My Advice to Every Business Leader Regarding AI
Talal Abu-Ghazaleh
I have spent more than five decades at the intersection
of commerce, technology, and institutional development. I have watched ideas
become industries, and I have seen empires dissolve. What separates the
enduring from the ephemeral is never the speed of adoption, but the quality of
judgment. I write this because I believe that judgment — the one irreplaceable
human faculty — is at risk of being abandoned at precisely the moment it is
most needed.
Artificial intelligence has entered the world with a
velocity unlike anything I have witnessed before. And unlike previous
technological waves, this one carries a peculiar danger: it produces outputs
that look intelligent. It speaks in full sentences. It cites facts. It offers
recommendations with apparent confidence. Because it sounds authoritative, far
too many leaders are treating it as though it were. They are not. They are
dealing with a system that has no conscience, no accountability, and no concept
of consequences, a system that, by its own designers’ admission, remains in its
infancy.
What concerns me most is not the technology itself, but
the human response to it. Across industries, organizations are adopting AI not
because they have identified a genuine need, but because they fear being
perceived as behind. Fear has never been a sound strategy. When a company
implements AI to signal modernity rather than to solve a real problem, it does
not gain a competitive advantage it accumulates a quiet liability. It builds on
ground it does not fully understand, toward outcomes it cannot reliably
predict.
The evidence is already accumulating. In software
development, a well-documented pattern has emerged: AI systems generate code
that passes every unit test and appears structurally sound, yet the resulting
application is three and a half times larger in memory and performs two
thousand times more slowly than the original, completely unusable in any
production environment. The AI succeeded by every intermediate measure and
failed catastrophically by the only one that mattered. This is what happens
when organizations measure progress by the volume of output rather than the
quality of outcomes.
The problem extends far beyond software. AI systems are
producing research reports that sound authoritative while containing invented
citations. They generate financial analyses with internally consistent logic
built on factually incorrect premises. They offer legal summaries with
misapplied precedents. In each case, the output looks like professional work.
In each case, uncritical trust in that output creates liability. A global
accounting firm was required to refund a government client in Australia after
an AI-generated report contained material errors that would have been caught by
even basic human review. This was not a small firm. It was a global institution
with vast resources and experienced professionals. That it fell into this trap
is not an indictment of AI. It is an indictment of the governance failure that
allowed AI outputs to be delivered as professional work without adequate
oversight.
One of the most consequential shifts underway is what I
call the democratization illusion. It is celebrated that non-technical staff
can now build software, automate workflows, and generate analyses that once
required years of specialized training. In some respects, this is a genuine
achievement. But it also means that organizations are now deploying systems
built by people who cannot audit them, cannot debug them, and cannot foresee
their failure modes. These systems will not announce their vulnerabilities.
They will function silently until they do not. When AI-generated layers are
added to complex infrastructure without rigorous governance, the risk does not
merely add; it compounds invisibly.
The deeper danger, however, is philosophical. AI speaks
with fluency. And fluency, in human psychology, has always been a powerful
proxy for credibility. We are wired to trust confident, articulate voices. AI
exploits this tendency without intending to — it has no intentions at all — and
the result is that its outputs are too often accepted without scrutiny. In
consulting and professional services, incentive structures accelerate this
problem: partners are rewarded for revenue, directors for reducing costs, and
associates for speed of delivery. In such an environment, AI-generated work is
not reviewed, it is passed through. It moves from model to client without a
knowledgeable human ever truly owning responsibility for it.
The financial sector that specializes in pricing risk has
already begun to respond. Insurance underwriters are actively exploring how to
exclude AI-generated work from professional liability policies. Some are
pressing regulators for explicit carve-outs. When the institutions whose entire
purpose is the accurate pricing of risk begin withdrawing from a category,
business leaders should treat this as a serious signal. Insurance companies do
not retreat from profitable markets without cause. They are telling us
something we should hear.
A reckoning is coming. Organizations that have deployed
AI without governance frameworks, without clear accountability, without
meaningful human review at critical checkpoints, will face it. They will face
legal challenges from AI-generated errors presented as professional
deliverables. They will face reputational damage when those errors surface
publicly. They will face pricing pressure as clients demand fee reductions upon
discovering that work once billed at the rate of expert human judgment was in
fact generated by an AI system in minutes. This is already happening. It is not
a theoretical future — it is the present, advancing.
I speak with particular concern for our region. The Arab
world is at a pivotal moment in its institutional development. Many of our
governments, enterprises, and professional bodies are still building the
frameworks — legal, regulatory, and cultural — that more mature economies spent
decades constructing. In that context, adopting AI without governance is not
merely risky; it is potentially generational in its consequences. If our institutions
embed AI into their foundations before those foundations are sound, the errors
will be structural, not incidental. The Arab world has an opportunity to lead
in responsible AI deployment — to build governance-first rather than
governance-after. That requires our business leaders to be more deliberate, not
less, than their counterparts elsewhere. We cannot afford to learn these
lessons the expensive way.
At Talal Abu-Ghazaleh Global, we have approached AI with
both conviction and discipline. We believe in its transformative potential — we
have invested in it, built with it, and embedded it across our operations and
services. But we have insisted on governance: on human ownership of AI outputs,
on review processes, on institutional accountability. We have built training
programs not to teach uncritical reliance on AI, but to teach people to use it
with wisdom and rigor. Because a tool of this power, deployed without wisdom,
is not an advantage. It is an accelerant for error.
There is a debate raging about whether AI will eliminate
jobs. I believe this debate, while important, distracts from a more fundamental
question: not whether AI will replace workers, but whether it will replace
thinking. An organization can survive losing headcount. It cannot survive
losing the capacity for independent judgment. I have seen what happens when
institutions hollow out their intellectual core — when they mistake the
execution of instructions for the exercise of wisdom. It takes years to build a
culture of rigorous thinking and very little time to dismantle it. If leaders
allow AI to become a substitute for thought rather than a support for it, they
will find themselves, within a decade, presiding over organizations that are
technically capable and intellectually empty.
My advice to every business leader is this: adopt AI, but
with discipline. Use it as you would any powerful instrument, with full
awareness of its limitations, with oversight at every critical juncture, and
with the clear understanding that accountability cannot be outsourced to an
algorithm. The winners of this era will not be those who adopted AI the
fastest. They will be those who adopted it with the greatest intelligence,
governed it with the greatest rigor, and preserved — above all else — the
irreplaceable quality of human judgment.
In practice, this means four things. First, never deploy
AI in a workflow without designating a named human who owns accountability for the
output, not the tool, not the team, but a specific individual. Second,
establish review checkpoints proportional to the consequence of error: the
higher the stakes, the deeper the human review must be. Third, train your
people not just to use AI, but to interrogate it to ask what it might have
missed, what assumptions it has embedded, and where it has substituted
confidence for knowledge. Fourth, measure AI's contribution to your
organization not by cost saved or hours reduced, but by whether the quality of
your decisions and the integrity of your outputs have improved. Speed and
efficiency without quality and accountability are not gains. They are deferred
losses. The future belongs to those who know how to combine human wisdom with
technological power. Not to those who mistake the appearance of intelligence
for the substance of it.