AI in Tunisian Banking: Real Uses, Rules and Gaps

AI in Tunisian banking is real but still early. The most visible uses so far are chatbots, remote identity checks and credit scoring tools, often built with local tech firms, while the sector still has no AI-specific rules. For fintech founders, that gap is both an opportunity and a risk.

Where AI is already used in Tunisian banking

Public information on bank AI projects in Tunisia is limited, but several concrete cases are documented.

Customer service chatbots

In September 2020 Amen Bank launched AmenBot, presented as the first intelligent banking chatbot in Tunisia. Built by the bank's internal teams with Tunisian startup Deepera.AI, it used natural language processing to answer questions about products, collect complaints and guide users through online account opening with Amen First Bank, handing over to human agents when needed. It was launched on Facebook Messenger, available around the clock.

Remote identity checks

Remote onboarding is the area where AI is moving from experiment to regulation. A BCT circular of 28 February 2025 allows banks to onboard customers electronically, provided identity verification is at least as strong as in person. For payment institutions, circular 2026-10, published on 25 September 2026, requires at least two-factor authentication and a presence check by facial recognition or an equivalent method, with the technology audited at least every two years, according to Challenges and Webdo. Liveness detection and document verification typically rely on computer vision models. Our guide to e-KYC and AML rules covers the details.

Credit and risk tools

Lending is the next frontier. Islamic microfinance institution Zitouna Tamkeen put into production in June 2025 a credit management platform from Tunisian vendor Talys Solutions, covering scoring, screening and AML checks. Startup EasyBank, which raised TND 1.2 million in January 2025 according to Wamda, says it uses AI to personalise advice and streamline loan eligibility checks. As early as 2021, the BCT's first sandbox cohort included projects using AI for remote customer identification.

Several banks are also signalling interest publicly. UBCI, for example, partnered with the Global AI Congress Africa held in Sousse in November 2025 and said it was relying on AI to strengthen its agility and innovation.

The policy backdrop

Tunisia does not yet have an AI law or a published sector rulebook for AI in finance.

• National AI strategy. In February 2025 the Ministry of Communication Technologies said a national AI strategy was in the final drafting phase, with an operational plan to follow. We could not confirm its formal adoption as of October 2026.
• Development plan. The 2026–2030 development plan lists digital transformation among its priorities, alongside financial inclusion and regional balance.
• Data protection. The main constraint on bank AI is Organic Law 2004-63 on personal data, overseen by the INPDP. It predates modern machine learning and is widely described as under revision. See our article on data protection for fintechs.

In opinion pieces published in L'Economiste Maghrébin and Business News in 2026, Abdelwaheb Ben Moussa, an IT manager at a Tunisian public bank, argued that the BCT and the Conseil Bancaire et Financier could issue AI-specific guidelines on model validation, stress testing and reporting. He suggested a gradual path: document automation and chatbots first, then fraud detection and alternative SME scoring, and only later dynamic credit allocation.

The two-speed problem

The same author describes a "two-speed" modernisation. Private banks have more room to invest in AI for scoring, fraud detection and predictive analytics. Public banks face ageing information systems, weak asset quality and limited budgets, and AI does nothing to fix non-performing loans or governance gaps.

There are also structural risks that apply to all banks:

• Dependence on foreign tools. Most AI solutions used are developed abroad, which limits banks' ability to audit algorithms or detect bias.
• Explainability. Automated credit decisions can be harder for customers, auditors and supervisors to understand.
• Data quality. Models are only as good as core banking data, and many banks are still migrating to new systems.
• Cybersecurity. More automation means more attack surface. See our article on cybersecurity in Tunisia's financial sector.

Opportunities for Tunisian AI and fintech startups

Tunisia has a strong AI talent base, illustrated by the success of InstaDeep, acquired by BioNTech (see lessons from InstaDeep). For founders, the openings in banking are practical:

1. Arabic, Tunisian dialect and French language models for customer service, where foreign tools often perform poorly.
2. e-KYC and liveness detection tailored to Tunisian ID documents and the new BCT audit requirements.
3. Alternative credit scoring for small businesses, farmers and thin-file borrowers, using transaction and wallet data, within data protection limits.
4. Fraud detection for fast-growing mobile payments, which rose 80.5% in transaction volume in 2025 according to the BCT.
5. Compliance tooling for AML screening and regulatory reporting.

To sell to banks, startups should expect long procurement cycles, strict data-hosting requirements and questions about model explainability. Building with auditability in mind is likely to become a selling point once supervisors start issuing AI guidance.