From career navigation to human work infrastructure.
The Philippines is our first laboratory for a broader problem: people lack good evidence about what they can do, realistic routes from where they are now, and the economics of the next move. AI is simultaneously creating demand for human judgment, evaluation and verification. Pukot and Sulog test whether those problems can share infrastructure without collapsing capability into destiny.
A dense concentration of route-choice problems.
Three signals matter because they describe distinct transition failures rather than one monolithic “jobs problem.”
Youth disconnection
PSA reported a 15.2% NEET rate among 15–24 year olds in July 2026.
Education-job mismatch
PIDS research estimated 39% of employed individuals were overeducated for their jobs, while over a quarter were undereducated.
Migration as a route
PSA estimated 2.19 million OFWs in 2024. Overseas work is not an edge case in the Philippine decision set.
These are population and labor-market indicators. They do not by themselves validate Pukot/Sulog demand, willingness to pay or product-market fit.
A route is a transition, not a personality label.
Each route should be evaluated on feasibility, economics, time-to-credibility, downside risk, reversibility, progression and the person's own preferences.
Degree → first work
Translate education and early evidence into credible first jobs without treating the degree title as destiny.
Degree bypass / middle-skill
Compare university with technical training, certification, apprenticeship, portfolio work or direct experience.
BPO → adjacent work
Carry forward service, operations, QA, training and support evidence into higher-value roles.
Remote / freelance
Identify whether current evidence supports remote work and the smallest gaps blocking credibility.
Work abroad / move / return
Compare net economics and progression, not headline salary alone.
Career change without reset
Find transferable capability, adjacent buyers and the smallest real proof gap.
Disengagement → re-entry
Explore low-friction re-entry for populations poorly served by conventional pipelines.
Entry into AI human-work
Use bounded tasks to establish reliability, then progress toward more valuable task classes.
The relevant export unit is a repeated labor-market condition: educated underemployment, high NEET rates, migration-heavy household economics, nearshore/BPO economies, informal work with weak portable evidence, or long-term labor-force detachment. Local labor data and buyer economics must be rebuilt market by market.
The ILO projected a global youth NEET rate of 20.4% for 2024, around 259 million young people. That establishes the scale of the transition problem, not a single global customer segment.
The payer already exists. The moat still has to be earned.
AI companies already buy evaluation, preference data, annotation and verification. That gives Pukot a concrete buyer and work unit. The danger is becoming a low-margin labor-arbitrage service.
What buyers should pay for
Task design, routing, independent responses, quality control, disagreement retention, provenance and structured delivery.
What should improve over time
Task-specific reliability, quality per dollar, throughput, information gain and the ability to route harder work to proven contributors.
What should not be the moat
Cheap geography by itself. If the system cannot improve quality or routing, it is a services company with replaceable labor.
Seven buyer routes, each with a separate proof burden.
1. Managed AI evaluation
Project and program fees from AI/ML teams. Current strongest wedge.
2. Human-work marketplace
Margin or take rate on paid tasks routed to qualified contributors.
3. Sponsored career cohorts
Universities, workforce programs, communities and employers pay for navigation and outcome reporting.
4. Recruitment / placement
Employer subscription, placement or qualified-introduction economics if capability + direction improves conversion.
5. Outcome intelligence
De-identified cohort analytics around route attempts, interventions and outcomes.
6. Capability-routing API
Software for platforms that already have human work but need better assignment and reliability evidence.
7. Consumer decision/search
Paid high-intent periods around job search and consequential decisions; useful as validation even if B2B becomes the primary payer.
Useful only if every arrow can be measured.
Sulog finds a real transition problem
A person arrives because something is changing, not because they want another generic career profile.
Pukot builds capability evidence
Real or calibration tasks show where a contributor is useful.
Buyers fund human work
AI evaluation, research, employers or institutional programs create paid demand.
Outcomes update the system
Task resolution and career outcomes change later routing and route quality.
What would change our mind.
If those do not materialize, the current venture thesis should narrow or change rather than being protected by more product complexity.
Evidence used for the problem framing.
PIDS — education-job mismatch
Philippine labor-market overeducation and undereducation estimates.
Source →Research boundary
Labor-market statistics establish that transition problems exist. They do not establish that our product solves them, that customers will pay, or that the same route works unchanged across countries. Those claims require product and commercial evidence.