Subscription software has been on a steady trajectory to overtake the entire software market. I don't have data handy, but it's easy to run through landmark software products, from Microsoft Office to Apple iTunes/Music to Adobe Photoshop, and map out the transition. Once upon a time, we purchased these products once, up front, and we'd get a CD-ROM or MP3 file in return; today they are all subscriptions. Video games have resisted to some extent (Apple Arcade and Google Stadia, R.I.P., are not world beaters). But when is the last time you purchased a disk for anything else? The subscription model is so common that one might say it has reached its middle age — and we can start to think about what comes next.
AI is the latest craze, so perhaps we can learn something about future business models from AI startups. However, if we look around, AI startups are consistently following the established model: charge $10-100 per seat per month for a suite of AI tools, accessed by web & mobile apps — like a SaaS in all ways but the underlying AI technology layer. I imagine that smart minds at these companies see subscriptions as a comfortable transition strategy to something new. And, if AI reaches even a fraction of its enormous promise, that something new could be much much, more expensive.
The whole premise of AI, and the scaling laws that drive it (see the Bitter Lesson, Moore's Law) is that AI will soon achieve see near-human-level intelligence, and it will be available to anyone via simple APIs.
The first AI product that can truly offer human-level capabilities as a drop-in replacement for a remote worker should (in theory) be priced close to the full salary of that worker. After all, the AI offers the same value as an employee, with lots of added benefits like on-demand scaling. There's a common argument that AI will dramatically reduce the cost of labor, and that's possible in the long term, but one could also argue that the incredible value created by AI should be priced according to what it brings to a customer. And the value of an assistant as smart and capable as a human would be a lot higher than an Office 365 subscription.
And we're not talking about just one employee: This technology is expected to replace many workers — so a company with 100 employees on payroll could choose to hire, say, 10 AI employees to complement their workforce. That's an AI bill on the order of 10 salaries, or about a million dollars a year, which is 10-100x the price of purchasing a subscription software product for the 100 employees on staff.
That's a giant recalculation, both for the AI provider and the AI customer. The dynamics are going to be a lot different under this new model. I can imagine a few important changes:
First, as an AI customer grows its workforce, that company must decide how much capital to devote to AI vs. human talent. These resource allocation decisions will mean that AI revenue is going to be more volatile than subscription revenue. AI companies will have to constantly pitch and repackage services, the way that IT services companies do today. To combat this, they will likely repackage some simple capabilities into cheaper subscriptions, but there will potential for greater rewards in selling AI talent.
Second, AI talent is going to mean different things to different customers at different times. Consumers might be willing to pay $1000 for an AI tax professional to review 100 documents, while a business is willing to pay $10,000 — but the cost of delivering that intelligence would be largely the same, assuming cost scales approximately with input tokens. Subscription services are based on low-marginal-cost technology, but AI adds a layer of significant marginal cost, and the scaling inputs are going to vary from application to application. A lot of thought and effort is going to be spent translating AI cost structures into reliable pricing models.
Third, bottlenecks like GPUs and energy are going to play a big role in cost and pricing analysis. Demand-based pricing is going to be a bigger topic than ever. AI companies already work through these issues today, but, in a world where customers are paying for AI employees, the AI companies will have to find smart ways to pass costs and availability concerns on to their customers.
Fourth, AI customers are going to need to adapt to the idea of spending relatively huge amounts of money on software. CIOs and CFOs are going to have to adapt to this new world. And the AI companies that can help them adapt are going to have a huge advantage. Packaging AI talent in a way that satisfies the constraints of enterprise customers will be hard. Metrics (like ROI) and promises (like SLAs) will mean something totally new. AI companies are going to have to innovate in this space. It's going to be a lot less exciting than training the next great LLM, but the financial rewards will be gigantic.
Fifth, and most obviously, a ton of product innovation is going to be required for AI employees to seamlessly integrate with existing companies. Customers will need customizable security behavior (can AIs get spearphished?), audio and video avatars, API integrations, oversight dashboards, scaling tools, and much more.
All of this presumes that we do not reach AGI or some kind of singularity first. My opinion — which I am not going to try to defend here — is that AI is going to be smart, but not scary-smart, for quite a while. What I hope to convey is this: Assuming AI tools get to be pretty darn capable before reaching AGI, a few AI companies are going to be fantastically successful, in part because they learn to price AI talent efficiently — and those companies may not have been founded yet.