B12 Magazine

Daily from the four markets B12 Group works in

Technology

How to adopt AI in your company: a guide for leaders

Where AI pays off, how to prepare your data, how to govern the risks, whether to build or buy, and how to measure the return: a clear guide for leaders in the UAE.

GuideB12 Group10 min read

Every board in the Gulf is now asking how to adopt AI in the company, and most are asking the wrong first question. The question is rarely which model or which vendor. It is which decisions and processes in the business would be meaningfully better if a machine could read, summarise, draft, classify or predict at scale, and whether the organisation has the data, the discipline and the governance to let it do so safely. Companies that start there tend to see results within months. Companies that start with a tool tend to collect pilots that never reach production.

This guide is written for chief executives, owners and senior managers in the UAE. It covers where AI creates value, how to prepare your data, how to govern the risks, when to build and when to buy, how to measure the return, and how the UAE's national AI strategy, charter and ethics guidelines should shape your approach.

Why the UAE context matters

The UAE was early to treat artificial intelligence as national policy. The government launched the UAE Strategy for Artificial Intelligence in October 2017, describing it as the first of its kind in the region, with targeted sectors that include transport, health, energy, water, education and technology. Since then the federal government has created a UAE Council for Artificial Intelligence to oversee adoption across government and education, introduced the role of chief AI officer in ministries and federal entities, and published a guide to generative AI covering around one hundred use cases.

For private companies, this context has two consequences. First, customers, regulators and government counterparts in the UAE increasingly expect AI to be used, and used well. Second, there is a clear national statement of values on how AI should be developed and used, which gives leaders a ready-made reference for their own policies. A company that aligns with those values early will find later regulation easier to absorb.

Start with use cases, not technology

The most reliable way to find value is to look for work that is high in volume, heavy in documents or language, and governed by rules that can be written down. In most companies, that points to a familiar list.

Customer service and sales support: answering routine questions in several languages, drafting replies for staff to approve, summarising conversations and routing enquiries to the right person.

Document-heavy operations: reading contracts, invoices, customs papers, insurance claims or onboarding files, extracting the key fields and flagging anything missing or unusual.

Knowledge work: drafting reports, proposals and policies, searching internal documents in natural language, and translating between Arabic and English with human review.

Analysis and forecasting: demand planning, pricing support, risk scoring and the detection of anomalies in transactions or operations.

Software and IT: assisting developers with code, tests and documentation, and helping IT teams triage incidents faster.

For each candidate, write down the current cost, time and error rate, who owns the process, what data it depends on, and what would happen if the AI made a mistake. Rank candidates by value and by risk. The ideal first project is valuable enough to matter, contained enough to control and visible enough to build confidence across the organisation.

Data readiness: the foundation most companies underestimate

AI systems are only as good as the information they can reach. Before committing budget, leaders should ask four plain questions. Where does the relevant data live, and is it in systems the AI can access securely? Is it accurate, complete and current, or will the model learn from errors and outdated records? Who is allowed to see it, and will the AI respect those permissions? And is any of it personal data, which brings legal obligations?

The last question matters a great deal in the UAE. The federal Personal Data Protection Law, Federal Decree-Law No. 45 of 2021, applies to the electronic processing of personal data, requires organisations to keep that data secure and confidential, restricts processing without consent outside defined exceptions, gives individuals rights to correct and to restrict processing, and sets requirements for transfers across borders. Latham & Watkins notes that the law requires lawful processing of personal data in AI applications. Companies in free zones such as DIFC may fall under their own data protection regimes. Before feeding client or employee data into any AI service, establish where it will be processed and stored, whether it will be used to train someone else's model, and whether that meets your obligations. Check the current rule with the UAE Data Office, your free-zone authority or legal counsel.

Practical data readiness rarely means a grand data lake. It usually means cleaning and organising the specific data a first use case needs, connecting it to the AI through a secure interface, applying the same access rules that already govern staff, and keeping a record of what the system was given and what it produced.

Governance and risk

AI introduces risks that traditional software does not. Models can produce confident but wrong answers. They can reflect bias in the data they were trained on. They can leak sensitive information if prompts and outputs are not controlled. They can be manipulated by malicious inputs. And they can raise questions about intellectual property, both in the material used to train them and in what they generate. Latham & Watkins points out that the UAE's copyright law, Federal Decree-Law No. 38 of 2021, is relevant to third-party material used in training data and outputs.

Good governance does not need to be heavy. It needs to be clear. Appoint an accountable executive for AI. Keep a register of every AI system in use, what it does, what data it touches and who owns it. Classify each use by risk: a tool that drafts internal meeting notes needs lighter oversight than one that influences credit, pricing, hiring or medical decisions. For higher-risk uses, require human review of outputs, test for accuracy and bias before launch and periodically after, and document how decisions can be explained to a client or regulator.

International frameworks help structure this work. The NIST AI Risk Management Framework, released in January 2023 and intended for voluntary use, organises AI risk management around governing, mapping, measuring and managing risk, and NIST added a profile specifically for generative AI in July 2024. NIST notes that the framework is being revised, so use the current version. The OECD AI Principles, adopted in May 2019 and updated in May 2024, set out five values-based principles, including transparency and explainability, robustness, security and safety, and accountability, and have 47 adherents according to the OECD.

Security deserves its own line in the plan. AI tools connect to email, documents and business systems, which makes them attractive to attackers. The UAE Cyber Security Council lists a national cyber security policy for artificial intelligence in its policy library, and it is worth reading alongside your existing security controls. At minimum, decide which AI services staff may use, block the unsanctioned ones for sensitive work, and make sure your security team reviews any AI system that can take actions on its own.

The UAE AI ethics guidelines and charter

Leaders do not need to invent their own ethical framework. The UAE has published several reference documents. According to Latham & Watkins, the UAE AI Office issued an AI ethics guide in December 2022, a non-binding framework built around principles such as fairness, transparency, accountability, privacy and safety, and Digital Dubai has published its own AI ethics principles and guidelines covering similar ground, including security and sustainability.

More recently, the UAE Charter for the Development and Use of Artificial Intelligence, published in 2024, sets out twelve principles. The UAE Government portal lists them as strengthening human-machine ties, safety, algorithmic bias, data privacy, transparency, human oversight, governance and accountability, technological excellence, human commitment, peaceful coexistence with AI, promoting AI awareness for an inclusive future, and commitment to treaties and applicable laws. Its stated objectives include ethical and responsible use, protection of privacy and data security, and greater transparency and accountability. Latham & Watkins describes the charter as non-binding and as supporting the national AI strategy.

Non-binding does not mean irrelevant. These documents describe what UAE authorities consider responsible practice, and sector regulators have built on them. Latham & Watkins notes, for example, that financial regulators including the Central Bank of the UAE, the Securities and Commodities Authority, the DFSA and the FSRA have adopted non-binding guidelines for financial institutions using enabling technologies such as AI, and that the Dubai Health Authority's policy on AI in healthcare requires independent validation, risk assessment and clear disclosure. A practical approach is to translate the charter's principles into a one-page internal AI policy, map each principle to a concrete control, and review it annually as the rules evolve.

Build, buy or partner

Most companies should buy before they build. Mature AI capabilities are now available inside the software businesses already use: office suites, customer relationship systems, service desks and accounting platforms. Activating these features under proper governance is often the fastest and cheapest route to value.

Building makes sense when the use case is central to your competitive advantage, depends on proprietary data, or requires integration that off-the-shelf products cannot provide. Even then, building rarely means training a model from scratch. It usually means combining a commercial or open model with your own data through retrieval, connecting it to your systems, and wrapping it in the controls your business needs.

Between the two sits partnering. A capable technology partner can help select models, prepare data, design the architecture, integrate securely and hand over a system your team can run. When choosing one, ask where your data will be processed, who owns what is built, how the system will be monitored for accuracy and drift, how security is handled, and what happens if you change providers. Insist on a clear exit path and on documentation. Firms such as B12 IT Solutions, which design, build and secure business systems, illustrate the value of treating AI as part of the wider architecture rather than as a standalone gadget.

Measuring the return on AI

AI projects fail most often not because the technology does not work, but because no one defined what success meant. Before launch, record the baseline: how long the task takes today, what it costs, how many errors occur, how satisfied customers and staff are. Then measure the same things after deployment, over a period long enough to smooth out novelty.

Useful measures include time saved per transaction, throughput per employee, reduction in errors or rework, faster response times, higher conversion or retention, and avoided costs. Count the full cost on the other side of the ledger: licences, computing, integration, data preparation, training, oversight and ongoing maintenance. Be honest about hours saved: they create value only if they are redeployed to more useful work.

Treat the first projects as a portfolio. Some will exceed expectations, some will disappoint, and a few should be stopped quickly. Set review points at thirty, ninety and one hundred and eighty days, and be willing to scale the winners and close the rest.

People and change

The human side decides whether AI sticks. Staff need to know what the tools are for, what they must never be used for, and how their own roles will change. Training should be practical and specific to each function. Encourage teams to propose use cases, because the people closest to the work usually know where the friction is. And be clear that responsibility for decisions remains with people, which is also a central theme of the UAE charter's emphasis on human oversight.

A first ninety days

In the first month, appoint the accountable executive, agree a short internal AI policy based on the UAE charter, list the AI tools already in use across the company, and shortlist three candidate use cases with their baseline costs and risks.

In the second month, prepare the data for the strongest candidate, confirm where it will be processed and that this meets your data protection obligations, choose whether to activate an existing product, build or partner, and run a controlled pilot with a small group of users and clear human review.

In the third month, measure the pilot against its baseline, review security and accuracy findings, train the wider team that will use it, and decide formally whether to scale, adjust or stop. Then repeat the cycle with the next use case, carrying forward what the first one taught you.

The B12 view

Adopting AI well is less about technology than about management: choosing the right problems, preparing the right data, setting clear rules, measuring honestly and bringing people along. The UAE offers an unusually supportive environment, with a long-standing national strategy, a published charter and practical ethics guidance that any company can adopt. Leaders who start small, govern carefully and scale what works will find that AI becomes an ordinary, dependable part of how their business runs, which is exactly where it creates the most value.

More from Technology

Call us free