The Future of Embedded Technologies

it's embedded all the way down

The basic point of this piece is to argue that embedded technologies will become exponentially more important over the next 10 years of software.

Capital efficiency is the name of the game. Abstracting away complexity behind well formulated APIs will continue to increase this dynamic especially as AI continues to be served to end users via API as well.

For this reason, it’s still incredibly early to be investing in, founding, and building with embedded infrastructure.

A Prerequisite:

If you had all the time and money in the world, you’d probably build most things yourself.. Call this the Uber way… or even the Jeff Lawson approach. Faced with building or buying, both Twilio and Uber defaulted to building software in-house. We don’t really live in that world anymore.

It’s even questionable whether Twilio or Uber were advantaged by building a ton of their software stack internally. I’ve seen no formal calculus done on this, but my gut instinct says that neither Uber or Twilio’s terminal value was heightened by their decision to prioritize building internal software where they could have bought it.

In fact, you could make a damn good argument that not only did it have no impact on the terminal value, it impacted the capital efficiency as measured by ROCE.

Build or buy is of course an important decision for all software purchases, but embedded is increasingly becoming the most important sub-category for vertical SaaS.

Embedded, Revenue, and Churn

The “embedded” vernacular is typically applied to fintech initiatives - think your payments, capital programs, banking, and more.

These are all great products that can dramatically impact revenue per user on platform. However, this is becoming a myopic view that is missing out on the ways embedded infrastructure is beginning to factor into other aspects of the customer experience on platform.

Embedded initiatives drive their importance not only from the revenue they create, but additionally from the ecosystem they uphold.

Multi-product users are churn-resistant users. And churn, especially in the mid-market and small business context, is typically the single-biggest determinant of a vertical SaaS company’s staying power. If you’ve got a great sales process, but you bleed 20% of new customer acquisitions after 12 months, you’ve got problems.

To put it in more technical terms, embedded initiatives improve ROCE by reducing the capex required to increase RPU, LTV, and NDR.

If layer cakes eat churn for breakfast, embedded initiatives get companies there faster. Put bluntly, by increasing the LTV of a SaaS company’s customers with minimal R&D costs, teams get more time and resources to add wholly-owned products with high margins.

And because of this LTV increase through churn reduction, today, embedded companies underprice the business impact they have for software companies.

Lastly, embedded initiatives are in turn creating new types of ecosystems ensuring that the platform itself is the conduit of activity for the value chain they serve.

Embedded to Reduce Friction (NDR)

The first 30 days for a user are where they have the highest propensity to churn.

As such, the highest returns on embedded initiatives stem from embedded attempts to reduce the pain of user onboarding.

Unified APIs tend to fit squarely in this bucket.

Plaid and Unified APIs

Plaid in particular nailed this. Nobody wants to go find their bank account and routing number to receive payments or some service from a platform. And SaaS platforms don’t have the time to build this out and accommodate every single bank in North America. As a result, Plaid is directly responsible for some ridiculous number of bank linkings. The friction reduction they have facilitated is absolutely massive with an almost unquantifiable impact on churn.

More unified APIs will follow suit. Most SaaS platforms could benefit from easily importing data into their application with ease. And we are quickly approaching the point where it will be easier to go to Merge API, Finch, or another unified API player whom has expertise in the auth flow, in the customer experience, and in the integration maintenance, than to build your own integrations.

In the same way that Plaid elevated bank integrations , the experience for the end business will be elevated, and churn in the onboarding flow will be heavily reduced.

That is the sign of a fantastic embedded opportunity.

Embedded as a Revenue Opportunity (ARPU/LTV)

Embedded payments and other fintech initiatives promise directly quantifiable increases in revenue per user. Payments of course is the big driver here and will always be the most steady-state way to predict revenue per user. But many others are now creating embedded opportunities for companies to create a financial ecosystem.

Embedded payments for the end business come with two really great effects.

1. It effectively subsidizes the cost of business software

2. It makes it far easier to run and manage a business from one operating system

These alone are substantial impacts.

Prior to Stripe and others, payments were mostly sold with a referral model. Software vendors (ISVs) would “refer” a customer to payments processor in exchange for some nominal bps on the payments.

This was an unholy alliance for two reasons:

1. SaaS companies didn’t have actionable insight into the sales side of their customer’s business.

2. It was too much of a sweetheart deal for processors whom got deal flow at very low costs.

APIs and new payment facilitators models effectively changed this to give us the new embedded payments world we have today. Embedded payments players navigate the network of acquirers, issuers, and more while SaaS platforms are able to monetize more effectively through increased take rate on embedded payments while acting as distribution vehicles for payments players.

The embedded fintech movement hasn’t stopped there and we are now arriving in a new era with capital, bank, and issuing.

In 2022, 44% of Square gross profit came from sellers that used four or more monetized products, an improvement of more than 15 points from three years ago. Square was among the first to realize that smaller sellers want seamless cards, banking, and other financial products in one partner.

These incremental financial products add up to drastically increase RPU.

Embedded as an R&D Driver

Perhaps the most important embedded experience now occurring in software centers around the use of foundational models.

There’s an argument to be made that partnering with OpenAI, Anthropic, or another player to launch an AI copilot shouldn’t be classified as an embedded technology.

Sure, except the end product that comes from these partnerships is an AI that an end user is directly interacting with.

The benefits of these partnerships are crystal clear. Most SaaS companies are not going to burn the tens of millions of dollars in GPU spend to monetize AI effectively. Nor staff a team of AI engineers to do so

Instead, the technologies undergirding AI products are more akin to embeddable technologies, only somewhat contextualized by a SaaS partner.

The decrease in capital expenditure to embed these models is perhaps the most exciting trend in software right now. They can create new revenue lines that customers are actually excited about with minimal R&D costs.

Perhaps more importantly, by giving end users a new way to interact with your software product in hopefully increasingly frictionless ways, they stand to create more expanded customer adoption of the SaaS solution. With that increase in LTV, SaaS companies can redeploy capital into other product offerings at a faster clip.

That in turn is creating a virtuous cycle, where customer adoption stands to go up, faster and faster.

I’m most impressed these days by Filevine, whom is cramming useful AI models into every aspect of their platform: lead generation, document management and creation, case management, and more.

I’m assuming their calculus is something like this: if we can make product adoption of a certain module go up by 10% by deploying an AI solution within the module, we can reinvest dramatically into our platform and create more modules with heightened adoption.

AI played correctly is an accelerant to ecosystems. If you can directly influence customer adoption through AI copilots, it stands to create massive benefit to accelerating product roadmaps, taking more bets, and growing an ecosystem.

Where do we go from here?

It seems to me that we are still in the early innings around pushing the limits of embedded technologies.

After core product development is done, there are all these complex products that can increase the utilization of the SaaS solution on the whole.

And while contextually, it may look different from platform to platform, it’s becoming more capitally efficient to go multi-product faster and faster, accelerate product utilization with AI, and continue to parlay these gains into revenue and retention.

Accounting and other financial products continue to be low-hanging fruit. But what about other products like business valuation for business acquisition? What’s to stop Baton from developing an embeddable business valuation tool?

Or when do AI products strike direct partnerships with SaaS platforms? Wouldn’t it be nice to have a BPO service abstracted via an AI interface?

There seem to be plenty of areas where this stuff occurs. Platforms that are the source of truth for their industry will

Likewise, many aspects of business management will continue to get simplified for end users whom will also experience more capitally efficient businesses.