The Rise of Software-Delivered Services
Written by
Sean Linehan.
Published on Oct 15, 2024.
There is a big shift underway in the software industry. Fast, cheap,
and powerful multi-modal models have massively expanded the scope of
what's possible with software.
A new breed of AI-powered companies are jumping into the fray to put
these capabilities to work. These companies look quite different
from the previous generation of SaaS. They sell work, not workflow.
They sell outcomes, not tools. It's a whole new type of company, and
we need a new name for it.
A recent (and excellent)
Sequoia article
promoted the term "Service-as-a-Software". The label is funny, but
doesn't hit the mark. It's grammatically awkward, shares an acronym
with the companies it's trying to differentiate from, and most
importantly, it fails to be meaningfully descriptive.
I propose a different name:
Software-Delivered Services (SDS).
Below I sketch out what I think makes Software-Delivered Services
special.
Becoming Accountable
SaaS companies have always had to meet certain standards. For
example, most service level agreements (SLAs) include a provision
for uptime. But delivering actual work massively ups the ante. SDS
companies can't just keep the lights on and call it a day. They have
to consistently deliver high-quality results.
If you use
Outreach
to send a bunch of mediocre emails, well, that's your fault. If
11x
sends a bunch of weak emails, that's
their fault. Previous generation software companies didn't
have to grapple with this level of accountability.
Breakout companies will have to maniacally focus on delivering
quality. To achieve this goal, the org chart of Software-Delivered
Services companies will likely look pretty different from their
Software-as-a-Service predecessors.
In particular, the oft-neglected QA team will balloon in importance.
Rather than being an understaffed team to fight bugs, it will be a
center of technical excellence and become mission-critical. Maybe
it'll even get a rebrand to something catchy like the "Eval team".
Fortunately, quality will be a key source of accumulating advantage.
When SDS firms consistently deliver great,
measurable results, their clients will gradually downsize
teams previously handling those tasks. Just like AWS coaxed a
generation of companies into going without a team focused on racks
and servers, great SDS companies will see their customers shed
internal know-how and come to rely on them. The cost advantage SDS
firms offer should just be too compelling to DIY.
The Billing Dilemma
This evolution in service delivery and quality assurance naturally
demands a parallel change in how SDS companies capture value.
The previous generation of software companies primarily billed their
services on a per-seat basis. This made sense for SaaS firms selling
workflow tools used by many employees. But Software-Delivered
Services are fundamentally different—they're doing much of the work
itself. This naturally should mean fewer human users. At the same
time, these new AI-driven services often have serious marginal
costs. Running complex, multi-step AI workflows isn't cheap.
So these companies face a dilemma: How do they bill for this new
thing?
They can't use the old per-seat model. Nor can they bill on a fixed
basis, as that leaves no room for account expansion and doesn't
account for their variable costs.
The solution many are converging on is a piece-work model. Instead
of billing per seat, they're billing per unit of value delivered.
Despite the investment community's love of per-seat recurring
business models, many SDS companies are going usage-based.
This isn't entirely new in the tech world. Cloud vendors like AWS
have been using consumption billing for years. But what's
interesting is how this model intersects with a much older, larger
market: professional services.
Professional services firms have always billed for discrete units of
work. Now, Software-Delivered Service companies are doing the same
thing. They're turning labor into software, then selling that
software-powered "labor" on a per-unit basis.
Navigating the New Frontier
SDS firms can offer a clear value proposition against incumbent
professional services firms: buy the same thing from us, but at 10%
of the cost.
That said, professional services can often be pretty idiosyncratic.
Customers often show up with weird or somewhat vague needs. While
LLMs greatly expand what's possible, they still can't fully match
the flexibility of human-led services.
To succeed, SDS firms must find opportunities that thread the needle
with what the models can do and what customers want. This could mean
finding categories where expectations are clear enough to compete
head-on. They will probably need to build up substantial
customer-facing teams to smooth over gaps that LLMs can't yet fill.
In essence, successful SDS companies will embody the "full-stack startup" concept introduced by Chris Dixon a decade ago.
Perhaps an enterprise VC associate will make an SDS market map. But
here's a few interesting examples I see:
- Sierra: Founded by Brett Taylor, Sierra offers customer service and charges per closed ticket. I don't know them well, but I assume they integrate with existing support software and plug their AI right into it.
- Lighthouse: Solving the visa problem for exceptional technologists. They sell on a per-visa basis and obsess over quality. On the face of it, it's a services business but it's deeply LLM driven.
- Crescendo: Backed by General Catalyst, Crescendo acquired an existing call center and is reimagining it with GenAI. Similar to Sierra, they only charge when pre-defined performance metrics are hit.
- Micro1: Using AI agents to vet and interview job candidates. They either sell direct staffing services or they bill in a way that maps to the number of interviews conducted by their agents.
- Exec.com: At Exec, we develop AI-powered training for sales and customer-facing teams. Our AI conducts live roleplays with employees and provides real-time feedback. We bill on a per-trained employee basis.
It's still early days and there will be a wide variety of
experiments on both scope of work and billing models.
Closing Out
Over the past two decades, software permeated every corner of the
economy. Trillions were poured into "digital transformations" to
bring offline processes online. This mostly meant encoding workflows
and standardizing data. But now software can go further.
Before LLMs, it felt like the software world was approaching a
plateau for new startups. The SaaS playbook was well-established and
left only small, fiercely competitive niches for newcomers. Now the
aperture has widened and the grounds are fertile for startups.
So, the long march of software eating the world continues. SDS firms
can credibly attack new markets measured in the
trillions
of dollars, one billable unit at a time.
The companies that figure out how to do this will reshape entire
industries. They'll combine the scalability and efficiency of
software with the high-touch, outcome-focused approach of
professional services. That will require a maniacal focus on quality
and a adoption of new billing models. It'll be hard, but the prize
is huge.
Thank you to Ben Braverman, Amelia Adams, Chris Power, Sam
Kirschner, Justin Mares, Pete Huang, AJ Chan, Minn Kim, and Nick
deWilde for reading early drafts and helping refine my thinking.