MANILA-based Thinking Machines Data Science is one of 30 companies worldwide recognized as an OpenAI Advanced Partner. The designation, announced at an Aug. 14 media briefing in Makati City and formalized in a joint statement four days later, expands the firm’s role in reselling ChatGPT Enterprise, ChatGPT Business and ChatGPT Edu to Philippine organizations. Founded in 2015, Thinking Machines says it has served more than 150 clients and trained 12,000 professionals since, with 95 percent of its artificial intelligence (AI) pilots reaching production. Panelists kept coming back to one thing that morning: whether all that activity was paying off for Filipino companies.
“We built Thinking Machines on the belief that the Philippines can build up and scale world-class AI capability,” said Stephanie Sy, the company’s founder and chief executive. Niek van Veen, its vice president for commercial, framed the Advanced Partner status as validation of a decade of work.
Both companies made the same case: access used to be the problem; now it’s what companies do with it. Mark Jeffrey, OpenAI’s director of partnerships for Southeast Asia and Taiwan, said the Philippines is already among the top 15 markets globally for ChatGPT users. Close to 40 percent of messages on ChatGPT’s consumer platform in the Philippines are work-related, according to OpenAI. Enterprise use is shifting toward delegated work, with Codex accounting for 64 percent of combined Codex and ChatGPT output tokens among enterprise customers as of June. Companies in the top 10 percent of AI usage now generate 8.3 times as many output tokens per active user as typical firms, up from 2.6 times in January. But OpenAI itself calls token volume “an imperfect measure of business value,” a proxy for depth of use, not proof of what that use is worth. One executive floated a companywide target: a 15 percent valuation increase within two years.
The discussion kept circling that split through the morning without naming it outright. Repeated advice for companies starting out: pick one narrow, painful workflow, run it for 60 to 90 days and prove the value before building an AI platform. Precious Lim, country manager for Amazon Web Services (AWS), said organizations that start with an AI strategy and try to build a platform first, before proving value on one problem, are the ones that get stuck.
That gap showed up outside the case studies, too. Earlier that day, Jek De Chavez-Hermida, Cebu Pacific’s director of data services, described the airline’s own path with Thinking Machines. Business units had been running disconnected AI pilots without a shared way to measure value, she said, until leadership required each initiative to carry a documented case tied to revenue, cost savings or productivity. She said the airline still has not settled who holds final authority over AI adoption and said it had run into added cost from the Bureau of Internal Revenue’s withholding tax treatment on its OpenAI enterprise subscription.
Thinking Machines pointed to two named clients as evidence that value can be shown. With RCBC, it said it paired business leaders with AI champions to build workflow-specific assistants ready for pilot deployment, citing a Net Promoter Score of 90. With NST Apparel, a manufacturing client, an AI-supported system for purchase document intake and email automation delivered a claimed 32 times faster turnaround on that specific workflow. Both figures come from Thinking Machines’ own materials, not the client companies directly, and the manufacturing case measures one narrow process, not the business as a whole. The reported workflow is much faster. It does not show what happened across the rest of the company.
Every slide Thinking Machines presented that day carried the tagline “a Temus entity,” alongside the “strategic investment” language Temus used when it announced the investment two weeks earlier, on July 30. Neither company disclosed the size of Temus’ stake or the resulting ownership structure. Temus chief executive Sng Ren Yeong called it a shared mission between “two teams that were each built by people who came home to build,” with Thinking Machines gaining Temus’ reach and resources, and Temus gaining a decade of AI delivery experience.
Also unresolved: how clients should weigh the two deployment paths Thinking Machines now offers: ChatGPT Enterprise directly or OpenAI’s models through Amazon Web Services’ Bedrock, which fits a company’s existing AWS setup.
Advanced Partner status gives Philippine organizations one more local channel to buy and deploy OpenAI’s tools, backed by a firm with a decade of delivery history and Southeast Asian reach through Temus. Cebu Pacific’s account shows what adoption can look like from the client side, with AI projects already underway while measurement and governance questions are still being worked out. The RCBC and NST Apparel numbers that anchored the case for impact — a Net Promoter Score of 90 and a workflow running 32 times faster — still come from Thinking Machines, not from RCBC or NST Apparel themselves. Until RCBC or NST Apparel puts those results on its own record, they remain Thinking Machines’ account of what its work achieved.