Mohammad Abu Sheikh and the Art of Being Understood

At sixteen, Mohammad Abu Sheikh wanted a place where young people could sit together and exchange ideas. He started Kitab in Jordan and added coffee. It is an appealingly practical response to the question of how to encourage intellectual life: give it somewhere to sit, and something to drink.

Years later, his work involves Arabic speech models, training data and enterprise software. The machinery has changed considerably. The interest in helping people reach knowledge, and in building around the way they actually communicate, is easier to recognize. His career keeps finding new obstacles between an idea and the person who could use it.

Abu Sheikh is now co-founder and CEO of CNTXT AI, the Abu Dhabi company he founded with Hassan Abu Sheikh. The business prepares data, develops AI products and helps organizations put them to work while retaining control of their information. The founder’s short description of himself is refreshingly economical: “I just like to build things.”

A place for ideas, with coffee

The café led to another opportunity. Kitab had a large kitchen sitting unused; Abu Sheikh turned that spare capacity into a cloud-kitchen business. There is something revealing about the sequence. An idle room became the beginning of another company. Opportunity arrived wearing an apron.

His subsequent ventures moved into education and audio. Lamha offered Arabic knowledge in short, accessible forms. For someone interested in how information travels, the attraction is clear: a useful idea should be able to reach a person outside a classroom, or fit into a day already occupied by other things.

He grew up in Amman, and the gap he saw in access to knowledge became a recurring business interest. His education ventures connected students and teachers; the audio work made learning available in smaller portions. These were ways of working with an audience he understood, in a language whose needs were often poorly served by the available tools.

Looking back across those businesses, Abu Sheikh describes a repeated encounter with local problems that required local work. Arabic content needed better access. Organizations needed usable information. The AI business grew out of that same territory, where the details of the region were part of the job from the beginning.

Arabic, as people actually speak it

Consider the customer at the other end of a voice service. That person has a dialect, a pace and a way of asking for something. They may move between languages. A system that handles a carefully delivered formal sentence still has to cope with an ordinary caller. The caller is unlikely to regard this as an interesting research problem.

Abu Sheikh’s answer has been to build around Arabic from the outset. Munsit, CNTXT AI’s voice technology, grew from demand for speech recognition that could handle the language on its own terms. He has described recruiting local expertise and assembling the data needed to reflect regional speech.

This makes the team itself part of the product. Linguists and dialect specialists bring knowledge that an engineering qualification alone cannot supply. They can judge whether a result fits its setting, whether a phrase carries the intended meaning, and whether the software understands a speaker who has made no special effort to accommodate it.

For Abu Sheikh, cultural context belongs throughout development. It affects the information collected, the people reviewing it and the standards applied to the result. Building from that position changes the assignment: the team has to understand what a successful interaction sounds like before it can decide that its model has produced one.

“We always build out of necessity.”

Mohammad Abu Sheikh

The early Munsit work also required a practical answer to the expense of preparing speech data. Abu Sheikh has described an automated process that generated, filtered and cleaned annotations for training. Scaling the material available to researchers meant changing how it was prepared. A language product begins well before anyone presses the microphone button.

Who holds the keys?

That concern with the material underneath AI leads into Abu Sheikh’s argument for sovereignty. He wants organizations to retain ownership and control of the data, infrastructure and models they depend on. For a buyer, these are questions about the continued operation of a system as much as its initial performance.

He has put the market-entry problem plainly: “You cannot drop a global stack into a new market and adapt later.” The choices made during construction can determine how much adaptation is possible. Language, deployment arrangements and the customer’s existing operations all enter the calculation.

This is why CNTXT’s work includes preparing data and delivering systems around particular workflows. A company may have plenty of information without having it organized for AI. Before an application can be useful, someone has to turn that accumulated material into something a model can reliably work with.

Abu Sheikh’s position gives the less visible work a central commercial role. Collection, annotation and evaluation may lack the theatrical appeal of a talking machine. They have the considerable advantage of helping determine whether the machine says something useful. In his account, that foundation is where the business begins.

A voice that can finish the job

In June 2026, CNTXT AI acquired Actualize, a company building Arabic voice agents for Gulf customers. The acquisition added conversational technology and workflow automation to its portfolio. It also brought Actualize’s co-founders into senior technical roles: Muhammed Shabreen as chief technology officer and Khalid Ghiboub as vice president of AI models.

The attraction is straightforward. Understanding a request is one stage of a service. Completing the booking, updating the record or carrying out the transaction requires connections to the systems where the work happens. Actualize’s capabilities extend the conversation into those actions.

For Abu Sheikh, the deal brought together language technology and people familiar with the demands of regional deployment. The emphasis was on moving customers from trials into services used during the working day. That is where a voice agent encounters its less glamorous examinations: unusual requests, existing software and the expectation that an agreed task will actually be completed.

The funding supplied another piece of that expansion. CNTXT announced a $60 million Series A co-led by AI71 and BlueFive Capital, with plans to support product development, new markets and deployments for enterprise and government customers.

AI71 already had a connection with the founder: it had acquired his earlier venture LocAI. Its return as an investor gave the financing an operating history behind it. The relationship linked an earlier application business with the next stage of his work on the data and systems supporting AI.

The other side of the table

Abu Sheikh also backs founders through SMPL. In February 2025, SMPL Fund I announced a $250,000 investment across AI and technology companies within the Qora71 portfolio. The announcement described a $10 million fund combining seed capital with practical support.

The arrangement gives his company-building interests another outlet. A founder needs money, but also encounters decisions for which experience is useful: who to hire, which customer problem to prioritize, and how to avoid discovering too late that growth has outrun the operation supporting it.

He has spoken candidly about that last problem. Earlier ventures taught him that rapid expansion could expose weaknesses underneath the business. At CNTXT, he emphasizes validation before scale. It is a less carefree philosophy than simply launching quickly, and one informed by having had to deal with what broke.

His approach to leadership likewise makes room for other people’s knowledge. The company needs technical skill and an understanding of local language and culture. The investment work extends that dependence on relationships beyond his own team, connecting founders with experience and contacts as well as capital.

The road now runs in both directions

In October 2026, CNTXT AI announced its first US office in the San Francisco Bay Area and reported more than $300 million in committed revenue, much of it from American customers. Those are company-reported commitments. The figure describes revenue commitments, distinct from recognized revenue.

The expansion gives Abu Sheikh’s argument an international test. A company that builds around Arabic and regional requirements can also serve businesses elsewhere that need those capabilities. The local knowledge becomes something customers abroad have a reason to buy.

CNTXT’s work now also includes training data for physical AI, collected from real-world activity. The subject has broadened from language into how machines learn to operate in varied settings. The familiar concern remains: the training material has to capture enough of the world in which the system will be expected to work.

Abu Sheikh wants the region to produce and export more of its own technology. He also wants younger founders to see a business built here and imagine building one themselves. Kitab began with people exchanging ideas over coffee. His current undertaking carries that interest into another setting, with more infrastructure, more capital and a rather larger bill for the equipment.