On the distance between what leaders believe about AI adoption and what their teams actually report, and the national research now measuring it.
By Eli Harrell, CEO, EmergePH
Ask almost any Philippine organization how their AI adoption is going and you will get a number. Licenses purchased. Seats activated. Attendance at the enablement session. Percentage of staff with access.
Ask the same organization whether behavior changed, and the room goes quiet.
This is not a failure of effort. It is a failure of measurement. We have built a national conversation about AI on top of numbers that describe procurement rather than practice. A license count tells you what an organization bought. It tells you nothing about whether a single person spotted a problem in their work and reached for AI to solve it without being told to.
That distinction is not academic. It is the difference between an investment that compounds and one that quietly stops.
The framework already asked the right question
In her recent articulation of the EUREKA framework, Atty. Jocelle Batapa-Sigue makes a point that deserves more attention than it has received. She argues that progress should not be measured only through the number of tools deployed or institutions using AI, and that knowledge-driven governance means evidence before automation.
The framework is right. The problem is that the evidence does not currently exist.
There is no Philippine dataset describing how AI is actually behaving inside our organizations. We have adoption surveys that count usage. We have global studies calibrated to markets that do not share our labor structure, our management culture, or our particular relationship between a team and the person above them. What we do not have is a behavioral baseline for our own market.
You cannot govern what you have not measured. You cannot build workforce policy on assumptions about behavior that nobody has tested. And you certainly cannot tell whether the national AI agenda is working if the only instrument available reports how many people were given access to something.
So we built the instrument.
What actually decides whether adoption holds
The AI Behavior Benchmark is original research measuring how Filipino teams and leaders turn AI spend into P&L impact and better lives for people. It is free, it is open to any organization in the country, and it is authored and run by EmergePH.
It does not measure tool usage and it does not measure sentiment. It measures five behaviors.
Ownership. Spotting a problem in your own work and using AI to solve it without being instructed to.
Modeling. Leaders making their own AI use visible to their teams, including the attempts that did not work.
Experimentation. Trying a new way of working and treating a failure as information rather than as embarrassment.
Learning autonomy. Teaching yourself the next skill instead of waiting for the organization to schedule it.
Value articulation. Being able to describe your work in terms of the value it creates rather than the tasks it contains.
Underneath those five sit three conditions that determine whether any of them are even possible.
Whether admitting you used AI carries a professional cost. Whether people believe that using AI makes them more valuable or more replaceable. And whether leadership has said, specifically and out loud, what the AI investment means for people’s jobs.
Most AI programs install the first five and never touch the last three. In our experience that is precisely where adoption goes quiet, and it goes quiet in a way that no dashboard will ever surface, because the behavior it produces is concealment rather than refusal.
The gap you cannot see from your own seat
Here is the part that makes this research different from a survey.
The benchmark asks leaders and their teams the same questions, then measures the distance between the two sets of answers.
That distance is the finding. In most organizations, leadership and the frontline are not arguing about whether AI matters. They already agree that it does. What they disagree about is what is already happening. A leader reads the rollout, the licenses, and the energy in the room and concludes that adoption is landing. The team is quieter about what they actually do, what they hide, and what they are not confident enough to attempt where anyone can see.
Both groups are answering honestly. Both are describing the same organization. The two descriptions do not match.
Early pilot readings across a small number of Philippine organizations suggest the widest divergence sits on Modeling, where senior leaders rate the visibility of their own AI use dramatically higher than their frontline colleagues do. There is a second pattern worth watching in the middle layer, where managers report high willingness to learn alongside low sense of mandate to act. That combination describes people who want to move and do not believe they have permission.
These are early signals from a small pilot, not published findings. They are the reason the full benchmark exists, and they are exactly the kind of claim that should not enter policy conversation until it rests on a real sample. That is what we are building now.
What we are asking of Philippine organizations
The benchmark takes about eight minutes to complete. Individual answers are never shown to anyone, and nothing is ever reported from a group smaller than five.
For leaders, there is a second thing worth knowing. Anyone who takes part receives a private link for their own organization. When four or more of their people answer through it within seven days, we return a private readout comparing what that leader believes is happening against what their team reports, behavior by behavior. Together with the leader’s own response, that makes the group of five that our reporting threshold requires.
There is no cost for this, and there is no obligation attached to it. We are asking for participation because the research needs a real national sample, and because an organization that has seen its own gap is better equipped to do something about it than one working from assumption.
The instrument is open now at emergeph.com/ai-behavior-benchmark. The benchmark publishes in November 2026.
Why this matters past any single company
There is a version of the next few years where the Philippines measures its AI progress the way it measured its early digital progress, in units of access. Devices distributed. Connections established. Seats filled. Those are worth counting. They were never sufficient, and we learned that slowly and at some cost.
There is another version where we insist on behavioral evidence from the beginning. Where a regional council, a national agency, or a company board can ask not how many of our people have AI, but whether the behaviors that make AI valuable are actually present, and where the honest answer is a number rather than an anecdote.
The second version requires somebody to do the unglamorous work of building the baseline. We would rather that baseline be Filipino, built on Filipino organizations, and published in the open where anyone can argue with it.
Ethics defines intent. Knowledge strengthens decisions. Between those two sits the measurement, and until now, we have been governing without it.
https://emergeph.com/
Eli Harrell is the CEO of EmergePH, a Philippines-based behavior change consultancy, and the host of the Lead Human podcast. The AI Behavior Benchmark is EmergePH’s flagship research, open to any organization in the Philippines at emergeph.com/ai-behavior-benchmark.
