Betting on Service-as-Software. What I've Learned from My Recent Investments

A few months ago, Julien Bek at Sequoia wrote Services: The New Software, arguing that the next trillion-dollar company will be "a software company masquerading as a services firm." His logic: services are a six-times-larger market than software, and AI now lets a new breed of "autopilots" sell outcomes directly to customers instead of selling tools to professionals. As models improve, delivery gets faster and cheaper, and the moat gets deeper rather than eroding.

I agree with him, and not just in theory. Over the last 12-15 months Notion has backed several companies operating within this thesis.

Naboo helps corporates procure and manage their events. Rather than handing companies a piece of software and leaving them to figure it out, Naboo takes them through the entire process, from sourcing to execution, as a “done-for-you” service.

TERN is building AI-employees to scale home healthcare in Europe, starting in Germany. Tern uses AI products to fix the main structural problems that have historically capped this industry that include poor technology and workforce shortages.

And I recently invested in a company (soon to be announced) in the construction space that is helping developers check their building plans against regulation during the pre-construction face, and they are doing this as a service rather than selling the software to construction players.

All of these are examples of the Service-as-Software model where the technology is what makes the outcome possible, but what the customer is actually buying is the outcome itself rather than the software.

Why services have the edge over software right now. Outcome-based pricing has the potential to capture far more value per customer than seat-based software pricing ever could, because you're priced on the job done, not on how many people log in. Sales cycles are often shorter, since customers are buying an immediate solution to a problem that is extremely valuable to them. When companies are selling software, someone needs to approve a budget to implement and in certain cases also replace existing solutions. Buying a service is a softer decision, because there's no implementation, adoption or rip-and-replace to get wrong; if a software rollout fails, the person who championed it owns that failure personally, with everyone pointing back at them. That is not the case with a service as the buying process can happen gradually, often in parallel to the existing solution.

Gross margin in service-as-software businesses. The instinct is to compare service-as-software business's GM to a SaaS business's and conclude it looks worse. I don't think this comparison is fair. In early stages in which Notion Capital is investing, what matters is the GM trajectory rather than the starting GM. A technology-led service may start at a 35-40% margin that looks like a traditional agency, and still be a good business if the cost to deliver each unit of output is falling every quarter as AI absorbs more of the task. The businesses I'm most excited about are the ones where that curve is bending hard toward software-like economics within a few years of launch.

The metrics I actually track. Early on, I care less about ARR and more about revenue per individual delivering the service as well as time-to-deliver. These are the closest ways to measure how much more efficient businesses are using AI compared to traditional services in the same space. Utilisation is a good metric to determine whether we are scaling the team faster than demand, and metrics like number of automated workflows or tasks show us what share of a task is handled by AI versus a person, and whether that share is growing on a quarterly basis. This is in fact the best predictor of where GM is heading. And on the growth side, renewal and expansion of contracts matter as much as net revenue retention (NRR) does in SaaS. These metrics are also a good indicator of satisfaction or in other words, the NPS score.

What is still unclear to me. SaaS has standard, well-worn playbooks for areas such as go-to-market, pricing and customer success, among others. By now we have a good sense of what a healthy ACV looks like within a given customer segment, what a good NRR looks like for that segment, and how to structure a contract so delivery isn't delayed and revenue is paid annually in advance. None of that is settled yet for Service-as-Software. We know what a Series B or C investor is looking for in a SaaS company almost by heart at this point: the ACV range, the NRR band, the retention curve that tells you the cohort is healthy. I've just laid out the metrics I believe matter for Service-as-Software; however, I don't have any clarity about the values that Series B and C are expecting to see in these businesses to invest. For example, is a GM of 30% a good GM for a Service-as-Software business? We haven't worked with enough of these businesses for long enough to answer that, and we might never get a single clean answer at all if it turns out these businesses are far less similar to each other than SaaS companies were. Until those benchmarks are better understood, I don't think we will fully know what growth investors will expect to see in each of these metrics, or how they'll weigh them against each other when it comes time to put a value on the business.

I've always been obsessed with good products. Products that are easy to use and actually deliver the value they promised. What excites me most about this Service-as-Software category is that it closes that gap by design: the business only wins if the customer actually gets the outcome. As suggested above, sales cycles in these businesses are probably shorter, but customers will not renew with a provider unless the outcome is at the level of quality expected. Ultimately, services are easier to replace than software so product quality matters even more. The product on its own is what makes a great Service-as-Software company.