Adding a chatbot is easy to imagine. Building an AI layer the whole company can rely on is a different question. Oluyinka Abubakar is an engineering lead at MTN Nigeria with a background in software architecture and backend systems.

The talk asks teams to move beyond isolated experiments and decide where intelligence should fit in a real system. That begins with understanding what the organisation is ready to support.

If an agent can take action, whose authority is it using? Abdulmalik Salawu is a senior DevSecOps engineer at Bluebulb Financials.

With Yusuf Sanusi and Lateefah Bello, the conversation will move from a useful idea to the boundaries that make it safe to deploy: what an agent can see, what it can do, and what a person must approve. What action would you never let an agent take without human approval?

Cloud access is only the first step. Who helps a new builder make something that lasts? Oladosu Ibrahim is a Cloud and DevOps engineer at Turing and founder of CodeSphere Hub.

The panel will look at the gap between discovering powerful tools and gaining the skills, support, and opportunity to put them to work. Oladosu’s experience building learning pathways makes that gap a practical question for Ilorin.

An agent reaches real users. What changes first: the architecture, the cost, or the controls? Lateefah Bello @Teefah_codes brings a cloud perspective from Arthurite Integrated to “Build, Secure, Scale: Developers and Builders in the Agentic Era.”

A promising build needs more than a launch button; builders also need visibility, sensible limits, and a way to recover when usage changes. What would you monitor first as an agentic service starts to scale?

A demo works once. A dependable system knows what to do when it fails halfway through. Yusuf Sanusi is a senior backend engineer at Autochek Africa whose public work includes loan-processing services and document workflows.

Yusuf joins Abdulmalik Salawu and Lateefah Bello to examine what changes when agents move from answering questions to taking actions for real users. If an agent stops midway through a task, what must it remember before trying again?