Agentic AI set to reshape financial fraud in Kenya, BCG warns

Kenyan banks have a narrow window to build defences against a new generation of AI-driven financial fraud before criminals gain access to the same technology, according to Boston Consulting Group (BCG).

Generative AI is already enabling criminals to create convincing synthetic identities, cloned voices, fake documents and realistic digital interactions at scale. As live deepfake technology becomes more accessible, scams are becoming harder to detect.

BCG says the same technology gives banks an opportunity to strengthen fraud detection and respond faster to emerging threats, but only if they act while agentic AI is still in its early stages.

The risk is acute in Africa, where mobile-first banking and digital payments are accelerating financial inclusion. As digital ecosystems expand, so does the scope for criminals to target consumers and financial institutions with AI-powered scams.

Data from TransUnion’s H1 2026 Update: Top Fraud Trends report shows that while suspected digital fraud rates declined in 2025, criminal activity has become more organised and targeted, with greater exploitation of digital identities. Median consumer fraud losses in Kenya reached Ksh108,482, while South Africa recorded unusually high suspected fraud rates in login attempts.

The bigger threat, BCG says, lies in agentic systems capable of running scams end to end without human intervention. The firm estimates the cost of running scams could fall by 90 percent or more once this happens, allowing criminals to launch far larger volumes of attacks and adapt quickly against new defences.

“What makes agentic systems so concerning for Kenya is that they could automate the entire fraud value chain in one of Africa’s most digitally connected financial ecosystems,” said Toivo Hensgens, partner at BCG in Nairobi.

He said the falling cost of running scams would allow criminals to launch highly personalised attacks at unprecedented scale and speed, particularly in a market where mobile money and digital banking are central to daily life.

“Banks can act now to build adaptive, AI-enabled defences, redesign operating models for scale, strengthen ecosystem coordination, and prepare for surge events,” Hensgens said.

Scams and fraud already cost consumers and businesses about $442 billion annually worldwide, according to the Global Anti-Scam Alliance. Scams involve customers being deceived into authorising payments themselves, while fraud involves criminals gaining access to victims’ accounts directly.

BCG said banks across Africa increasingly face operational costs beyond consumer losses, including disputes, complaints, credit-reporting issues, debt collection and customer-servicing costs, even where liability frameworks vary by market.

Agentic AI capabilities have improved tenfold each year since 2024, BCG said. If that pace continues, models could within one to two years run scams lasting days or weeks, allowing individuals and small teams to carry out fraud that currently requires organised syndicates.

Frontier AI models typically carry safeguards against misuse, but open-source models tend to catch up with leading capabilities within six to 12 months, giving banks a limited window before criminals gain access to similarly capable tools.

Henok Eyob, managing director and partner at BCG in Nairobi, said banks that invest early in fraud prevention will be better placed than those that wait.

“Kenya’s leadership in digital financial services has opened enormous opportunities for financial inclusion. But it has also created a greater need for stronger, more sophisticated fraud defences,” Eyob said.

He said banks need models that detect suspicious activity early, intervene in real time, stop high-risk transactions and protect customers from increasingly sophisticated scams.

BCG has set out five steps for banks to prepare. These include increasing threat monitoring through richer behavioural and transactional data, red-teaming their own detection models and tracking the capabilities of open-source AI systems.

Banks should also use the lead frontier models hold over open-source alternatives by building internal AI teams and faster approval pathways for new capabilities, BCG said.

Other recommendations include building operating models that can scale quickly during volume spikes, deepening collaboration with payment providers, telcos, regulators and law enforcement, and designing “fire breaks” such as surge playbooks for periods when fraud activity spikes.

“The transition to agentic scams and fraud is unlikely to be gradual. It could happen faster than many institutions expect,” Eyob said.

He said banks that act now to redesign how they prevent, detect and respond to fraud will be better placed to protect customers and limit losses, while those that delay risk falling behind adversaries that are learning faster than they are.

Download the FREE Kaspersky Next Enterprise  Security Guide here to explore the complete framework for simplifying security operations and building cyber resilience.

Go to TECHTRENDSKE.co.ke for more tech and business news from the African continent and across the world.

Follow us on WhatsApp, Telegram, Twitter, and Facebook, or subscribe to our weekly newsletter to ensure you don’t miss out on any future updates. Send tips to editorial@techtrendsmedia.co.ke