Many companies have embraced artificial intelligence (AI), but getting those projects into day-to-day operations remains a bigger challenge than adopting the technology itself.
Thinking Machines said that the problem is rarely the AI models or the technology behind them. More often, companies struggle because they choose projects that do not solve meaningful business problems. Thinking Machines Data Science, a Temus entity, is a Philippines-founded AI and data transformation company headquartered in Manila and operating across Southeast Asia.
Since 2015, Thinking Machines has served more than 150 clients, trained more than 12,000 professionals to apply AI, and delivered enterprise AI and data systems across financial services, retail, conglomerates, and civic organizations.
“I think there’s not so much of a technical bottleneck,” said Niek van Veen, VP for Growth at Thinking Machines. “The technology is there. Technology is never the issue. A lot of technology is already in place.”
He said organizations that already use cloud platforms such as AWS or Microsoft Azure have access to security controls, governance frameworks, and role-based access controls needed to deploy AI.
Instead, he said many businesses start with AI pilots that are difficult to justify or measure.
“The pilots they run are, maybe not the right pilots that prove business value,” van Veen said. “It’s about finding a big enough business problem and being able to measure that business problem. Once you have identified that, that’s when you scale it.”
Van Veen said companies should focus on one or two high-impact use cases before rolling AI out across the organization.
“Find one or two use cases that have enough pain that AI can solve,” he said. “All the downstream technical things you need to do, you can solve them. Other companies have solved them, and we have helped other organizations solve them.”
Van Veen said changing AI models should not discourage companies from adopting AI. According to van Veen, newer models generally improve performance, although businesses using AI in production should expect additional testing whenever providers retire older models.
“The models are just being replaced,” he said. “Whatever workload you are running on these models, they will just change, so you might need to change the instructions a little bit. Typically, they work just fine. Actually, the work is even better when a model gets upgraded.”
The bigger cost comes when companies have to switch to a different model after an existing one is no longer supported.
“When models change, or especially when they are deprecated, companies need to change to a new model and go through another round of testing. That’s the main cost associated with this constant change of models when agentic systems are in place,” he said.
As AI projects grow, van Veen said organizations should also invest in stronger data foundations.
“Data foundations are absolutely critical,” he said. “Once you start using more powerful models, you will quickly find that you need your data foundations in place. Investing in a data platform, data management, and data governance is absolutely critical to make AI produce the best outcomes for the organization.”
Thinking Machines began as a data company before expanding into AI consulting, giving it experience across software development, analytics, machine learning, and large language models. That background, van Veen said, helps the company recommend the right technology for each business challenge.
“It always starts with a business problem,” he said. “Sometimes AI is the solution. Sometimes agentic systems are the solution. Sometimes it’s just a simple script. Sometimes it’s a data platform. You need to be aware of that full spectrum.”
For organizations eager to scale AI, van Veen advises them to solve a real business problem first. Once companies can clearly measure the value AI delivers, the technical work of putting it into production becomes much easier.