I’m hanging my hat on services-as-software
Transitioning from traditional software into the AI world
When I started Aligned, I had been investing in B2B software (which, for my entire career until that point was effectively synonymous with SaaS) for seven years and building it at a startup before that. I conceived of the idea of Aligned in 2023, began to put the pieces in place in 2024, and went live in 2025. These were pivotal years in the transition to the AI era.
I’m not going to lie – I was intimidated at first. It was hard to hold a crystal ball and predict what would happen to software business models. To what layer of the stack would value accumulate? Had I already missed the boat on the foundational, or at least generational, opportunities? What would create defensibility for startups in this new era, if anything? What would companies choose to build in-house rather than buy as software engineering not only democratized but also became exceptionally fast, and therefore lower cost, to do?
One thing felt certain to me: people were only going to use more, rather than less, software at work. How their jobs transformed was hard to predict – frankly, for everyone – but there was no way work was going to become more manual or stay the same given the new possibilities unlocked by AI (and the general progressive direction of society at large), and so I felt comfortable starting a B2B software focused fund and figuring out where value would accumulate as I went. I’d take my time deploying (which, luckily, is baked into our low velocity model) and keep my finger on the pulse of the cutting edge by spending time with the smartest minds paving the future of work. Either software was going to be sold to solve problems directly, or it’d be sold to help people solve their own problems, so either way, I was in the clear.
Discovering the services-as-software paradigm
Before I began deploying, I spent nine months with founders, operators, and software buyers understanding where they believed the world was going and what was worth building in the new AI era. It occurred to me that software was no longer a hammer for purchase to help an end user solve their own problems. It for the first time had the ability to solve the problem itself, acting much more like a service provider than a tool. This created a massive opportunity to capture much more value from customers, ultimately “creating” larger market opportunities, because the value software could deliver was much broader and more tangible. It also called into question for me whether “software” was really the right term at all…sure, it was the means, but wasn’t the thing being sold just a service?
Eliana Berger first described what she was building to me as services-as-software in early 2024. I had never heard the term, but it made sense to me – software being leveraged under the hood to deliver a service.
It resonated because of the value capture opportunity. Sure, you could sell a tool that helps an existing service provider (in her end market’s case, a healthcare biller) get a job done faster or more efficiently, but wouldn’t you rather just be the AI biller?
I started noticing that I applied this logic to every company I saw that was fitting their product into the old paradigm of software being an efficiency tool. Why would you sell a tool when, for a high margin, you could sell the solution to a customer’s problem – which surely delivered more ROI?
This logic required a leap of faith – the belief that AI truly could deliver the solution without requiring excessive human intervention, ultimately making it low-margin and difficult to scale (at least for venture-backed companies). I had lived through the last generation of tech-enabled services that many believed would get to high margins…and yet, most if not all never did.
But AI was only getting smarter, and especially when trained to target a specific domain and set of use cases, it was only getting less error-prone. As long as there was a human in the loop to check its logic and outputs, it was already proving to be able to solve end-to-end problems largely on its own.
What this evolution means for business models
So are all software companies in the AI era going to be tech-enabled service companies? Just like before the paradigm shift, my belief has always been that companies should sell the way their customers want to buy. There are some problems that teams want to outsource – for example, in most cases, drug development services (like our portfolio company Dash Bio offers) for a number of reasons including the fact that they require infrastructure investments and regulatory compliance which benefit from economies of scale. Liability, core competency/fluency, network effects (i.e. benefit from being on the same “platform” as others), and limited in-house resources are all reasons why companies choose to outsource problems rather than manage them in-house.
But suppose there are things people want to own in-house – customer communications, for example. These domain areas are typically differentiating for the business (vs a necessary evil) and viewed as a core competency. In some cases, businesses will choose to build tooling to support these use cases in-house. But in other cases, where the business is not technology native (think businesses in the trades or healthcare providers, for example) or the efficacy of the solution benefits from usage by other businesses, it may desire a third-party solution provider’s help. This is where there’s opportunity, in my view, for services-as-software with the customer’s human in the loop (rather than the company’s human in the loop). It’s like a tech-enabled service, except the solution isn’t fully outsourced… the customer is, at the end of the day, in the driver’s seat. This is the other, in my view, primary business model for services-as-software companies.
Defensibility when development costs are plummeting
A common argument against investing in (or building) these types of businesses is the idea that there’s no defensibility – why won’t incumbents who already have distribution in a given vertical or department just usurp these startups given the nearing-zero cost of developing new capabilities? And what keeps another startup from popping up and taking market share?
Let’s address incumbents first. In my view, this depends on the market. Some incumbents are incredibly formidable – in fact, I can’t believe they’re incumbents at all…it feels like just yesterday, they were early-stage startups. Yet they have huge market share and trust in their industries, and they’ve managed to remain innovative. To me, this makes it much less interesting to compete with foundation model providers on core AI infrastructure, for example. Others have found themselves in their own competitive bloodbaths, with their core businesses being threatened by their largest incumbent competitors. These often serve as distractions from capturing net new opportunities… especially if these companies are not the most fast-moving and innovative at scale. Another consideration is channel conflict – some incumbents sold to service providers in their respective markets, meaning disrupting them would cannibalize their core businesses, which they can’t afford to do. Same with business model conflict, where incumbents may charge by seat in a world where the humans they used to sell to are being replaced by AI-enabled services-as-software. This similarly makes it difficult for those incumbents to react and compete in a new paradigm without putting their core businesses at risk.
In the case of new startups, I’ve always been a believer that to build a large enduring business in software, you need to have a defensible competitive advantage beyond “it’s hard to build.” It has only been getting cheaper and easier to build new software products over time (for example, we built so many things from scratch in my days as a PM that only 5-10 years later were all offered via third-party SDKs). These modes of defensibility fit into a few buckets:
A first-mover advantage paired with stickiness; examples of the latter include…
The solution becomes step function more valuable once it’s trained and invested in like a set of employees would be
It stores data and/or context that’s not transferable
It has been approved by third-party regulators or procurement processes with tons of red tape
A distribution advantage that allows unusually fast and/or inexpensive market capture paired with stickiness or a high barrier to entry; examples of the former include…
A partnership that scales distribution quickly & efficiently
Virality
Existing trust in a hard-to-penetrate industry
Note: In tight-knit vertical markets, this can be particularly effective as word-of-mouth virality is often inherent to how the ecosystem adopts new technology
A first-mover advantage paired with network effects; examples of the latter include…
A benefit of different market participants living under one umbrella
Economies of scale
Without these structural advantages, I agree with the concern – there’s less and less reason for companies to have staying power as the cost to replicate plummets. This is an incredibly important criterion in my personal investment evaluation process.
An exciting time to invest
We’re early in the transition, but one thing became clear to me in my first year running Aligned: the next huge opportunity in B2B software will come from services-as-software business models, regardless of whether the company’s or customer’s human is in the loop. If a buyer has the option to pick something more powerful with greater ROI, it’s only a matter of time until they’ll do so, and those following the old “hammer” paradigm will be left behind.
I’m thrilled to be investing out of The Aligned Fund I as the opportunity for step function change emerges in every vertical and department, transforming the way companies run. We’ll be backing the most promising services-as-software companies coming out of this era as they begin to exhibit early signs of product-market fit.
