Could AI Democratize Access to Education?

Joash Lee and Cherry Sung of Sedifly co-authored this article.

Photograph of a human-looking robot. Image courtesy of Taiki Ishikawa via Unsplash.

AI is advancing rapidly, and along with that, a swarm of “ChatGPT Wrappers” built atop OpenAI’s model and non-technical builders “vibe coding” have spawned. As EdTech builders, we’ve seen this trickle into the space. Yet amid the hype, AI has indeed transformed education — forcing schools to rethink the way they assess students while unlocking access to higher education, streamlining grading, and enabling personalized planning. This wave of change has affected the education consulting landscape too.

World-class education is seen as a “must-have” in Southeast Asia, and with a prevalent “kiasu” culture, parents are willing to fork up top dollar to secure their children's futures; rates at most educational consulting firms start from $10,000 and run into the millions for multi-year long packages beginning as early as age 12. For instance, The Koppelman Group offers a year of application support to 10 US colleges at US$40,000 as of 2024 (~S$52,000). It’s also not uncommon to see certain consultancies doubling their rates every year… talk about inflation!

Nonprofits such as Project Access and Questbridge serve lower-income families, though these often provide limited support and lack scalability as they rely on volunteers and donations. Yet, traditional consulting relies on humans, or as they’re more often called, strategists, which is labor-intensive and only accessible to the most privileged students.

Will access to educational consulting remain reserved for the elite few? Or could AI, as we’ve seen in other sectors, democratize access?

Is the future already here?

Thus far, we’ve seen the emergence of some AI educational consulting upstarts attempting to break the “pay-to-win” barrier where privileged students gain an unfair edge over the rest, and the space has been bubbling with acquisitions. Most of these players aim to lower the cost of guidance and feedback that historically required hours of one-to-one time with strategists. 

One such upstart, AdmitYogi, delivers real application examples and personalized essay insights with rates beginning at $100. Within several months of launch, the firm turned profitable and was acquired by Crimson Education. Despite initial momentum, one of its founders, Ananth Veluvali, publicly alluded to unsustainable unit economics due to the pricing model of such a tool which hinders long-term revenue growth, and the acquisition was for an allegedly small sum. Another AI educational consultant, Sups AI, was acquired by the U.S. News to bolster its CollegeAdvisor business.

From observation, we haven’t noticed significant progress nor developments in these upstarts since their acquisitions, suggesting these deals might have been pure “data plays” with the aim of tapping on the upstarts’ data sets to integrate with in-house offerings, rather than to run as standalone solutions. This checks out since the acquirers are the largest global educational consultancy and a major media company with over 42 million monthly visitors, respectively, so much value is unlocked by combining AI and human guidance.

We also spoke to several EdTech founders, including college admissions influencer Daniel Lim who previously built an AI educational consultant he says didn’t work out, and consensus is it’s not worth building such tools. Even with Daniel’s 500,000 followers and half a billion views which should’ve been a launchpad, his AI educational consulting startup eventually failed so he did what any good founder would when things go south: cut the loss and pivot to building something else.

The good, the bad, and the ugly

But why exactly is building a standalone AI educational consultant — or as we put it, a tool that serves the “sandwich class” by providing 80% of the service a traditional consultant would at 5% of the cost — so tough?

1. Unsustainable unit economics

It's tough to grow topline substantially when operating in a high churn industry with low average revenue per user (ARPU), and consequently, low lifetime value (LTV). This is because once students get into college, they never need to use the tool again, and charging anything more than $10 to $20 a month under a SaaS (or more accurately, AI-as-a-service) model seems like a rip off; AI is commodity as a result of the insane amount of VC funding poured into the space, and consumers aren’t willing to pay premiums for such tools with the plethora of AI models available at low-to-no cost. So SaaS models in this sector hardly, if ever, work, unlike for services that lock users in sticky ecosystems, such as Dropbox.

2. Total addressable market (TAM) is minute

The TAM for AI educational consultants is currently small as a result of the low ARPU and LTV. Though the number of students applying to colleges annually is large and the broader educational consulting market is growing fast, most revenue is generated at the top. This makes it tough to build a venture-backable AI educational consultant.

3. Challenges in scaling past individual consumers

Many startups have adopted a consumer-facing go-to-market strategy using social media platforms like TikTok, but this funnel is hard to nail and building a winner will require unrealistic levels of adoption given the unit economics. Compounded with the inevitable hurdles that come along with an AI educational consultant such as students dropping out post-application or admission, upstarts building in the space have struggled to find true product-market fit.

So is it all doom and gloom for the “sandwich class,” and will they lack access to quality guidance forever, or is there a way out?

Some key questions emerge as we think about the feasibility of building an AI educational consultant:

1. Would consumers be willing to pay a premium for an AI educational consultant that can do the job better than ChatGPT? In other words, could we increase ARPU over the long run?

2. Could we scale beyond the “essay tool + cheap subscription” model? In other words, increase LTV and grow the TAM by integrating more deeply into a student’s lifecycle (from school-fit → application → decisions → enrollment → post-enrollment retention), rather than a single point-service?

3. What would be required to successfully bring the solution to market? In other words, what’s the winning distribution strategy?

Working with and for several educational consultancies and speaking to thousands of students gave us answers:

1. Breaking the value barrier (pt. 1): Our thesis is that the cost of generic AI models like ChatGPT will exponentially decrease ultimately to zero over the coming years or even months. However, over the next 2 to 3 years, tools serving niche markets, such as an AI educational consultant, will come at premiums, and consumers will be willing to foot the bill. It’s not unlike how Google is free today, and generic data is available at no-cost on Google, but firms still pay top dollar for market intelligence tools or data providers like Bloomberg Terminal, Crunchbase, and Statista. In fact, we’ve already seen this happening, with tools like Rogo.ai serving investment banks. What differentiates the next generation of application-layer tools from a “ChatGPT” wrapper will be the upstart’s ability to build moats around data, distribution, and user interfaces.

2. Breaking the value barrier (pt. 2): Beyond an AI educational consulting tool that helps students get into their dream school through essay insights, extracurricular recommendations, sample profiles, and acceptance rate calculators, much value can be accrued through the addition of supporting services that extend past “termination day.” For instance, an AI tool helping students navigate college life, clubs, and careers might prove useful and be something consumers would be willing to pay for, past merely getting into college.

3. Breaking the awareness barrier: A B2C approach works fine and well for traditional consultants, given the high margins and exorbitant fees they charge, resulting in relatively fewer students they serve to make the business case. Pareto’s principle applies to the revenues of many educational consultancies — just 20% of clients make up 80% of a firm’s revenue. However, AI educational consultants must explore other means of penetration such as B2B (institutional partnerships) or B2B2C to attain critical mass quicker as a sole B2C approach won’t work given the unit economics.

No doubt these are formidable hurdles builders must overcome, but we believe in the long-term potential of AI educational consultants that can enhance, not replace student guidance as countless students would benefit from a “just enough” offering. In an informal survey we conducted involving more than 100 students who applied to a US or UK college without guidance from an educational consultant, over 95% mentioned they’d have paid for a lower cost solution if it were available to them. That said, we don’t think AI educational consultants will ever replace traditional consultants as the top end of the market will always exist; consulting is a business that is deeply human — that emotional connection is irreplaceable.

Building winners

Wrapping up, we’ll share some final thoughts on opportunities we see, metrics we’re tracking, and how to build a winner.

Funding and valuation of largest AI startups. Image courtesy of The Economist.

Today, the market is saturated with horizontal AI applications, with giants such as OpenAI, Anthropic, and xAI. Comparatively, the market for vertical AI applications is relatively nascent. Though certain sectors like LegalTech have rolled out vertical applications faster than others, there’s enormous potential in specialized solutions with integrated workflows, created and tuned by domain experts, and we’ll see the proliferation of vertical AI applications in sectors like EdTech soon.

Given the ever-changing landscape, we’re closely monitoring developments in AI, especially on when sector-specific applications coming at justified premiums become mainstream. In addition, we’re tracking our industry partners’ perceptions on and reception of AI educational consultants — we’ve validated the problem as feedback we’ve received is that Educational and Career Guidance (ECG) counsellors lack full context on global college admissions, but institutional AI adoption for most tools, even beyond EdTech, remains slow, with annual recurring revenue (ARR) presently driven by consumers.

How we think the educational consulting industry will pan out. Image courtesy of Sedifly.

As the educational consulting market stands, the number of students applying to colleges, as well as the proportion of those students working with educational consulting firms, is increasing. The pyramid on the left shows the state of the educational consulting industry today, and the larger pyramid on the right depicts growth of the sector and the layers we think will emerge. As we descend the levels, the rate an individual student pays decreases. Today, educational consulting is accessible only by the top 1% of income earners, but solutions serving the “sandwich class” are changing that — these upstarts are a form of disruptive innovation, and that’s where we think the greatest opportunity to build lies. Though the rate paid by an individual student in the “sandwich class” is lower than that of a student working with a traditional educational consulting firm, there are many more students in this middle segment, potentially resulting in a larger TAM over time.

Sedifly team in Singapore. Image courtesy of Sedifly.

The perennial question remains: How do founders build winners? Many would gravitate towards building “sexy” upstarts in Web3 or AI without actually solving real problems, before winding up in mere months. Instead, we’re first tackling the low hanging fruit — traditional consulting — with a profit with purpose model that serves the historically neglected middle segment. We call this “200% of the service at 20% of the cost,” and though not at the “5%” price level of an AI educational consultant, there’s a huge deadweight lost to be reclaimed. “Sexy?” Not so. But focusing on the “problem to be solved” has paid off — we closed a 3x oversubscribed pre-seed round not long ago and crossed six-figures in revenue shortly after. The takeaway? Solve problems; don’t create them. Because being a founder isn’t about doing what generates the most clout (ahem… AI B2B SaaS); it’s about adding the most value for your customers.

As to whether AI could democratize access to education, we believe humans are here to stay. That said, we’re bound to see an AI educational consultant take the crown soon.

Full disclosure: One of the authors, Joash Lee, has a personal stake in Dropbox and xAI, companies mentioned in this story.

Note: Variations of this story were published on Tech in Asia and Forbes.