A preliminary 2025 MIT Project NANDA report found that only a small share of the task-specific generative AI implementations it studied produced measurable business impact. Its frequently quoted “95%” concerns that studied GenAI context and a particular success measure. It does not describe every enterprise AI initiative, nor does it prove that 95% never reach production. The more useful question is why value fails to cross from a pilot into a working organisation.
I have spent the better part of two decades watching organisations attempt digital transformations of various kinds. The pattern is remarkably consistent. A leadership team gets excited about a new technology. Budgets are allocated. Vendors are selected. Pilots are launched with great fanfare. And then, somewhere between the proof of concept and the production deployment, the whole thing quietly dies. The vendor gets blamed. The technology gets blamed. Occasionally, the CIO gets blamed.
But the technology is rarely the problem. The problem is that organisations try to learn something new without first unlearning something old.
The First Law: Unlearn the Myth of the Silver Bullet
Every few years, a technology comes along that the market declares will solve everything. Cloud was going to eliminate IT costs. Blockchain was going to eliminate intermediaries. Now AI is going to eliminate inefficiency itself. The pattern is familiar: a genuine technological breakthrough gets inflated into a universal solution, and organisations rush to adopt it without asking the most basic question: what specific problem are we solving?
Imagine a leadership team asked to “implement AI across the organisation” because competitors appear to be doing so. If nobody can name the first decision, sponsor and intended business result, “do AI” is a reaction, not a strategy.
A useful strategy names the problem, defines the expected result and assigns an owner before selecting a model or supplier.
The first law of digital unlearning asks organisations to abandon the belief that any technology, including AI, is inherently transformative. Technology is an accelerant. It accelerates whatever is already there. If your processes are sound, your data is clean, and your people are aligned, AI will accelerate your advantage. If the processes are broken, the data is fragmented and people cannot challenge the output, AI may reproduce those problems more quickly and at greater cost.
The Second Law: Unlearn the Separation of Strategy and Governance
In most organisations I encounter, strategy and governance live in separate rooms. The strategy team dreams up ambitious AI roadmaps. The governance team writes policies that constrain them. The two groups meet occasionally, usually when something has already gone wrong, and the conversation is adversarial by design.
This separation is a relic of an older model of technology adoption, where the risk profile of a new system could be assessed once, at the point of deployment, and then monitored passively. AI does not work this way. A machine learning model that performs perfectly in testing can develop bias in production. A generative AI system that produces accurate outputs today can hallucinate tomorrow. The risk is not static. It evolves. And governance that is separated from strategy cannot evolve with it.
The second law demands that governance be embedded in the strategy itself, not bolted on afterwards. We use the metaphor of brakes on a car deliberately. Brakes do not exist to stop the car. They exist to allow the driver to go faster with confidence. A governance framework that is designed alongside the AI strategy does not constrain innovation. It enables velocity. It gives the board confidence to approve larger investments. It gives regulators confidence that the organisation is acting responsibly. It gives customers confidence that their data is being handled with care.
In the UAE, this is not a theoretical concern. The Personal Data Protection Law is maturing. The EU AI Act applies in phases and may reach certain providers or deployers outside the EU when their AI outputs are used in the Union; applicability must be assessed for the specific system and role. The DFSA is actively surveying AI adoption within the DIFC. Organisations that treat governance as a separate workstream from their AI strategy are building on foundations that will not hold.
The Third Law: Unlearn the Outsourcing of Capability
This is the one that makes consulting firms uncomfortable, and I say that as someone who runs one.
Some traditional consulting models can create dependency. A longer engagement generates more revenue, while a more complex deliverable can make it harder for a client to maintain the work independently. This model has worked for decades because the knowledge gap between the consultant and the client was wide enough to justify it.
With AI, that model can become actively harmful when it leaves the client unable to operate or improve the capability independently.
AI is not a system you install and walk away from. It is a capability that must be continuously managed, refined, and governed by the people who understand the business context in which it operates. An external consultant can design your governance framework. They can build your initial models. They can train your teams. But if, twelve months later, your organisation cannot operate and evolve those systems independently, the engagement has failed, regardless of how polished the final presentation was.
The third law of digital unlearning requires organisations to stop outsourcing capability and start building it internally. This means selecting partners who measure their success by your independence, not by the length of their contract. It means investing in your people, not just in technology licences. It means accepting that the most valuable deliverable from any AI engagement is not a model or a dashboard. It is the institutional knowledge that allows your team to build the next one without picking up the phone.
The Path Forward
A pilot that cannot cross into sustained use is not inevitable. It is often a sign that organisations applied old mental models to a different kind of challenge. The enterprises that will succeed with AI in this region will be the ones willing to do the difficult work of unlearning before they attempt to learn. Many will, because the ambition and investment are both extraordinary.
Unlearn the silver bullet. Define the specific problem before selecting the technology.
Unlearn the separation. Embed governance into strategy from day one.
Unlearn the dependency. Build internal capability that outlasts any engagement.
The organisations that do this will deploy AI in a way that lasts. In a market moving as fast as the UAE, durable capability matters more than a short-lived launch.