Last year, I wrote about the hybrid space between a person and an AI and named it kasangkapwa (companion in the journey toward shared identity). Watching my students build with agentic AI this trimester, I realized I need to pay attention to a different relationship.

These past two weeks, my undergraduate class in Management in the Digital World held its final presentations. Six groups, each embedded with an actual partner organization, mostly micro- and small enterprises, each running an action research project around an AI-assisted prototype. Group projects involved prototyping customized queuing and point-of-sale systems, sales pipelines, inventory forecasting, entrepreneurial incubator dashboards, knowledge management systems, and invoice processing. These are all undergraduate business students. Almost none of them code.

Two years ago, this output was out of reach. One year ago, it would have needed an engineering elective and a semester of suffering. This trimester, with agentic AI platforms that plan, execute, and revise with minimal supervision, it needed a good brief, a few weekends, and the discipline to check the machine’s work. 

Mid-presentation, an Eraserheads line from their song “Para sa Masa” (For the Masses) lodged in my head and would not leave: “Pinilit kong iahon ka, ngunit ayaw mo namang sumama.” (I tried so hard to save you, but you didn’t even want to go with me.)

The surprise of the trimester was that the scarce skill turned out to be restraint. Group after group had to be pulled back from burying a small business under features nobody asked for. One team’s first cycle failed exactly this way. They arrived with a polished solution; the owner received it politely and did not use it. The prototype worked. The relationship, initially, did not; until further alignments were done. 

The bottleneck has moved

For as long as I have supervised student consulting projects, the limit on what a project could deliver was the student. Analysis, design, execution: all rationed by skill and time. Our pedagogy is calibrated to that scarcity. We teach students to build better because building was the hard part.

Agentic AI dissolves that scarcity, and the bottleneck relocates to the other side of the table. The constraint now is the partner organization’s capacity to absorb change: the store owner’s hours in a day, her trust in three undergraduates she met in June, her willingness to alter routines that have fed her family for a decade. My students could suddenly build far past her readiness. That gap is where projects now fail.

From a pair to a system

Last year, in this column, I proposed the word kasangkapwa for AI’s in-between status: more than kasangkapan, a mere tool, yet incapable of the loob that makes another being kapwa (shared identity). I treated it then as a two-party matter, a person and a machine in a hybrid space, and I argued that the discipline it demanded was interior: dual awareness, active discernment, keeping your own voice.

My students’ projects seemed to outgrow that pairing. What actually operated in the class was a system with four seats. The student group, whose members had to divide roles and re-explain to one another work an agent did overnight. The agentic AI, no longer just a conversation partner but a worker that runs unattended. The professor (myself), needed as a third set of eyes on the process running between students and machine. And the partner organization, the one seat with no interest in the technology at all, only in ginhawa: relief, breathing room, a slightly easier day.

Kasangkapwa, it turns out, cannot remain a private virtue. Once the AI can act on the world, the ethics of using it extend past your interior state and into what your ease of building does to other people.

Mabait na pamimilit

The Eraserheads line is funny until you occupy the owner’s chair. Being handed a system you never requested, by student consultants who anticipated a need you have not yet felt, is a mild form of coercion no matter how good everyone’s intentions are. Pamimilit na nakabalot sa kabaitan. (Coercion wrapped in kindness). And agentic AI has made this coercion nearly free to produce. The feature costs nothing to add. The burden of refusing it, of explaining the refusal, or of absorbing an unwanted system falls entirely on the person we claim to be helping.

I do not exempt myself. I require my students to use AI, and some have called this out in my teaching evaluations. My defense is that they will compete with and alongside these systems the moment they graduate, and exposure under supervision beats discovery under pressure. But the same logic that makes me push my students is the logic my students used on their partners. Alam ko kung gaano kadulas ang daang iyon. (I know how slippery that path is.)

Negotiating the depth of change

The team whose first cycle flopped is the reason I remain hopeful. Their failure bought them something no feature could: the owner saw that they could take no for an answer. Trust followed. In the second cycle, the owner asked for more than they had originally dared to propose, and the point-of-sale system that finally shipped was co-constructed, fitted to routines, adopted because of trust between the group and the owner. 

The lesson is to rethink pakikipagkapwa-tao (human relations) in the advent of agentic AI. My student groups reflected: negotiate the depth of change; agree on it with the partner instead of delivering it to them. Match the intervention to the partner’s readiness rather than to the machine’s capacity, because that capacity is now effectively unlimited and has stopped being a useful guide to anything. Underneath sits pakikiramdam (empathy/sensitivity), sensing what the other can receive, which no agent can perform on anyone’s behalf.

What the professor is for

My own role shifted more than I expected. The students rarely needed me for technical answers; the agent outclasses me there. What they needed was accompaniment through the process: the first-month exercise of articulating desired outcomes for the individual and the group, consultations in and out of class, a colleague who co-sat the project defenses so a second pair of human eyes read what the students and their machines had produced together. Formative touchpoint after formative touchpoint, ending in an oral defense, because standing in a room and answering an unrehearsed question remains the one deliverable no agent can complete for you.

Two rules now govern the class. First, everything a student builds with AI is a prototype. It stays a prototype, however finished it looks, until people who know the domain and carry the stakes have validated it, and the validator who matters most is the partner, hindi ang propesor (not the professor). Second, we watch the process and not only the product, because a finished-looking product now proves very little about the learning behind it.

It seems fitting to end this with another Eraserheads line: “Gusto mo bang sumama?” (Do you want to join?) Kasangkapwa, a year later, is now about pakikipagkapwa-tao (human relations) augmented by the agentic capabilities of generative AI. Should this relationship lead to Alapaap (clouds), that is for us humans to decide. – Rappler.com

Patrick Adriel “Patch” H. Aure, PhD, is an associate professor at the Department of Management and Organization and founding director of the PHINMA-DLSU Center for Business and Society, Ramon V. del Rosario College of Business, De La Salle University. Email him at patrick.aure@dlsu.edu.ph.

Here are other Good Business columns on AI:

[Good Business] Steering generative AI towards nondisruptive creation

[Good Business] Humanistic national development in the era of AI

[Good Business] AI as a social good