By Paul Fakayejo, tech analyst
There is a palpable fear in the banking industry lately due to rumours of an AI that can detect fraud by merely listening.
The name of the AI is Safi and it is making banks nervous again.
Safi AI is owned by Insolify, one of the leading tech giants in Africa, according to tech industry sources.
According to Paul Fakayejo, an industry analyst, “When Safi first entered African banking halls, the controversy it caused was a bit simple however weird already: An AI was talking to customers. Now, the controversy is far bigger than imagined, because Safi is no longer just listening, it is thinking financially.”
“Safi has gone from being just a Voice Agent to being a Financial Intelligence Engine. Safi has quietly evolved beyond customer care AI into something banks rarely admit they need:
real-time Financial Intelligence (FI),” he further said.
“Every call, WhatsApp message, USSD interaction, and support request Safi handles is no longer
treated as “support noise.”
It becomes structured financial signals; such as repeated dispute patterns, unusual transaction complaints, behavioural inconsistencies during verification, sudden changes in customer language or urgency and cross-channel anomalies (voice vs WhatsApp vs app activity).
“Individually, these signals look harmless, but altogether, they expose fraud before money leaves the system,” he said.
Paul further emphasized that Safi functions in ways it makes fraud detection not needing to wait for alerts.
He emphasized further that traditional fraud systems rely on rules, thresholds, and post-transaction alerts. “Safi doesn’t wait. It detects fraud during conversation.
If a customer suddenly struggles to answer routine verification questions, switches languages unnaturally, pressures for urgent overrides or contradicts recent transaction history; Safi flags the interaction instantly, sometimes before the transaction is even completed,” Paul said.
Paul said “Critics call this intrusive. Banks call it something else:
the earliest fraud signal they have ever had.”
When responding as to why this changes banking operations, he said “Safi’s FI layer does not replace a core banking system, it feeds the core banking system. When integrated with systems like Fincore, Safi provides: context behind transactions, not just numbers, behavioural risk signals tied to real identities, cross-channel intelligence linking voice, chat, and account activity and early warnings that allow Fincore to enforce controls before settlement. In simple terms: Fincore protects the ledger.”
The industry expert said “Safi protects the conversation that leads to the ledger entry. That link did not exist before Safi.”
He described the innovation as “Multi-Channel Banking, Finally Acting Like One System”
According to Paul, “Most banks claim to be ‘omnichannel.’ Customers know better. Voice, WhatsApp, apps, and branches usually don’t talk to each other.
Safi changes that.”
“A customer who complains on WhatsApp, calls an hour later, and attempts a transaction in-app
is no longer three separate cases. Safi connects them into one financial narrative,” he further said.
“This unified intelligence allows banks to; detect coordinated fraud attempts, stop social-engineering scams in progress, assist customers seamlessly across channels, reduce false positives that frustrate legitimate users, for the first time, banking channels actually agree with each other,” he further said.
Addressing the new fear, Paul said it is “fraud prevention without friction”.
“Here is what really unsettles the industry, safi prevents fraud without extra OTPs, delays, or customer stress; no new forms, no extra buttons; no “please visit a branch”, just conversation analysis, financial context, and intelligence. That means: fewer blocked legitimate transactions, less customer anger, better regulatory reporting, stronger internal controls and no visible ‘fraud system’ for competitors to copy,” Paul said.
He further said “Regulators are paying attention quietly, because FI-driven fraud detection creates new questions; who is accountable when AI flags intent? Can conversational intelligence be audited? Should fraud prevention happen before or after customer authorization?”
“Safi is forcing regulators to think beyond transactions, towards intent, behaviour, and financial context. The real disruption is not AI, it’s awareness.
“Safi did not expose a technology gap. It exposed a blind spot.
Banks spent decades watching transactions, but ignored the conversations that created them.
Now an AI is listening to those conversations, connecting them to financial intelligence, and stopping fraud before it becomes a report.
That is why Safi is not controversial because it talks. It is controversial because it understands.
And once banking systems gain that kind of awareness,
there is no going back,” he concluded.
