# Services as Software (Service-as-Software): Meaning, Models and Economics

Source: https://trymimetic.com/guides/services-as-software

[Guides](/guides) / Revenue QA

# Services as software: what it means, who is building it, and how it gets paid

Service as software means selling finished work instead of a tool to help people do it. An AI system, usually with people supervising it, does the bookkeeping, resolves the support ticket, prepares the tax return or runs the campaign, and the customer pays for the result. This guide covers where the idea came from, the three business models built on it, the economics investors are underwriting, and the places it breaks.

## Where the term came from

The phrase has older uses, including a 2017 Analytics Magazine article, "Service as a Software: A new paradigm". Its current meaning was popularized by Foundation Capital in April 2024, in an essay by Joanne Chen and Jaya Gupta: **"AI companies are leading a transition from Software-as-a-Service to Service-as-Software"** ([Foundation Capital](https://foundationcapital.com/ai-service-as-software/), 2024).

Two years later, investors are making the same argument under other names:

- Sequoia's Julien Bek: **"A copilot sells the tool. An autopilot sells the work,"** and **"If you sell the tool, you're in a race against the model"** ([Services: The New Software](https://sequoiacap.com/article/services-the-new-software), 2026).
- Bessemer calls them AI-native services: **"AI can deliver the work itself and the outcome is the product"** ([Bessemer](https://www.bvp.com/atlas/owning-the-outcome-bessemer-s-ai-native-services-evaluation-framework), 2026).
- General Catalyst: **"Customers care about outcomes, not products"** ([The Future of Services](https://www.generalcatalyst.com/stories/the-future-of-services), 2025).
- Dwelly, a UK lettings rollup: **"AI makes Services the new SaaS"** ([Dwelly](https://www.dwelly.group/blog/dwelly-raises-170m-series-b), 2026).

## Why investors care: the services budget is bigger

The case rests on where the money is. Sequoia argues that **"For every dollar spent on software, six are spent on services."** General Catalyst's written estimate puts US services spending above $6 trillion against a software market of about $370 billion. These ratios vary by source and most are not fully sourced, but in every version the labor budget is many times the tool budget. Service-as-software goes after the labor budget.

## Three ways companies are building it

**1. AI-native services startups.** Build the service from scratch with AI doing most of the work, and sell it as a service: bookkeeping, claims, legal intake, support. Emergence Capital argues this is the stronger path: **"Organic-then-inorganic compounds. Inorganic-then-organic decays"** ([Emergence](https://emcap.com/thoughts/roll-up-or-build-why-the-fastest-path-to-a-venture-scale-ains-business-is-the-one-founders-keep-avoiding), 2026).

**2. AI rollups.** Buy established services firms with customers already in place, then rebuild how they work. Long Lake's CEO, Alex Taubman, on why he buys instead of selling software to these firms: **"it's actually really hard to sell software into these services industries ... if you can't beat them, join them"** (No Priors, 2026). Multiplier's founder Noah Pepper made the same switch: **"I realized I was barking up the wrong tree by trying to build a SaaS business"** ([TechCrunch](https://techcrunch.com/2025/06/18/multiplier-founded-by-ex-stripe-exec-nabs-27-5m-to-fuel-ai-powered-accounting-roll-ups/), 2025). We track 81 of these; see [AI rollups in 2026](/research/ai-rollups-2026) and [Long Lake, the largest](/guides/long-lake-ai-rollup).

**3. Software priced on outcomes.** Stay a software company, but charge per result. Intercom's Fin: **"Fin is priced at $0.99 per outcome"** ([Fin pricing](https://fin.ai/pricing)). Sierra: **"Sierra gets paid only when we complete a task for you"** ([Sierra](https://sierra.ai/blog/outcome-based-pricing-for-ai-agents), 2024).

## The economics being underwritten

- **Margins.** Mayfield's Navin Chaddha on AI-first services: **"You could have blended margins of 60% to 70% and produce 20% to 30% net income"** ([TechCrunch](https://techcrunch.com/2025/06/29/why-ai-will-eat-mckinseys-lunch-but-not-today), 2025). A traditional services firm runs at a fraction of that.
- **Growth and margin together.** General Catalyst's target for AI-enabled roll-ups is a "Rule of 60": firms growing at single digits reaching **"10-20% growth"** with **"30-40%"** margins.
- **Productivity at the desk.** A tax preparer at a Multiplier firm: **"Prior to Multiplier, I sent out 14 tax returns in a 14-hour day. This year, I was able to send 24 tax returns in a 10-hour day"** ([Multiplier](https://www.multiplierholdings.com)).

These are targets and company claims. None is audited. General Catalyst's Marc Bhargava: **"we're only right if the AI enablement works"** ([Cognitive Revolution](https://www.cognitiverevolution.ai/the-revolution-in-services-general-catalysts-ai-rollup-strategy-with-marc-bhargava/), 2025).

## How it gets priced

The natural pricing for service-as-software is per outcome, and Foundation Capital's original essay argued exactly that: **"An outcome-oriented approach aligns the cost of the unit of software with the associated business value."** Adoption is still early. In Kyle Poyar's 2025 survey of 240+ companies, **"5% of survey participants said their primary pricing model is outcome-based right now. However, 25% said they expect it to be outcome-based by 2028"** ([Growth Unhinged](https://www.growthunhinged.com/p/2025-state-of-b2b-monetization), 2025). Foundation Capital expects the shift to be **"messy, with false starts, hybrid tiers, and plenty of margin surprises."** For how this plays out in marketing services, see [outcome-based pricing for AI-native agencies](/guides/ai-native-agencies-outcome-based-pricing).

## Where it breaks

- **Liability stays human.** Multiplier's Pepper: **"An AI's answer is worth zero until a professional puts their name on it"** ([Pulse 2.0](https://pulse2.com/multiplier-raises-35-million-series-b-to-acquire-professional-services-firms-and-build-ai-tools/), 2026).
- **Efficiency can eat the revenue.** Nathan Benaich and Nikola Mrksic: **"Services businesses aren't inefficient by accident. They're inefficient by design"** ([Air Street](https://press.airstreet.com/p/ai-rollup-mirage-technology-2025), 2025). A firm billing by the hour earns less when the work gets faster, unless it changes how it charges.
- **The outcome has to be counted.** Foundation Capital: **"Outcome is ultimately a function of the customer's product, not just your software."** Intercom says it still relies on **"basic rules to track whether a resolution was 'successful'"** ([Stripe case study](https://stripe.com/customers/intercom-pricing)). When the outcome is a lead, a booking or a sale, the count comes from the customer's website and analytics, and in our audits of 936 live sites one in five had duplicate tracking corrupting those counts (see [conversion tracking setup](/guides/conversion-tracking-setup)).

## What this means if you run or buy a services business

1. **Decide what you sell: hours, seats or results.** Service-as-software only pays off if the pricing moves with it.
2. **Define the outcome before you price it,** and name the system of record when two systems disagree.
3. **Fix the front door.** Most of the growth in the investor models comes from serving more customers with the same team, and those customers arrive through the website, the phone and the booking flow. In our check of 146 companies bought by AI rollups, 35% showed no Google Analytics, Tag Manager or HubSpot tracking on their homepage ([AI rollups in 2026](/research/ai-rollups-2026)).

## How Mimetic helps

Mimetic checks a business's site the way customers use it, and whether its tracking adds up: conversions counted twice, consent settings dropping them, tags that stopped firing, and records that disagree with analytics. For a company being paid on outcomes, that is the check that makes the outcome number trustworthy.

## Common questions

**What is service as software?**
A business model where AI does the work itself, such as bookkeeping, support or tax preparation, and the customer buys the result rather than a tool. The term was popularized by Foundation Capital in April 2024.

**What is the difference between SaaS and service as software?**
SaaS sells software that people use to do their work, usually priced per seat. Service as software sells the finished work, often priced per outcome, with AI doing most of it and people supervising.

**Who coined service as software?**
The phrase has older uses, including a 2017 Analytics Magazine article. Its current AI meaning was popularized by Joanne Chen and Jaya Gupta of Foundation Capital in April 2024.

**What are examples of service as software?**
Intercom's Fin charges $0.99 per resolved conversation and Sierra charges per completed task. AI rollups such as Long Lake, Thrive Holdings and Multiplier buy services firms and run them on shared AI platforms.

**Is service as software the same as an AI rollup?**
No. An AI rollup is one way to build it: buying existing services firms instead of starting from scratch. AI-native services startups and outcome-priced software companies are the other two.

Written by Coleman Hindes, founder, Mimetic.
Published 2026-11-22.

### If you get paid on outcomes, check that the outcomes are counted right.

[Check a site](/)
