The country’s chief economist expects employment to rebound in the fourth quarter of 2026, helped by increased economic activity during the Christmas season, after the unemployment rate rose to six percent in July, its highest since June 2022.
Department of Economy, Planning, and Development (DEPDev) Secretary Arsenio M. Balisacan told reporters on the sidelines of last week’s Luzon Economic Corridor (LEC) Investment Forum that some of the economic headwinds experienced in the third quarter could also subside toward the end of the year.
To recall, the number of jobless Filipinos climbed to 3.14 million in July, the highest since December 2021, when 3.27 million were unemployed, according to the latest Philippine Statistics Authority (PSA) data.
The Cabinet-level Development Budget Coordination Committee (DBCC), composed of President Ferdinand R. Marcos Jr.’s economic managers, has also raised its 2026 unemployment rate projection to 5.3 to 5.8 percent from the previous forecast of four to five percent, according to documents on the proposed ₱7.2-trillion 2027 national budget.
Against this backdrop, Balisacan said he views the growing use of artificial intelligence (AI) as an “opportunity to get into higher value,” instead of it replacing actual people.
The DEPDev chief added that closer collaboration with other sectors and upskilling could help address the high unemployment rate even as companies increasingly use AI for certain jobs.
Beyond employment, Balisacan is advocating a measured expansion of AI across the public sector, where the technology could make government monitoring and decision-making faster and more responsive without replacing human judgment and institutional accountability.
During last week’s 13th National Monitoring and Evaluation (M&E) Network Forum, Balisacan said AI could help the government monitor programs more frequently, process field information more quickly, recognize patterns across large and complex data sets, and identify implementation concerns earlier.
“Used well, these capabilities can shorten the distance between evidence and action. An early warning can prompt a timely course correction. Information drawn from multiple sources can help decision-makers see where implementation is lagging and why,” Balisacan said in his speech.
However, he stressed that faster processing and more sophisticated technology do not necessarily produce better evidence.
“Speed is not the same as insight. An output is not automatically evidence merely because it was generated by a sophisticated system. Credible evidence still depends on sound questions, appropriate methods, reliable data, careful interpretation, and an honest acknowledgment of limitations,” he said.
“AI can support this process, but responsibility cannot be delegated to an algorithm. Human judgment and institutional accountability must remain at its center,” Balisacan added.
For government, Balisacan said adapting to AI should go beyond acquiring new software. It requires reliable data systems, personnel who understand the technology and can critically evaluate its outputs, clear lines of responsibility, and safeguards for privacy, fairness, transparency, and the integrity of public evidence.
Many government M&E systems, however, continue to contend with fragmented reporting, uneven data quality, manual processing, and delays in delivering information.
“AI may help address some of these limitations, but it cannot repair weak data or a poorly designed process. Applied without care, it may allow errors to travel faster and become harder to detect,” Balisacan said.
“Our starting point, therefore, must be the decision or implementation problem we are trying to address—not the technology we are eager to deploy,” he added.
Balisacan also warned that AI-supported systems could reproduce errors and biases embedded in their data and design, potentially producing seemingly precise results that are nevertheless wrong or perform poorly for particular groups or communities.
“Meaningful human review is therefore indispensable. Officials and practitioners must understand what an AI-supported process has done, examine the basis for its results, challenge findings that appear implausible or incomplete, and remain accountable for the decisions that follow,” he said.
“AI can support monitoring and evaluation, but AI itself must also be monitored and evaluated,” Balisacan added.
“We must assess whether AI-supported systems achieve their intended outcomes, work under actual operating conditions, perform fairly across groups and locations, and deliver benefits that justify their costs and risks,” he said.
Before scaling an AI-enabled tool, Balisacan said agencies should define the problem and expected results, test the system, document its limitations, and specify who is responsible for review and corrective action.
Agencies should also determine what data are being used, whether these are accurate and representative, who may access them, and how privacy and security are protected.
“Ultimately, our progress should be measured by whether these technologies help us produce better evidence, make better decisions, use public resources more effectively, and achieve better development outcomes,” Balisacan said. – with reporting from Danielle T. Bayani