The Store and the System: Why Philippine Businesses Stall Between the AI Pilot and the Payoff

Manila from Harbour Square. The capital to modernise Philippine operations exists; the question is what it gets spent on. Photo: Aeron Oracion on Unsplash

The prototype worked. The client said so. It was, by the client’s own account, exactly the inventory system the business needed. Then the quote for building the real thing arrived, and the conversation went where Bien Herrera says it often goes.

“They’re going to say, ‘Oh, if it’s that much, I’ll just build another store, because that’s predictable for me.’”

Bien Herrera, Managing Partner and Head of Philippines, Embiggen X¹

Herrera runs the Philippine side of Embiggen X, an AI and data systems firm serving mid-market companies in Switzerland, Poland, Singapore and the Philippines. He told the story in June at the first episode of Grit Chronicles, the founder conversation series from Starting Block Ventures. It drew a knowing laugh. It should also worry anyone who cares about Philippine productivity.

Choosing the store is not foolish. A store is a known quantity. But repeated across thousands of boardrooms, the choice becomes a habit of putting capital wherever returns are easiest to forecast. In the age of AI, that is rarely where they are highest.

A company that nearly ran out of road

Herrera did not found Embiggen X. He joined a year or two in, straight out of Ateneo de Manila University, thinking it was a consulting firm. On day one he found a business model canvas on the wall and realised it was a startup. He applied as an intern, was hired as an analyst, and worked his way to partner and country head.

The company changed course twice along the way (Figure 1).

The middle step nearly killed it. In 2023 the firm went all in on corporate venture building and landed a big client. Within months it was clear the market was not buying. “Revenue wasn’t flowing,” Herrera said, “but our overhead was the same.” A team that had peaked at about 50 full- and part-time staff shrank until only Herrera and the CEO were left (Exhibit 1).

Read as a case, three things stand out. The pivots were adjacent: each kept the firm’s core skill and changed only the customer, which is why the company could survive losing its revenue. The crisis came from operating leverage, not a bad idea: a fixed cost base built for a market that showed up late. And in the end, survival came down to one person deciding not to leave.

The cheapest moment to take a risk

At the low point, Herrera went six months without pay while corporate recruiters called. He stayed, and his reasoning was pure economics.

“I realised there was no other point in my life where I could absorb this much risk.”

Bien Herrera, Embiggen X¹

Asked what he still uses from his economics degree, he named one idea: opportunity cost. His decision shows why the concept is subtler than the textbook version. The real question is not only what you give up, but when giving it up costs least. Finance has a name for this. Real options theory holds that the value of keeping a door open grows with uncertainty and with the time you have left to walk through it.² At 23, with no dependants, Herrera traded a modest salary for a long-dated option on his own skills.

That has a social edge. Many young Filipinos are breadwinners early, and for them the same bet puts a household budget at risk. If the Philippines wants more people to take Herrera’s bet, it has to make the bet cheaper, not just applaud those who can afford it.

The Batad rice terraces in Ifugao were built by hand over generations: compounding is a process, not an event. Photo: Omri D. Cohen on Unsplash

Why the store wins the meeting

Embiggen X now works in three phases: find the priorities, prove value with a prototype, then deploy. An audience member asked the question every firm in the business asks itself. Why do so many projects die between phase two and phase three?

The prototype is rarely the problem, Herrera said. His team can build one in a day, and clients like it. The problem is payback. A full build costs a known amount today and returns an uncertain amount over one to two years. A new store costs a known amount today and returns something the owner can estimate. Many choose the store.

Three forces make the system look worse in the boardroom than it is in reality.

Familiar risk feels safer. Decades of decision research show that people prefer a bet whose odds they know over one whose odds they don’t, even when the unfamiliar bet is no worse on average.³ An owner who has opened ten stores can guess what the eleventh will sell. Nobody in the room has deployed this AI system before.

The numbers lag the value. Economists studying general-purpose technologies have found that their payoff depends on hard-to-measure investments in new processes, skills and data. Early on, that makes productivity look worse than it is; later, better.⁴ A finance team judging a system on year-one results is measuring the bottom of the curve.

The books favour bricks. A store goes on the balance sheet as an asset. Much of what makes an AI system work, such as training and process redesign, is expensed as it happens.⁵ One looks like an asset, the other like a cost.

None of this is uniquely Filipino. An MIT study of enterprise generative AI, drawing on hundreds of interviews, surveys and public deployments, found that roughly 95 percent of corporate pilots produced no measurable profit-and-loss impact. Its authors traced the failures mainly to poor integration with how companies actually work, not to weak models (Exhibit 3).⁶

The lesson is not that AI projects fail. It is that projects judged like stores fail the test, and projects run like software installs fail in practice.

“A glorified Google search”

Herrera was blunt about how companies grade themselves. Leaders think they have adopted AI because staff use ChatGPT, Claude or Gemini every day.

“If you’re using it that way, it’s a glorified Google search.”

Bien Herrera, Embiggen X¹

Real adoption, in his definition, means rebuilding processes to run on AI, not “slapping AI on top” of old ones. Think of it as a ladder (Exhibit 4). Each rung demands more change than the last, and most of the value sits at the top.

Climbing it is a leadership job before it is a technical one. Give a team that lives on paper reports a new ERP system, or even a spreadsheet, Herrera said, and many will simply not use it. Logic about efficiency does not change habits. Direction from the CEO, the board and the COO does. “Leadership mindset comes first.”

The MIT research points the same way. Projects built with specialist partners succeeded roughly twice as often as purely in-house builds, and the best returns came from unglamorous back-office work rather than the sales and marketing tools that attract most budgets.⁶ For a mid-market Philippine firm, the message is simple: start where work is repetitive and measurable, and borrow expertise rather than building it.

The polite answer problem

The sharpest question of the night came from the floor. Embiggen X starts by diagnosing operations, interviewing floor managers and frontline teams, not just the C-suite. But hiya, a sense of shame and propriety, and pakikisama, going along to keep the peace, run deep in Philippine workplaces. Staff often hide bottlenecks to avoid friction. How does a one-week assessment get past the polite answer?

Herrera admitted his team is still learning. Employees assume the questions are a prelude to layoffs. What works is presence: showing up at the office every day, talking, and making clear the conversation carries no risk. “It’s not written anywhere,” he said. “It’s just how you build a genuine human connection.”

That is more than a soft skill. Filipino psychology has long described these values as part of a culture that prizes smooth relationships, which makes workplaces humane but filters out bad news.⁷ Management research makes the complementary point: teams surface problems only when people believe candour is safe.⁸ Feed an AI system the version of a process people are comfortable telling you, and it will automate that version, including the steps nobody actually follows. Herrera’s broader line applies here too: trust is not transactional. It is built in the relationship, and the quality of the diagnosis depends on it.

The same choice, at national scale

Scale the store-or-system dilemma up and it describes the Philippines’ IT and business process management industry. Its industry association projected about 1.9 million jobs and US$40 billion in export revenue for 2025, rising to about US$42 billion in 2026, and listed responsible scaling of AI among its top priorities. Its chief executive has warned that the formula that built the industry will not carry it through the next phase.⁹

The IMF estimates that about a third of Filipino workers are highly exposed to AI. Roughly 60 percent of them are in roles where AI is more likely to help than replace them. Outsourcing carries the largest share of jobs at risk of displacement (Exhibit 5).¹⁰

Herrera framed the challenge as capturing AI’s gains “without it destroying our current industry.” Our view: the outsourcing model is a store. It grows predictably by adding seats. The next phase means selling systems, meaning redesigned processes delivered by fewer, more skilled people and priced on outcomes. Providers that make that shift keep the client. Those that bolt AI onto existing seats will watch clients do the redesign themselves.

The upside reaches beyond services. Herrera pointed to agriculture, where neighbouring countries already use AI in ways the Philippines mostly does not. His prescription was not invention but hyper-contextualisation: take what works elsewhere and adapt it to local crops, farm sizes and supply chains. Local knowledge is the moat.

The same bias shows up in capital markets. Herrera argued that Philippine investors want short-term returns from ventures that take six to ten years, and too often send money abroad rather than into local companies solving local problems. Our earlier paper, Fewer Bets, Bigger Cheques, found the same pattern in the deal data.

What to do about it

• Boards and CEOs: put an executive with authority over the process, not just the IT budget, in charge of each AI programme. Measure cycle time, error rates and cost per transaction, not logins.
• CFOs and investors: fund deployment in stages, each with a milestone and a rule for stopping. That turns one unfamiliar bet into a series of smaller, known ones. Show payback as a range, and judge systems over years, not quarters.
• Operators and young professionals: take career risks when your obligations are smallest, and earn trust beyond the deliverable.
• Industry bodies and policymakers: aim reskilling at the roughly one in five workers whose jobs AI can complement, fund the adaptation of proven regional AI tools to Philippine farms, and track how deeply firms redesign processes, not merely whether they use AI.

The process is the point

Asked for the one lesson he hoped the room would keep, Herrera skipped technology entirely. Find your why, he said, and keep going back to it. Earlier he had put the same idea in operating terms.

“If I’m always chasing the result, I’m forever disappointed, because the goals will always move. But if I focus on the process, something will happen.”

Bien Herrera, Embiggen X¹

That is the difference between the store and the system. The store is a result: visible, countable, predictable. The system is a process: slow to show and hard to measure, but it compounds. Philippine business has become very good at buying results. The companies that learn to invest in processes will be the ones that turn the country’s exposure to AI into an edge.

Notes and references

1. Starting Block Ventures, Grit Chronicles Episode 1: Bien Herrera (Embiggen X) in conversation with Winston Binauhan, vOffice Sun Plaza, 26 June 2026. Session recording; quotations lightly edited for clarity.
2. A. K. Dixit and R. S. Pindyck, Investment under Uncertainty, Princeton University Press, 1994.
3. D. Ellsberg, “Risk, Ambiguity, and the Savage Axioms,” Quarterly Journal of Economics 75(4), 1961, pp. 643–669.
4. E. Brynjolfsson, D. Rock and C. Syverson, “The Productivity J-Curve: How Intangibles Complement General Purpose Technologies,” NBER Working Paper 25148, 2018; American Economic Journal: Macroeconomics 13(1), 2021.
5. Philippine Accounting Standard 38, Intangible Assets (equivalent to IAS 38), on the expensing of training and research-phase costs.
6. MIT NANDA, The GenAI Divide: State of AI in Business 2025, as reported by Computing, August 2025.
7. V. G. Enriquez, From Colonial to Liberation Psychology: The Philippine Experience, University of the Philippines Press, 1992.
8. A. Edmondson, “Psychological Safety and Learning Behavior in Work Teams,” Administrative Science Quarterly 44(2), 1999, pp. 350–383.
9. Context.ph, “IT-BPM marches toward 2M jobs, exports milestone,” 30 December 2025, reporting IBPAP projections and remarks by IBPAP President and CEO Jack Madrid.
10. M. Cucio and T. Hennig, “Artificial Intelligence and the Philippine Labor Market: Mapping Occupational Exposure and Complementarity,” IMF Working Paper WP/25/43, 21 February 2025.

This case study is based on remarks made at a public Starting Block Ventures event and on the published sources cited, which are summarised in our own words. Company figures are as recounted by the speaker and have not been independently verified. Exhibit 2 is a stylised illustration, not based on any client or company data. Derived figures, including the split in Exhibit 5, are Bull Run Consulting calculations. The views expressed are Bull Run Consulting’s and do not necessarily reflect those of Embiggen X, Starting Block Ventures or the speakers. This paper is for information only and does not constitute investment, accounting or legal advice.