check is a fight against the clock.

• The trap of constant “copy and paste”: the alert concerns a suspicious €120,000 transfer to a third-party company. To verify the identity of the client and the counterparty, Thomas opens the KYC tool. The file is incomplete. He then logs into the bank’s CRM to look for an up-to-date supporting document, and then opens a third internal tool to manually extract the transaction history of the last six months.
• Documentary investigation beyond the walls: next, Thomas has to leave the bank’s applications to open the commercial registry website and laboriously search for the ultimate beneficial owners (UBOs) of the third-party company. Once he has those names, he moves to the Negative News platform the bank provides to check whether their reputation is tainted. The tool works, but it returns dozens of articles: it is Thomas who has to read every paragraph and manually filter out the namesakes.
• The race against the clock: digging deeper, Thomas discovers that his client is linked to three other accounts within the bank. To analyse this interconnected cluster of 4 people, Thomas must repeat the entire procedure (KYC, transactions, commercial registries, sanctions lists, Negative News) four times in a row. Isolated in front of his screens, handling the data is so heavy that each individual file takes between 30 and 60 minutes. For this single complex cluster, the overall processing brushes against the absolute limit of human productivity: nearly 1 hour 30 at most, without a single break.
• The blank-page syndrome: 10:00 a.m. Exhausted by this data gathering, Thomas must now write the investigation summary for TRACFIN (the financial intelligence unit). He types his report by hand, trying to summarise from memory the financial flows and press alerts he himself has compiled.

Thomas spends 90% of his time handling tools and only 10% exercising his expert judgement.

Perspective 2: Thomas, the analyst augmented by the Sentinel solution

Let’s replay the same day, at the same time. 8:30 a.m. The same alert about the suspicious €120,000 transfer appears. But this morning the institution has deployed Sentinel, the augmented investigation solution designed with our support.

• Automatic activation of agentic AI: Thomas doesn’t need to open five different programs or manually look up the ownership tree in the commercial registries. As soon as the raw signal was detected, Sentinel took over in the background. Without any human action, a team of specialised virtual agents connected instantly to all the sources in the ecosystem (Core Banking, KYC, filtering) and to the external regulatory registries.
• Sentinel’s orchestration work: while Thomas was having his coffee, Sentinel smoothly executed a structured investigation journey:

-It aggregated the complete client profile and transaction history (Circle 1).

-It queried the Negative News tool and the sanctions lists, smoothing out and eliminating false positives due to namesakes on its own (Micro-Vigilance).

-It scanned the commercial registries to reconstruct the beneficial owners of the third-party company.

-Finally, it automatically mapped and analysed the interconnected cluster of 4 people (Circles 2 & 3).

• Exponential performance: instead of brushing against an hour and a half of intensive manual investigation, Thomas opens his dashboard and finds a fully pre-analysed, consolidated and sorted file in exactly 5 minutes.
• Full and auditable trust: Sentinel proposes a clear recommendation (for example, an escalation to the compliance unit and a TRACFIN report). But Thomas keeps absolute control: the system applies the strict Human-in-the-Loop principle. Sentinel provides full explainability across 6 dimensions in the form of an immutable audit log. Thomas sees exactly which document or which Negative News article triggered the suspicion, associated with a factual confidence score.
• The sovereign decision: Thomas validates the conclusions, exports the complete, sourced investigation report as a PDF and electronically signs the decision. His validation instantly enriches the solution’s evolving memory for future investigations. In less than 10 minutes, the decision is made and perfectly documented.

Conclusion: reinstating the business expert

Augmented investigation replaces neither the bank’s existing tools nor its analysts: it builds the intelligent bridge that was missing. By entrusting Sentinel with the tedious tasks of navigation, mass reading and document cross-checking, AI gives the analyst back their true role: that of sovereign, top-level judgement.

To ensure the success of such a setup, the architecture relies on a Compliance by Design approach: local (in-situ) data processing for strict compartmentalisation, systematic anonymisation of data flows before they are processed by the AI, and absolute respect for banking secrecy and the GDPR.

We support you step by step along this path thanks to our proven 4-maturity-level methodology (from conceptual Foundations to governed Operations) and our leading expertise in deploying agentic AI solutions applied to compliance.