What does AI mean for your job?

For most of us AI is already starting to change the workday. We are handing over more and more of our most time consuming and tedious tasks to Claude, ChatGPT, and Gemini, and maybe even to our own agents but how much responsibility are businesses really ready to hand over to the machines?

We looked into all of that and more on the jobs panel at the EnterpriseAM Egypt Forum — the AI edition. Klivvr CEO Nils Bachtler, Visa’s vice president and head of products and solutions for North Africa, Levant, and Pakistan, Nader Soliman; and Windfall CEO Youssef Rizk discussed how they use the technology, where agents are taking over, and what still requires human judgment.

Setting the stage: A faster workweek doesn’t necessarily mean a shorter one. Bachtler’s schedule remains busy, but AI has changed how Klivvr operates, delegates work, and makes decisions. The young credit business has “used the technology from day one, helping it grow faster and improve performance,” he said. For Soliman, the biggest change has been less time spent on gathering and collating information and more time interrogating it, weighing strategies, and spending time with clients. He sees AI as a “force multiplier” that frees him to focus on judgment and product innovation. “At the end of the day, the ultimate outcome is still my accountability and my responsibility,” he said.

At an AI-native startup, the shift goes much further. Rizk, who is also CTO of Y Combinator-backed Wondercraft, described a business where engineers work with Claude and Codex, while marketing, sales, and SEO also run on AI. “What you used to need 10 people to do, you can now realistically do with around two to three people,” he said. Those employees take ownership of planning their functions, with AI doing much of the execution.

Most businesses are still at the read-only stage. Through Windfall’s work helping companies implement AI, Rizk sees usage concentrated on summarizing, translating, and retrieving information. That already saves time, but the bigger opportunity comes when systems can act on what they learn. His framework has three components: Intelligence, access to data, and the tools to interact with the world. Connecting those pieces allows AI to complete tasks in a loop, with varying degrees of human intervention.

Getting there starts with understanding your own business. Established companies cannot necessarily rebuild themselves around AI as easily as a young software startup, Rizk said. Giving AI access to business data is a useful starting point, allowing employees to request insights that previously required dashboards and specialist work. The next step is equipping it to take action.

Access to the tools is only part of the job. Visa has more than 1,200 AI champions worldwide helping employees develop their skills through demonstrations, hackathons, and practical applications, Soliman said. “AI literacy is as important as digital literacy is today.The starting point should be identifying the problem and the desired outcome, rather than adopting AI for its own sake.”

At Klivvr, agents are moving beyond improving existing processes. The company initially used AI models for underwriting, collections strategies, and fraud prevention. Its newer agents are helping create workflows and accelerate policies, collections strategies, and customer acquisition campaigns. Critical areas, including underwriting and collections, still go through human approval, Bachtler said, although the company is testing how agents perform against its teams before giving them greater autonomy.

Customers are already interacting with AI. Klivvr’s customer-facing AI agent answered more than 40,000 inquiries in the month before the panel, Bachtler said — questions that would otherwise have gone to its call center. The company is also deploying an agent there and sees potential for AI to improve collections performance and customers’ understanding of repayment obligations.

Bachtler envisages agents handling customer campaigns, messages, and offers without individual approvals, alongside agents overseeing governance. A human layer would continue monitoring performance. For Rizk, deciding when to remove direct supervision requires evaluations: Measure the existing process, test the agent against that baseline, and establish whether it performs better. He cautioned that AI is not yet consistently comparable with the alternatives.

Your next purchase could be made by an agent. Soliman pointed to agentic commerce as a major application: Moving from AI that researches a holiday or compares products to AI that completes the purchase. “Visa Intelligent Commerce is designed to enable that transaction. But making the payment is the straightforward part,” he argued. “Establishing that the agent is acting on the customer’s intentions is harder.”

Businesses must distinguish legitimate shopping agents from fraudulent ones. Blocking every agent could cost them sales; admitting all of them increases risk. “Visa’s Trusted Agent Protocol is designed to help trusted agents communicate securely with merchants,” Soliman said. The questions now extend beyond verifying the cardholder to whether the agent is authorized and its actions match the customer’s intent.

As execution speeds up, planning becomes more important. Rizk expects more people to become planners and orchestrators of AI systems. But human planning may struggle to keep up with machines’ ability to execute. His expectation is that businesses will put more effort into experimentation and finding new applications. “I’m more optimistic about the future of businesses, and that they’re going to do more with AI, not less.”

The upside extends beyond large companies. Soliman sees AI opening up capabilities that once required substantial time, money, and effort to individuals, professionals, and small businesses. For product teams, it also shortens the distance between an idea and execution. Rizk is excited by how much easier it makes building something: The initial friction of getting started falls when you can simply ask a system to begin work. The concerns are about keeping up.

A double-edged sword. Although the panel was generally optimistic about the future of AI for businesses they did have some concerns. Bachtler worries education is lagging behind. “The education system still needs to catch up. Kids right now are not learning as much as they should, and this could have an effect in the future.” Soliman highlighted the gap between innovation and governance. Rizk pointed to the growing burden of reviewing AI-generated code and cyberattacks unconstrained by human working hours.

Tap or click here to read the panel’s full transcript.