Why I Killed CollegeCraft.AI

Part of what I'm doing with Future Obvious — my startup studio — is trying things. After building and selling InductiveHealth, I entered a phase where I could finally pursue the ideas I'd never allowed myself to put time into while scaling a company. That's a gift. It's also a setup for some hard lessons.

Not every venture works. Some fail because the product is wrong. Some fail because the market isn't ready. And some could work — the product is solid, the mission is real — but they don't map to your particular superpowers, and doing ambitious things without that alignment is a slow bleed of time, money, and focus.

CollegeCraft.AI was that third kind. I shut it down last November. Here's why, and what it taught me.

This One Was Personal

I went to an elite private high school in New Orleans. I had a college counselor. On paper, I was well-served. In reality, I was consistently advised to aim down.

I ended up at Emory — which turned out to be exactly where I was supposed to be. But I didn't find out until I was already there that I would have comfortably qualified for the Woodruff Scholarship, one of Emory's most prestigious merit-based awards. The catch was that my high school had to nominate me, and they only got two nominations per class. Those went to kids whose parents were better connected in this upper-crust New Orleans community. Neither of them even attended Emory.

That was a huge cost impact to my parents - because of politics, not merit.

I got into every school I applied to. I received scholarship offers from safeties and stretches alike. The system worked for me in every way except the one that was supposed to be about recognizing the best candidates. That failure wasn't about me — I landed fine. It was about what it revealed: the process rewards connections and insider secrets over capability, and the people guiding students through it have incentives that don't always align with the student's best interests.

Now multiply that by every kid who doesn't have well-off parents. Who didn't go to an excellent private school. Who doesn't have a safety net when the counselor gets it wrong. Situations like mine and many even worse would change life outcomes for those students. And I know it happens every day.

I wanted to provide a floor — a level of quality, trustworthy guidance that every student could access regardless of their zip code or their parents' network.

What We Built (And It Worked)

Let me be clear: CollegeCraft was not a failed product. This was a full-featured AI college counselor — not a prompt and a website.

We modeled it on how a real counselor walks students through the process at specific times. Contextually relevant conversations about planning impactful summer experiences in January, when there's still time to act. Financial aid planning. Cost estimation. Application strategy. Everything a human counselor would cover, but with superhuman breadth and availability.

High school students actually used it without me having to pay them. That's a bar most edtech products never clear.

One student had the system prepare her for a selective summer program interview. She passed with flying colors. The college search feature drew real praise — as it understood a student's preferences over time, it suggested schools they hadn't considered. "If you like College A, you might like Colleges B, C, and D for these reasons." Students said it was one of the most useful features they'd ever encountered in any tool.

We put the student at the center. No selling student data to third parties. No manipulating their journey to benefit advertisers. In a space where many services quietly make the student's data the product, we believed trustworthy advice was the foundation. This is the most important juncture in these kids' lives so far, and accuracy and integrity had to come first.

We actually built the system twice. The first version was a tool for practicing counselors — what my wife wanted in her own practice. We quickly discovered that most counselors don't do nearly as much as she was doing with her students. So we rebuilt it as the product I'd originally envisioned: a direct-to-student AI counselor. The models met us there, delivering excellent results. We spent hundreds of thousands of dollars. Three full-time engineers, a full-time college counseling expert, plus my wife and me working uncompensated.

It worked. So why did I walk away?

Three Walls

The Channel Problem

The audience for CollegeCraft is high school students and their parents. The channel to reach high school students in 2025 is TikTok.

I am not a TikTok person and have real issues with it as a platform, but it's where the students are.

We hired a video producer to interview students. We tried to make it work. But founder-market fit is real, and the honest truth is that I do not have the instinct, the energy, or the shamelessness required to win on that platform. This isn't false modesty — it's a founder recognizing when the go-to-market demands a skill set that isn't his.

We also ran Google Ads with a free essay review as a lead magnet. We generated hundreds of sign-ups. But the quality was terrible — these were students who were already behind, looking for last-minute shortcuts. Many tried to get the system to write their essays outright. It wouldn't. That was by design, and I'm proud of it. But it meant our best lead gen channel attracted exactly the wrong users.

The Incentive Problem

This is the lesson I think about most.

We talked to several schools. Administrators would send us to the college counselor to evaluate the technology — which was appropriate. These are the domain experts.

At one Atlanta private school, after an in-person demo, the counselor told us: "This is amazing. It does everything I do. And that's why we're not moving forward."

Read that again. The product was too good. It threatened the gatekeeper's role, so the gatekeeper killed it.

I had less of an issue with this at private schools — those counselors are often genuinely excellent and have small enough caseloads to serve their students well. It is a key part of what parents pay for in these institutions. But the public school counselors, the ones with 600-student caseloads who do very little with the vast majority of those students? The ones who could have used this tool to transform outcomes for hundreds of kids? They had the same reaction. And that felt closer to malfeasance than professional judgment.

We were selling to people who felt — whether consciously or not — that they might be replaced by this technology. That feeling wasn't accurate, but feelings rarely are. And when the evaluator is also the person most threatened by the product, you have a structural problem that no amount of feature development can solve.

The Herd Problem

Even when we got past the counselor, we hit the parents. And parents making the single biggest financial decision of their child's life to that point... do what other parents do. They ask their counselor. They ask their friends. They follow the herd.

This isn't irrational — it's actually a reasonable heuristic when you're overwhelmed and the stakes are high. But it means adoption of a new approach requires either a critical mass of social proof (which takes years) or a channel through the existing trusted advisors (the counselors who felt threatened).

Students, meanwhile, split into two camps: the ones who genuinely engaged with the tool and got tremendous value from it, and the ones who just wanted the system to do their work for them. The ones who wanted to cheat found that cheating was easier with other tools that had no guardrails — which meant our integrity was, perversely, a competitive disadvantage with a meaningful segment of our users.

The Discipline of No

I made the call to move on last November. I kept the site up for a while, and entertained a few meetings with potential partners who might carry it forward. But even that residual time wasn't free — every hour spent on CollegeCraft was an hour not spent on Force Multiplier, and Force Multiplier demands irrational focus.

That kind of focus is something I know well. It's one of the primary factors that made InductiveHealth successful. When you're bringing something genuinely new into the world, you can't split your attention and expect to win. You can do several things, but not all at once and well. Creation is hard. It requires everything you've got.

Does it sting to have a failure? Of course it does. Would I prefer CollegeCraft to have worked? Absolutely. But part of this post-exit phase is trying the things I never allowed myself to pursue while building and scaling Inductive — and part of that is accepting that not all of them will pan out. The return on focus is still the most reliable force in building a company. I know this because I've lived it. Recognizing when to apply that focus somewhere else, even when the mission still pulls at you, is a kind of maturity I'm still earning.

What I'd Tell Another Founder

Direct-to-consumer tech is not my superpower. Not yet, maybe not ever. My strengths are in B2B, in complex systems, in selling to business owners who are trying to figure out how to do more with less. CollegeCraft needed someone who lives on TikTok, who has deep relationships in the education world, who can navigate the politics of school systems and the herd dynamics of anxious parents. That person exists. It just isn't me.

The broader lesson is one I keep relearning: ambitious things are hard in direct proportion to how far they sit from your natural advantages. The mission can be right. The product can be good. But if the go-to-market requires you to become a fundamentally different founder, you have to be honest about that gap.

I'm taking that honesty into Force Multiplier AI, where the founder-market fit is as strong as anything I've ever built. That's where my focus belongs.

If You Want It

The system is built. It's 100% functional. I turned it off to stop the infrastructure burn, but it could be turned back on tomorrow.

If you're in my network and you can see yourself making something of this — if you have the founder-market fit I don't have, the connection to this mission, and the willingness to eat glass for the years it'll take to crack the go-to-market — reach out. I would do a great deal with the right person to move this forward. But I need someone who can run with it. No additional cash burn from me on this one.

The mission deserves it. The students deserve it. It's just not my path right now.

— Matt