A note for private equity and venture people who think about education, fragmented services markets, and what AI actually does to both. Written by me, because I answer my own email.
Eugene Toh, Founder, ETG Economics and Genius Plus Academy
I have taught A Level Economics since 2007 and put nineteen consecutive cohorts through the national exams. I wrote the H1 and H2 TYS answer keys and 50 Model Essays, published through SAP and Shing Lee and sold at Popular bookstores. ETG ranks first on Google for its subject in Singapore, holds a 4.9 rating across more than 500 public reviews, and publishes its results openly: a long run distinction rate roughly double the national norm, self reported cohorts, caveats stated where anyone can check them.
With my wife Eileen, a former MOE teacher, I also run Genius Plus Academy: Primary and Secondary Math across four locations, built on more than a hundred proprietary learning materials she authored herself.
Two facts frame everything else on this page. The JC student population in Singapore is demographically shrinking, and our student numbers are growing. And in nineteen years we have never raised a cent.
Strong margins, prepaid revenue, multi-year customer relationships, and one fatal flaw: total dependence on an irreplaceable founder. The knowledge lives in one person's head. The trust lives in one person's name. Key person risk makes these businesses nearly impossible to invest in, integrate or scale, and every operator in this industry knows it.
Almost nobody fixes it, because fixing it means writing down twenty years of tacit knowledge and rebuilding your own company as a system while continuing to run it.
I did exactly that. And the reason why is personal enough to be worth a section of its own.
Here is how I operate. I treat every mistake as a review. I treat every adversity, every looming challenge, as an instruction to rebuild and do better. So around four years ago, when I looked hard at AI and concluded it was a threat to everything on this page, I did not write a defence memo. I started levelling up.
Since then I have automated almost every process this company runs on.
I built all of it myself: no agency, no dev team, no consultants. In my niche, I am operating at the frontier of what AI can do, and I say that carefully, because this page has already promised you can check.
I work harder than anyone I know, and with AI I now ship like a hundred person team. The site you are reading is several hundred hand coded pages, its content engine, its redirect architecture and its analytics, rebuilt by me directing AI agents, in weeks, not quarters.
And when I face a problem, I do not solve it once. I build a repeatable system so the problem stays solved after I have moved on. That habit, applied for years to every corner of two companies, is what produced the machine below.
Every process in the company, hiring, onboarding, daily operations, escalation, compensation, performance management, culture, is codified into machine readable modules. They are not shelfware. New administrative hires train against them, pass scenario based assessments at an 80 percent gate, and run a centre shift solo after eight shifts across two weeks. The same modules are the ground truth our AI agents query and execute against. That number is the kind of thing you can test on a site visit: pick a new hire, pick a shift, watch.
The usual objection to systematised tuition is that the pedagogy stays locked in the founder. Mine is in print. The TYS answer keys. Fifty model essays through SAP and Shing Lee. A decade of worked national exam answers, hundreds of open model essays and notes published on this site, structured for machines as deliberately as for students. At Genius Plus Academy, classes already run with tutors who are not the founders, hired, trained, gated and promoted through the same system. The method is not a mystique. It is an asset you can hold.
Built for us, by us: staff training with auto generated assessments and engagement gating, an institutional knowledge base the whole team queries, a hiring pipeline, student records. The integration infrastructure a consolidator normally builds after the deal, running before one.
Administrative operations across every location run on two full timers and a small bench of part timers, several of whom are with us for a season by design. Teaching is never staffed this way; the admin layer is. Staff leave, operations do not degrade, because the knowledge lives in the system, not in heads. AI fluency is a stated expectation in every contract and every review. We solved key person risk at the staffing layer first, on purpose, because that is where it is cheapest to prove.
I still teach every ETG cohort myself. By choice: the classroom is where the material gets sharper every year, and the results above are the return on that. What I have removed is everything around the teaching that used to require me, and, through the published corpus, most of what was only in my head about the teaching itself.
And the system has so far been proven inside two brands we built ourselves. It has not yet been installed into a business we did not build. I am not going to pretend otherwise on a public page.
But notice what that means. The expensive, slow, usually impossible part, turning a founder's twenty years into an executable system, is done and inspectable. Running it on an asset we did not build is the next experiment, and it is exactly the experiment worth running with the right partner.
Tuition in Singapore, and across Asia, is large, fragmented, cash generative and run almost entirely on founder hero mode. Consolidation in this industry has always foundered on the same rock: integration. The asset walks out the door with the founder.
A codified, AI operable system attacks that rock directly. Codify, systematise, then scale quality. Everything we build is held to an internal engineering constraint: it must work at five times our current student volume without compromising what makes the results real. That is a design spec we build against, not a forecast I am asking you to underwrite.
So here is what a partner actually gets, said once and plainly. In one person: the CEO, the CTO, the COO, the CMO and the chief AI officer, each at working depth, not slide depth. Plug me into a larger system and I will do to it what I did to my own: work through every element, every process, and rebuild each one repeatable, in a very short time. And the part that should interest you most, given everything this page has admitted: I train people, and agents, to replace me.
Making myself removable is not the risk with me. It is the product.
The businesses fund themselves and always have, so this is not a raise. What interests me is partners who back operators rather than decks: buy and build in this space with the system as the integration layer, strategic capital toward the five times constraint, or a structure neither of us has named yet.
Financials, cohort data and the system itself are shown live, in person, under NDA. This page is deliberately the surface.