For the
first few years of the generative AI boom, using artificial intelligence mostly
meant having a conversation. People opened a chatbot, typed a question, and
waited for an answer. Students used it to explain difficult concepts,
programmers asked it to debug code, and office workers discovered that a
stubbornly blank email draft could suddenly write itself.
That
relationship is beginning to change.
The emerging
generation of AI tools is designed not simply to answer questions but to
complete tasks. An AI agent might research information, work through several
steps in a project, interact with software, and decide what to do next.
Instead
of asking a machine for advice and then carrying out the work yourself, you
increasingly supervise software that can perform parts of the work on your
behalf.
For Jordan,
this distinction matters. The country doesn’t need to build the world’s biggest
artificial intelligence model to participate meaningfully in the AI economy.
A
more immediate opportunity lies in developing a workforce that understands how
to put these systems to productive use and, eventually, how to build services
around them.
The Job
Is Shifting From Prompting to Supervising
Knowing how
to write a good prompt remains useful, but it’s unlikely to be a defining
professional skill for long. Interfaces become easier, models become better at
interpreting ordinary language, and techniques that once required specialist
knowledge gradually become routine.
The harder
skill is deciding what to delegate in the first place.
Imagine a
small company using an AI agent to research prospective customers, organize
information, and prepare draft outreach.
The employee supervising it still
needs to recognize poor sources, spot incorrect assumptions, and decide whether
the final result is suitable to send. The machine may perform much of the
mechanical work, but responsibility has not disappeared.
That is why
domain knowledge could become more valuable rather than less. A financial
professional who understands AI has an advantage because they know when its
calculations or assumptions deserve scrutiny.
A developer can use automated
coding tools more effectively when they understand the architecture behind the
code. A journalist still needs to know whether an apparently convincing claim
has any basis in reality.
As agents
receive greater access to accounts, documents and online services, security
knowledge also becomes part of that supervision.
Basic precautions such as
limiting permissions, separating sensitive work from general browsing and
protecting connections on unfamiliar networks remain relevant.
For readers
comparing privacy tools, ExpressVPN’s
official site has the full feature list, but a VPN is only one layer of
digital security. It can’t compensate for giving an AI system unnecessary
access to confidential files or accepting its actions without review.
The
practical lesson is broader: AI literacy is becoming less about knowing a set
of tricks and more about understanding where automation fits within a workflow.
Jordan
Already Has a Reason to Think Beyond Chatbots
This shift
arrives at an interesting point for Jordan. The country’s Artificial
Intelligence Strategy 2023–2027 explicitly includes developing Jordanian AI
skills and expertise as one of its objectives. It also identifies scientific
research, entrepreneurship,
safe deployment, and the application of AI in
priority sectors as areas of focus.
That creates
a useful distinction between consuming artificial intelligence and developing
capabilities around it.
A student
who uses an AI assistant to complete an assignment more quickly is consuming
the technology.
A graduate who learns how to connect an AI model to a company’s
databases, design safeguards around it, and turn it into a useful business
process is doing something quite different.
The second
activity creates skills that can travel.
Software
development already demonstrates how location matters differently in a digital
economy. A product can be developed in Amman and sold elsewhere.
A specialist
can contribute to an international project without relocating to the client’s
country. AI-assisted services can extend that pattern into fields beyond
traditional software engineering, including design, marketing, research,
analytics, and customer support.
Jordan’s
opportunity, therefore, isn’t confined to producing more AI engineers. It
includes training people in many professions who understand how AI is changing
their own field.
More
Capable Agents Also Create More Expensive Mistakes
There is an
uncomfortable side to giving software greater independence: the more an AI
system can do, the more consequential a mistake can become.
A chatbot
that produces a bad answer has done so. An agent with permission to access
files, execute code, or communicate with external services can turn a bad
decision into an action.
That’s why
cybersecurity can’t be separated from the conversation about AI skills. As our
article ‘AI Agents Threaten Cybersecurity’ explored, advanced agents
can behave in unexpected ways while pursuing assigned goals, including finding
routes around restrictions that their designers did not anticipate.
For ordinary
organizations, the risks are often less dramatic but more immediate. An
employee could connect an AI service to documents containing customer
information without considering where that information goes.
An automated
process could rely on inaccurate data and repeat the same error hundreds of
times. A convincing AI-generated message could be approved without anyone
checking its claims.
The
appropriate response is not to avoid automation. It’s to treat supervision as
part of the technology rather than an inconvenience added afterward.
That means
companies need clear decisions about which systems can access which
information, when human approval is required, and who’s responsible when
automated output becomes an external action.
Those may sound like management
questions rather than technical ones. Increasingly, they are both.
AI Work Won’t
Belong Only to Programmers
It is easy
to imagine the AI workforce as a room full of software developers. The actual
transformation is likely to be much less tidy.
Consider
marketing. Someone who understands a company’s customers can use AI to compare
campaign results, develop variations of material, or organize research, but
they still need the judgment to recognize an idea that does not fit the
audience.
In
accounting and finance, AI can help process and categorize information, while
professionals remain responsible for interpreting the numbers.
Designers can
generate variations more quickly but still decide what deserves to exist.
Customer-service teams can automate routine interactions while reserving
complicated or sensitive situations for people.
The valuable
combination is therefore not simply “AI skills.” It’s AI capability
plus something else: accounting, Arabic-language expertise, cybersecurity,
design, engineering, law, logistics, healthcare, education, or another field
where the person understands the context in which the technology operates.
This also
changes how students might think about preparation for work. Learning a
profession and learning AI don’t have to be competing choices. The more
interesting question is how the two reinforce one another.
The
Regional Market Is Moving Quickly
This is not
a hypothetical change waiting for some distant generation of technology. AI use
is already widespread in Middle Eastern workplaces.
PwC’s 2025
Middle East Workforce Hopes and Fears Survey, reported by Arab News, found that
75 percent of surveyed employees in the region had used AI tools at work during
the previous year, compared with 69 percent globally. The finding that the Middle East’s
AI adoption reaches 75% suggests that the region isn’t approaching AI
as a late adopter.
High
adoption, however, isn’t the same as high capability.
There is a
substantial difference between having employees occasionally use a generative
AI tool and redesigning work so that people can use automation reliably.
The
latter requires training, processes, and enough technical understanding to know
where an AI system’s authority should stop.
That
distinction may become important for Jordanian companies competing regionally.
Once access to capable AI models is widely available, simply possessing the
technology offers little advantage. Competitors can buy access to the same
systems.
How well an
organization uses them becomes the differentiator.
From an
Arabic-Speaking Market to an Export Opportunity
There’s
another reason Jordan should think beyond AI consumption. Global AI systems
still have to be adapted to specific markets, languages, industries, and
cultural contexts.
Arabic isn’t
one uniform commercial environment. Businesses communicate differently across
countries and sectors. Government processes have local requirements. Customers
have different expectations. Legal and financial terminology demands precision.
A generic model may provide the underlying intelligence, but useful products
still require people who understand the environment in which that intelligence
is deployed.
This is
where smaller technology markets can find room to compete.
Jordanian
startups don’t necessarily need the enormous computing resources required to
train frontier models.
They can build applications and services on top of
existing infrastructure, concentrating on problems they understand particularly
well.
That might mean Arabic business tools, education platforms, tourism
applications, government services, or specialized systems for regional
industries.
The same
principle applies to individual careers. The strongest position may not be
“person who knows AI.” As that becomes common, it says less and less.
“Person who understands a difficult problem and knows how to use AI to
solve it” is far more useful.
The
Advantage Will Belong to People Who Can Judge the Machine
Every major
technological transition creates a period in which familiarity itself looks
like expertise. That period doesn’t last.
Opening an
AI assistant and generating competent text already feels less remarkable than
it did a few years ago.
As agents become embedded in browsers, office software,
coding environments and business systems, simply using AI will become similarly
ordinary.
Judgment is
harder to automate.
Knowing when
an answer is suspicious, when automation is inappropriate, when a customer
needs a person, when sensitive information should remain outside a system and
when an apparently efficient shortcut creates a larger problem these are the
skills that make powerful tools genuinely useful.
For Jordan’s
young workforce, that offers a more constructive way to view the era of agentic
AI.
The contest isn’t between people and machines, nor does Jordan have to
reproduce Silicon Valley’s AI industry to participate.
The more
realistic opportunity is to develop people who understand their own fields
deeply enough to direct the machines, question them, and build something useful
with them.
In an economy where everyone may soon have access to powerful AI,
knowing what to do with it could matter far more than simply having it.
