BCT circular 2025-17, KYC, alert triage, regulatory reporting: we deploy AI agents that run these workflows end to end, with an audit trail on every step. Human in the loop, data in Tunisia.
In a regulated financial environment, an AI system that is almost right is a liability. We design for the framework you already operate in: your compliance obligations, your core banking stack and your own risk appetite, with the monitoring and documentation your regulators expect.
The problems we solve every day for organizations like yours.
Finance, accounts, procurement and compliance teams spend 30 to 40% of the day copying figures off invoices, checking purchase orders, assembling KYC files and preparing regulatory submissions. Manual entry mistakes alone delay 15 to 20% of loan applications. And when an auditor asks for the full trail behind a transaction, it has to be stitched together from three or four different systems. AI agents run these workflows end to end and leave a traceable trail on every step.
Customer, account and transaction data sit in hundreds of systems that don't talk to each other. The integration layer becomes a mess of conflicting definitions, duplicates, missing fields and mismatched formats. Loan officers re-enter the same data, credit analysts reconcile exposure reports by hand, reporting comes out inconsistent and decisions wait. 93% of banking executives name legacy systems and integration complexity as their biggest modernization problem.
Rule-based systems fire so many false alerts that analysts spend their day clearing noise instead of chasing real risk. Fraud schemes change faster than the rules do. AI agents watch transactions as they happen, catch the multi-variable patterns people miss, and pull the time from detection to response down from hours to minutes.
Since 22 December 2025, BCT circular 2025-17 requires a formal AML/CFT risk assessment, immediate declarations through goAML, and a new quarterly frozen-asset return. No transition period. At the same time the sector's cost-to-income ratio has reached 48%, its highest since 2019, so the load lands on teams already stretched. AI agents prepare the files and the consistency checks, validation stays with your teams.
We do not discover your obligations mid-project. We design for the framework you already operate in.
Formal risk assessment, immediate declarations through goAML, quarterly frozen-asset return. Our compliance workflows are built to feed these obligations, not to work around them.
Human in the loop, traceability and documentation: the direction the BCT and the Conseil Bancaire et Financier converge on in their guidance.
Loi organique n° 2004-63 on the protection of personal data. Hosting and processing inside your perimeter, in Tunisia. Direct identifiers are detected and masked or tokenized before storage, and the design accounts for what can still be inferred from the fields that remain.
Every decision leaves an exportable trail that can be presented as-is to a control or an external auditor, with no manual reconstruction.
Concrete applications that deliver measurable outcomes.
Score transactions in real time and prioritize alerts so investigation teams focus on genuine risk instead of false positives.
Extract and validate data from KYC files, loan applications, and contracts, with audit trails for every decision.
Automate regulatory report preparation and consistency checks across systems, keeping humans in the loop for sign-off.
Handle account inquiries, card issues, and onboarding questions around the clock, escalating sensitive cases to your staff. 57% of financial services executives put customer service at the top of their AI agent wish list, and agents keep the cost per interaction down without adding headcount.
Our core solutions tailored to the requirements of your sector.
Customer-facing assistants engineered for the tone, security, and escalation rules of financial services.
Automate document-heavy and compliance workflows while keeping accountability clearly defined.
When off-the-shelf AI is not enough, we design systems around your constraints, not the other way around.
Inside your perimeter or a private cloud you control, in Tunisia. No model training on your data, and no transfer outside the framework of loi organique n° 2004-63.
Every decision leaves a traceable trail: input, model, version, human validation, timestamp. Your compliance team exports the full trail for the BCT, the CTAF or an external auditor without rebuilding it by hand.
We connect to your existing stack. No migration, no replacement: agents read from and write to your current tools through controlled, logged interfaces.
Ambiguous or high-stakes cases are escalated to a human, never decided alone. Confidence thresholds are set with you, and a model can be pulled from the flow without stopping it.
No. Your data runs your workflows, it does not train our models. The scope is written into the contract. Where we need records to develop or test against, we work on synthetic data that mimics the statistical shape of yours, never on the originals.