Presight’s data analytics help UAE’s biggest companies conquer AI fears

As the UAE carves out a role in developing artificial intelligence tools with Microsoft, Google and OpenAI, Abu Dhabi-based Presight is harnessing big data analytics to help governments, energy producers and other industrial giants sharpen their performance.

The $3.5 billion company got off the ground less than three years ago as a subsidiary of G42, the UAE’s umbrella AI firm owned by Sheikh Tahnoon bin Zayed’s Royal Group and backed by the Mubadala sovereign wealth fund. Now, after a March 2023 IPO, it’s expanding through acquisitions, most recently buying a majority stake in AIQ, the data arm of national oil company ADNOC.

In an interview with The Circuit on the sidelines of the World Governments Summit in Dubai, Presight CEO Thomas Pramotedham said the company aims to decipher AI for some of the largest state-owned enterprises in the UAE and elsewhere in the Middle East and Africa while breaking down hurdles to wider adoption. Before moving to Abu Dhabi, he was CEO of Esri Singapore, which helped build the city-state’s Smart Nation program using data analytics and digital know-how to improve government services.

“We fear something we don’t understand,” Pramotedham said. “We will continue to evolve AI but also bring other technologies and apply them in an impactful way, where the technology truly hits the ground running.”

The interview has been edited for length and clarity.

The Circuit: How would you describe Presight’s mission and where do you see areas for growth?

Thomas Pramotedham: In my view, we’re definitely on the cusp of a transformation era. Presight has a very strong foundation in national platforms, applied technology, applied big data analytics and smart cities. And then, you have generative AI coming in for disruption.

We’re a company based in the UAE, focusing on the Global South – countries in Africa, Central Asia, and ASEAN. We focus on applying technology. That’s a key part of it. I believe AI will become more accessible, especially with recent advancements in open-source models that are more efficient.

Presight always operates on a national platform – we’re not really a B2B company. We work with large state-owned enterprises, primarily in Abu Dhabi, as their digital transformation and AI partner. That’s the trajectory I see. We will continue to evolve AI but also bring other technologies and apply them in an impactful way—where the technology truly hits the ground running.

Where does Presight fit in with the global energy transition?

There are two parts of the story. Either the rise of AI, everyone’s talking about where we are going to find the energy to support computing and no less so, buying the chips and creating the infrastructure. So that’s one part of the angle that we put in. The second part is this: the move from current energy transition into renewables is a journey that is taken. Our AIQ venture with ADNOC is really driven by ADNOC ‘s vision of AI in energy and energy for AI. What do I mean by that? AI is going to help the current energy operation transform, optimize, and therefore, they should be able to do more with less – and how do you achieve the requirement that is needed without an increased carbon footprint.

Then on the other side, [with Presight’s power-generation] “IntelliGrid” businesses, it’s really a transition of digitization or transmission and distribution network. So a smart grid, when you put the two together, what we’re expecting to see is soon traditional and renewable energy will interact. So you get two sources of energy providing, whether it’s compute or powering a city. Then what you really need is microgrids that are driven by AI algorithms that regulate this transmission. So really minimizing the wastage of electricity or optimizing, you know, how individual households consume electricity.

What are the key challenges in deploying AI at scale in critical infrastructure sectors like energy?

We are seeing a higher adoption of AI. We did for ADNOC, we launched the ENERGYai Large Language Model. It’s a generative AI workflow that helps optimize reservoir selections, seismic cube analysis, and in many ways, through AI automation, improves ADNOC’s upstream functions and workflows. So, helping energy companies optimize for better energy production.

On the city side, on the consumption of it, to work with cities, like what we’re doing now with Azerbaijan and many other countries, is really introducing a smart grid into the transmission networks that a country has. So it’s the adoption of the two. Of course, not every country. Power is still an issue. Electricity is not accessible to almost 700 million people around the world. But really, how do we bring this transition and make it available to developing or underdeveloped countries? It’s really one of the key ways to bridge the electricity divide – not just the digital divide, but the electricity divide.

How have you been actually limiting the electricity divide?

One is the work we’re doing with Azeri Gas. So, smart meter digitization is a piece of work that I think will continue. Through that, the optimization of transmission and distribution of utilities in the country can be improved. So that’s the work we do with IntelliGrid, our joint venture.

For, like I said, the acquisition of AIQ, it’s creating base models that can help not just large national oil and international oil companies but also small and medium oil companies optimize their operations. Using the models that we’re building with ADNOC, the model is trained on AdNoc, it’s applied to ADNOC. But the foundation model can be taken out to help a medium-sized oil company improve their production, improve their operations.

What impact will your work for ADNOC have on the broader industry?

I think ADNOC is an example of a large energy language model with agentic workflows that is converging with what we’ve seen in the last 18 months. So in 2023 and 2024, you heard a lot about large language models, and then last year, you heard a lot about agentic models. Large language models are foundation models that support broad queries. But when you specialize down to industrial language, that’s where agentic workflows come in.

So I could do the same at ADNOC and financial services. I could do it for the supply chain. I could do it for transportation, and what will happen is you get AI trained in a particular vertical, and from then, you develop different agents that help that vertical optimize itself. For example, the requirements I need in port operations and supply chains are quite different from the models I need in regular multimodal transportation.