Businesses are racing to adopt artificial intelligence. Yet new research suggests that only 4% of surveyed Gulf organisations have fully integrated AI into their operations. And we are now entering a new phase where AI does not just answer questions but can actually make decisions and take action. Are Egyptian businesses ready for this shift — and who takes responsibility when AI gets it wrong? Joining us is Muamar Magam, Managing Partner at Quintus LTC. Muamar, welcome back to Fintech TV.
Thank you Bassel. It is always a pleasure joining you again. Greetings to you and all Fintech TV viewers.
We are entering the era of agentic AI. What changes when AI moves from answering questions to actually making decisions and taking action?
The important change is that AI can now carry the work through rather than leaving every next step to a person. As Salim Ismail writes in Organisation Singularity — the assistant answers, the agent acts. Let us take an Egyptian entrepreneur seeking finance. An authorised agent could coordinate routine checks, leaving the banker to focus on seasonal income anomalies or unusual business patterns within the approved boundary. Human oversight moves from reviewing every step into the exception loop. But the real value comes when we learn from those exceptions — capturing the banker’s corrections, testing them, improving the process without allowing the system to write its own policy. That brings Egypt’s young talents together with experienced bankers who understand customers and markets.
New research suggests only 4% of Gulf organisations have fully integrated AI. What is preventing businesses from moving beyond experimentation?
Many organisations are introducing AI faster than they can change the way decisions move. If there is better analysis but it still waits on the same approval chain, the institution may simply have a longer queue of recommendations. The 4% figure is from surveyed organisations only — not every business. But importantly, it shows the gap between tools and integration. Government can help turn ambition into a delivery agenda. Last April, the UAE announced it is targeting agentic AI across 50% of federal government sector services and operations within 24 months — with 90-day entity milestones and regular monitoring. That gives local talent and suppliers a clear opportunity to contribute. For Egypt, the takeaway is: apply the same discipline to one important process with safeguards and customer outcomes. Give the team room to let the work demonstrate progress — then build the confidence to expand.
Egyptian fintechs like MMT Highland are already using AI to support lending and collection decisions. How far should financial institutions allow AI to go in making decisions independently?
I would give AI more independence only when the evidence proves — rather than assumes — that the system is ready for each decision. MMT’s machine learning scoring and collection automation does not establish unrestricted agent autonomy. A payment reminder is very different from distributing debt or managing customer hardship — where human judgement remains essential. We need to test existing services with governance, data, risk, and compliance involved. The experimentation output must not determine the customer outcome directly. Once the workflow proves its value and reliability, bring it back to the main bank and scale carefully. Each successful experiment gives the institution evidence for the next step — rather than asking customers to take the risk.
When an AI system makes a financial decision that harms a customer — who should be held responsible?
For the customer, the bank must remain a place they can turn to for explanation, correction, and genuine appeal. They should not have to resolve disputes between an institution and a technology supplier. The HSBC CEO put it very nicely — I need human judgement, I need human decision-making, and I need human accountability at scale. Organisation Singularity makes that practical through what is called fiduciary alignment — every autonomous decision connects to a named human able to intervene. The takeaway: delegating execution can improve bank capacity. But trust still depends on someone being willing and able to answer for the result.
Thank you very much for your valuable insights.
Thank you. It is a pleasure to be with you today. Thank you Bassel.
