Earlier this year, the UAE set a public deadline: move half of all government services to agentic AI within two years, and train every federal employee to work alongside it. Some entities have committed to 75%. The clock is already running, and it changes what "good with AI" means for anyone whose work touches a form, a ticket, or an approval.
In this month's Pulse Live conversation, Nexford CEO Fadl Al Tarzi sat down with Ayman Itani, founder of Think Media Labs, adjunct professor of AI strategy at Hult International Business School, and an advisor to UAE government entities on their agentic AI rollout. The session was unscripted and question-led, and the audience kept both of them busy.
If you missed it, here are the five takeaways worth acting on.
Ayman drew the line simply. Generative AI is a conversation: you ask, it answers, you do the work. Agentic AI takes the action. Submit a form with a missing document and the machine checks it, spots the gap, and writes back the same day asking for the file. That step used to sit in someone's inbox for days.
Half of UAE government services running this way means every process behind them gets reviewed: which steps still make sense, which manual workarounds crept in, which systems need to talk to each other, and which requests still escalate to a human. Ayman called it "immense work," and it is happening daily.
The Lesson
Learn the mechanics of a process before you learn the tools. That review work is where the jobs are.
Both speakers kept returning to this. Ayman has walked into many projects labeled "agentic AI" that turned out to be about process, data, budget, and who is authorized to approve what. Fadl shared a telco example from his Dubai days: thousands of angry inquiries a day, and the brief was "answer faster." The real problem was that customers had bought a service that never got provisioned. No response time fixes that.
Ayman's own habit when advising a business is to skip the slide decks and sit in the customer service department listening to calls. The complaints point to billing, product quality, and broken handoffs. Customer service is the frontline that tells you where the problem is; it is almost never the problem itself.
The Lesson
Fix the process first. Put AI on a broken workflow and you get a faster broken workflow.
"If you're going to build the software on a process that's inefficient, that's not going to help anyone."
— Fadl Al Tarzi, Founder & CEO, Nexford University
Ayman is a computer and communication engineer, so this carries weight coming from him. AI now handles much of the technical background work, which leaves the human gaps: frustration inside a company, frustration from users, problems a middle manager has been trying to solve for years. The person who steps into those gaps needs to be interested in talking to customers, mapping who does what, and understanding how a process moves from step one to step three.
This is the AI Translator role Fadl described at the top of the session. In his analogy, the AI engineer builds the engine, the data analyst reads the gauges, and the AI Translator decides where the vehicle needs to go, then explains the output to the people who have to act on it.
The Lesson
Business process analysis, customer insight, and the ability to explain AI output to a decision-maker are the skills in demand. Named tools like Python, SQL, and automation help you speak the engineer's language; they are not the entry ticket.
Fadl raised a loan-application example from the Q&A: an applicant marks "business owner," so route them to the business-owner process. That is an if-then rule with zero operating cost and zero variability. Agentic AI is probabilistic and it costs money, so using it there makes no sense.
Ayman added a case from his own team. A 12-slide client proposal came back grammatically correct, logically sound, and impossible to follow. The excellent solution they had sketched on a whiteboard had vanished into AI-polished prose. He now asks his team to skip AI on work where the point is to say something clearly and directly.
The Lesson
Deterministic logic beats a model when the rule is known. Reserve AI for the judgment work it is good at, and say so when it is not the right tool.
LLMs predict the highest-probability next word. Fadl noted that as models improve, their confidence in wrong answers has climbed with them. Ayman's filter: any answer that starts by telling him he is right gets ignored, and every source link gets opened before he relies on it.
Neither check works without subject matter expertise, because without domain knowledge you cannot tell a good report from a convincing one. For fresh graduates facing fewer entry-level jobs, Fadl's advice was to build critical thinking, curiosity, and learning how to learn. Critical thinking in practice means asking which parts of your problem the model cannot see (the product, the pricing, the market, the regulations) and adding that context yourself.
The Lesson
Pick one or two real problems at work or at home and run them through the tools yourself. You will find the limits faster than any course or conference will show them to you.
"You need a layer of subject matter expertise to do a good job with AI. Otherwise, you're coming in quite weak."
— Ayman Itani, Founder, Think Media Labs
On jobs at risk: Repetitive work AI already does well, such as translation, tidying English copy, drafting social posts, and freelance front-end page builds. The bigger risk Ayman sees is the person who refuses to touch the tools out of fear or cynicism and misses the warning signs.
On the skills that stay human: Reading the room in a meeting, speaking in public, and keeping real working relationships. Ayman stuttered as a kid and now speaks on TV weekly, so he counts public speaking as a trained skill.
On customer experience careers: Fadl suggested CX professionals look at product management and design. Years of first-hand customer pain points are what a product designer is trying to acquire, and AI now lets that person mock up a fix, test it with customers, and write the spec for the tech team.
On content and SEO: Ayman compared today to the early 2000s, when nobody knew how to rank on Google yet. The new skill is writing so that AI agents crawl your site, take an action, and recommend you.
"AI is not as scary as you see on social media, and AI is not as magic as you see on social media."
— Ayman Itani, Founder, Think Media Labs
Ayman's closing advice to new entrants was to stop holding out for a big-name employer. Two years as a junior in a 12,000-person company often teaches less than two years with real responsibility in a 35-person one. Take the role where you will own a process end to end, then bring the tools to it.
You can watch the complete live session with Ayman Itani on-demand here.
The AI Translator is the role behind every takeaway above: the person who understands the business objective and can direct AI to serve it. Nexford's BS in AI in Business (BSAIB) and MS in AI (MSAI) are built for exactly that job. Book a call to see which path fits where you are now.
The UAE gave itself two years. Pick a process you know, run it through the tools this week, and find out what breaks.