Across the Gulf, small and medium businesses, from trading companies in Deira to contractors in Riyadh, are being told that AI will transform them. Many owners are rightly sceptical: they have limited time, no technical team and a business to run. The good news is that AI adoption for an SME is more a management project than a technology one, and it can be done carefully and cheaply if the sequence is right.
Start with a problem, not a technology
The most reliable way to waste money on AI is to begin with a tool and look for somewhere to use it. Start instead with a list of tasks the business does often, that take meaningful time, and where mistakes are tolerable. Answering routine customer enquiries, drafting quotations, summarising supplier documents, categorising expenses, forecasting next month's stock needs and screening job applications all qualify.
From that list, pick one. The best first use case is usually unglamorous. It is frequent, so savings add up. It is well understood, so you can tell when the AI gets it wrong. And it does not touch anything where an error would cost a customer or a licence. A Dubai trading company that uses AI to draft first versions of quotations from a product catalogue, with a salesperson reviewing every one, is a model first project.
Check your data before you check the tools
AI works from data, and the honest question is whether yours is usable. For most SMEs it lives in accounting software, spreadsheets, email threads and messaging apps, often in two languages. That is normal and rules nothing out, but it shapes what is realistic.
Ask three questions. Do we have examples of the task done well, such as past quotations or answered enquiries? Is the data in a form a system can read, or is it photographs of invoices and voice notes? And do we have the right to use it, particularly where it contains customers' personal information, which UAE and Saudi data protection rules govern? A week spent tidying data for one use case is usually the best-spent week of the whole project.
Buy, configure or build?
For nearly every SME, the answer is buy or configure. Modern AI assistants, customer-service tools, document-processing services and forecasting features inside accounting and inventory software are mature, inexpensive and improving quickly. Many can be set up with the company's own documents and rules in days. Custom models make sense only when the task is unique, the volumes are large and the economics are proven, which is rare in the first year.
Two cautions apply. Check where the tool stores data and whether that is acceptable to your customers and regulators. And avoid tools that cannot export your data; the market is changing fast and you will want the freedom to switch.
Measuring value and bringing the team along
Before the pilot starts, record a baseline: how long the task takes, how often errors occur, how many enquiries go unanswered. Measure the same things after thirty and sixty days. Value for an SME is usually one of three things: hours returned to staff, mistakes avoided, or revenue won through faster responses. If none of these moves, the use case was wrong, and that is a cheap lesson.
The people who do the task every day will find uses no consultant can imagine, but only if they are involved from the start and are not afraid the tool is there to replace them. Be clear about purpose. Give two or three staff a few hours of hands-on training, let them experiment and ask weekly what is working. In a bilingual workplace, make sure the tool handles Arabic as well as English before anyone relies on it.
Common mistakes to avoid
- Starting with the hardest problem because it is the most exciting.
- Letting the AI act without a human check in the first months.
- Measuring nothing, then arguing about whether it helped.
- Treating the pilot as an IT project rather than an operations project.
- Pasting confidential customer data into tools whose data policy nobody has read.
A 90-day roadmap
Days 1 to 15: list candidate tasks, choose one, record the baseline and name an owner who is not from IT. Days 16 to 30: tidy the data for that task, select a tool, confirm its data handling, and train two or three staff. Days 31 to 60: run the pilot with every output reviewed by a person, collect feedback weekly and fix the obvious problems. Days 61 to 90: measure against the baseline, decide whether to expand, stop or adjust, and if it worked, choose the second use case using the same method.
Ninety days is enough to learn whether AI has a place in your business and where. It is not enough to transform the company, and it should not try. Businesses that take this measured path build something more valuable than a single tool: the habit of adopting new technology without drama, one well-chosen problem at a time.
