Service as a Software: the New SaaS

Service as a Software (noun):A software system that leverages humans to perform a service end-to-end, reasoning, acting, and delivering outcomes that were previously the exclusive domain of skilled labor leveraging software.

This redefines what "addressable market" means for technology companies. Global enterprise software spending is roughly $1 trillion.¹ Global services spending across professional, financial, and business services exceeds $6 trillion.² Every prior generation of software sold tools to the people doing the work, now Service as a Software will agentically execute the work itself increasing the addressable market for technology companies by 7x.

We Have Seen This Before

The advertising industry already completed this transformation, and the language it used is instructive. In 2008, roughly 80% of digital ad spending was negotiated by humans through insertion orders.³ By 2024, 91% of display advertising was transacted programmatically.⁴ The word the industry chose was "programmatic," meaning software that executes decisions autonomously based on rules, data, and optimization targets. Today we would call that same concept "agentic." The terminology has changed but the architecture is identical. An agent is a programmatic system with a foundation model as its reasoning engine. Google’s Performance Max campaigns generate creative, select audiences, allocate budget, and optimize toward a conversion target with zero human input on individual decisions.⁵ That is an agent. Google simply built it a decade before the word became fashionable.

In this vertical the media buyer did not get a better Software as a Service tool, rather the media buying service got abstracted into a Service as a Software platform that agentically buys media.

The same pattern will arrive in every service vertical where the work follows codifiable rules.

The Next Era for SaaS: Tool → Network → Fulfillment

This pattern has a consistent three stage arc. Software begins as a tool that makes the human faster. Then it becomes a network that connects supply and demand (often with integrated payments). The final frontier will be when it becomes the fulfillment layer that performs the work. Programmatic advertising completed this arc in roughly fifteen years. At Covalent, we believe AI will compress the timeline to five or fewer.

Vertical examples	Tool Era	Network Era	Fulfillment Era
Advertising	Ad servers (DoubleClick)	Ad exchanges (2008)	Performance Max (2022)
Transportation	Dispatch software	Uber marketplace (2014)	Autonomous vehicles (202n?)
Procurement	Coupa / Ariba	Supplier networks	AI negotiations/fulfillment
Customer Service	Zendesk/Intercom	Outsourced call centers	AI voice reps/chatbots
Marketing	HubSpot / Marketo	Affiliate networks	AI campaign generation/execution
Design	Sketch/Figma	Upwork	AI UI/UX generation

Uber illustrates this most intuitively: Today Uber operates a network; it connects riders to human drivers and takes a 25% commission. When autonomous vehicles reach scale, Uber absorbs the driver entirely. The take rate moves from 25% to nearly 100% gross margin on the ride.⁶ Uber becomes the service. The network becomes a fulfillment engine. This same progression is visible today in procurement (from purchase order software, to supplier matching networks, to AI that negotiates contracts and places orders autonomously), in AP (from invoicing tools, to payment rails, to AI that collects, reconciles, and resolves disputes), and in marketing automation (from email sequencing tools, to intent data networks, to AI that generates creative, selects channels, and optimizes spend without human involvement).

A Framework for Prediction

Which services will software absorb first? Sequoia's Julien Bek offers a useful lens: the ratio of intelligence work (complex but codifiable rules) to judgement work (taste, experience, instinct).⁷ I would add three additional variables: outsourcing penetration (has the buyer already accepted external delivery?), data density (does volume create a compounding advantage?), and regulatory friction (does licensure block autonomous delivery?).

Service as a Software: Readiness Scorecard

Variable	High Readiness	Medium	Low Readiness
Intelligence Ratio	>70% rule-based	40 - 70%	<40%
Outsourcing Penetration	>50% outsourced	20 - 50%	<20%
Data Density	High volume, structured	Moderate	Sparse, unstructured
Regulatory Friction	Internal/unregulated	Standardized licensing	Strict licensure

What About AI Thrivers?

For the right Services/SaaS company, every task completed generates proprietary training data that improves the model, which lowers cost and raises quality simultaneously. Traditional services firms face the opposite: growth requires proportional headcount, which compresses margins. In a true SaaS company, the marginal cost of fulfillment will equal the marginal cost of token processing, or near zero. The new SaaS economics will look like the old SaaS economics, but addressing a $7T opportunity!

The companies that win will look like software businesses on an income statement and services businesses on an invoice. Software multiples assume $1T of TAM. Service as a Software assumes $7T.⁸ This is a great area for Covalent to invest.