AI can catch you up on a meeting that you missed or joined late. But can it keep a refinery running, lower a factory’s electricity bill, or change the economics of an entire industry? For companies moving beyond experimentation, those are increasingly the questions that matter. The fundamental question that business owners are currently grappling with is rethinking how their businesses operate — and deciding how AI can help them do things differently and how much responsibility they can safely hand over to the bots.

That was the focus of the company panel at the EnterpriseAM Egypt Forum — the AI edition. Microsoft’s Middle East and Africa Growth Markets General Manager Mirna Arif, Capgemini Egypt CEO Hossam Seifeldin, and Qalaa Holdings founder and Chairman Ahmed Heikal discussed where AI is delivering value, what is holding adoption back, and why some of Egypt’s competitive advantages could come under pressure. You can read more on that in our post-conference coverage here.

The productivity gains are already part of the workday according to our panelists. Arif uses Copilot to summarize documents, prepare customer briefings, and catch up on meetings without asking colleagues to repeat themselves. There is also a more relatable application: Taking the frustration out of an email before hitting send. Her prompt? “Please make it more professional and make sure I don’t get fired.” Beyond those tasks, she sees agents taking on parts of workflows, requiring executives to think through permissions and how work should be organized.

For Heikal, the business case comes down to the cost of getting things wrong. “Every day a refinery is offline I lose USD 5 mn,” he told the audience. If a refinery goes down, it needs seven days to come back up, I lose USD 35 million in those seven days. So I don’t have the luxury of doing it wrong,” he said. Predictive maintenance is one application across his businesses, alongside automation, agriculture, and cybersecurity.

At Egyptian Refining Company (ERC), Qalaa’s USD 4.3 bn refinery, an AI pilot is looking for trouble before equipment fails. ERC’s Head of Digital Transformation Ramy Harfoush explained that the agent monitors operating data — including pressure, temperature, vibration, and flow — for patterns associated with historical failures. When it identifies a warning sign, it assembles previous maintenance orders, spare parts requirements, inventory availability, procurement lead times, and root cause analysis into a package for engineers to assess.

The engineer still makes the call. The aim is to allow the maintenance team to validate the findings and plan an outage before an unexpected breakdown forces one. “This doesn’t only save time, it keeps people on top of things,” Harfoush said. The pilot links an early warning to the information needed to act on it, giving engineers a head start on decisions that directly affect the refinery’s margins.

AI is also helping companies make more efficient use of energy. Ahmed Imam, the CTO of ASEC Automation, Qalaa’s engineering and industrial automation business, described an energy management application for cement grinding, an area he said consumes around 40% of a plant’s power. It learns from control system data and recommends operating parameters to improve energy consumption. The system provides “the best parameters and best values, or set points, for the operator,” Imam said. At Dina Farms, Qalaa’s agriculture arm, Heikal pointed to drones flying above the fields to identify crop problems like changes in color, to help determine fertilizer or pesticide needs and potentially improve yields.

In customer service, AI is already taking work off human agents’ desks. Capgemini’s “AI buddy” listens to service desk calls, summarizes them, and suggests solutions while the conversation is happening,” Seifeldin said. “It also updates the knowledge available for future calls. With one client, 56% of traffic was being handled without human intervention. Technology used to be an asset that depreciates with time, and AI is the first thing that is actually appreciating with time,” he said, pointing to the knowledge accumulated as systems operate.

The next step is changing business processes. With most projects Seifeldin still sees a focus on speed and productivity, but companies are starting to explore how AI could reshape business processes and business models. That requires more than access to a capable tool: Cybersecurity, data, regulation, and existing workflows all need attention. Capgemini is training teams across technologies so they can apply AI to areas including coding, project management, implementation, and modernization.

BUT- Legacy systems can stop that transformation before it starts. Inconsistent data, old applications, and undocumented code make it difficult for established businesses to capture AI’s full value. “You cannot run an AI system on such a platform. You need to have the right foundation in order to get the right outcome,” Seifeldin said. Startups have an advantage because they can choose that foundation from the outset. For larger organizations, modernization becomes part of the investment required to make AI useful.

Some responsibilities should stay with people. Arif frames adoption around three questions: What can AI accelerate, what can be delegated to agents, and what must remain human? A customer briefing can be automated, but she still wants the account owner or country manager to “read the room.” Difficult conversations with employees also remain a human responsibility and they will likely always remain so, according to Arif. “I will never delegate to an agent or an AI a difficult conversation with a team member, for example,” she said.

Leadership has to signal that AI use matters. Heikal, who sits on NYU’s President’s Global Council, brought roughly 250 people together with NYU professors for a week in Marassi on Egypt’s North Coast earlier this year to explore what’s coming with AI. Setting aside that time was itself a message. “If I am taking one week off for this, then it must be important. Everybody should try to get the best implementation.There are the local flows and the local processes but also, sometimes you may find that you are able to commercialize an agent, and that can bring some value over time,” said Heikal.

Cybersecurity is a particular concern given the potential damage from attacks on industrial businesses. “Cybersecurity [attacks] can kill you,” said Heikal. His role, he said, is to encourage adoption while putting ethical and legal safeguards in place.

For Egypt, the opportunity extends from finance and tourism to agriculture and education. Arif sees the dividing line running between companies willing and able to transform within each industry. Regulation, workforce training, and industry readiness need to advance together. Seifeldin sees scope for AI to help the public sector, including education, where agents could take on some teaching functions as the country grapples with teacher shortages and school construction costs.

But AI could also erode advantages Egypt has spent years building. Seifeldin warned that wider AI adoption could weaken the language skills advantage underpinning the call center industry. Heikal’s concern extends to robotics: If machines can perform textile work around the clock at lower unit labor costs, production could shift back toward mature markets. That prospect already affects his investment thinking. “There are certain classes of industries where I would be very, very hesitant to invest,” he said.

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