AI-Native Services: selling the outcome, not the software

TEN13 · Published July 2026 · TEN13 Investment Thesis

AI-native services (AINS) are companies that use artificial intelligence to deliver an entire professional service — legal work, insurance underwriting, healthcare administration, customer support — from start to finish, and then sell the completed outcome rather than a piece of software. This is different from a traditional software-as-a-service (SaaS) company, which sells a tool for a human professional to use: an AI-native services firm takes on the work itself, assumes the delivery risk, builds its own processes and IP, and captures the value that human labour used to capture. It matters because the global services market is roughly ten times the size of the global software market — more than $6 trillion against $695 billion — and AI has only recently become capable enough to operate inside it without a human in the loop. TEN13, a Brisbane-based Australian venture capital syndicate, argues that AI-native services (AINS) may be the most significant structural shift in technology investing since SaaS, and is backing companies in the category both globally and in Australia.

Executive summary

AI-native services (AINS) describes a new business model that TEN13 calls "service as software": a company builds an AI system that performs a professional service end-to-end and sells the delivered outcome, not the underlying software. Some of the market's most disciplined venture investors have converged on this thesis at the same time. General Catalyst, Sequoia, Emergence Capital, a16z and Lightspeed have collectively backed more than 25 AI roll-up platforms since 2024, spanning legal, insurance, IT managed services, accounting and healthcare — each underwritten to reach venture-scale outcomes, not the services-industry multiples expected of a traditional private-equity roll-up.

The scale of the opportunity comes from a simple spending gap. In 2025, for every dollar spent on software globally, roughly ten dollars were spent on services: the lawyers, insurance brokers, healthcare administrators, contact-centre operators, accountants and compliance professionals who do work that software has so far only assisted with. AI can now do that work itself, start to finish, with limited human direction. The bet behind the AI-native services thesis is that owning the delivered outcome is a fundamentally larger prize than selling the tool that assists with it.

Global software spend stands at $695 billion. Global services spend — covering legal, accounting, BPO contact centres, recruitment, healthcare administration and consulting, i.e. work traditionally delivered by human capital — stands at more than $6 trillion, and is now deliverable by AI systems at comparable accuracy. (Sources: WIPO Global Innovation Index, 2024; The Business Research Company, Professional Services Market, 2024.)

Global services spend covers legal, accounting, BPO contact centres, recruitment, healthcare administration and consulting — work traditionally delivered by human capital, now deliverable by AI systems at comparable accuracy. Sources: WIPO Global Innovation Index (2024); The Business Research Company, Professional Services Market (2024).

The scale of the incumbent response reinforces the thesis. In May 2026, OpenAI launched DeployCo, an AI services venture that raised more than $4 billion at a $14 billion launch valuation, backed by TPG, Advent, Bain Capital and Brookfield. As TEN13 Managing Partner Stew Glynn puts it:

"SaaS companies sell software. AI-Native Services firms do the work and sell the outcome."

The fact that the world's most valuable AI lab is not stopping at software, and is making a play for services too, is a signal that the category is being taken seriously at the highest levels of the industry. (Source: OpenAI announcement; press reporting, May 2026.)

What is an AI-native services company?

An AI-native services company is a business that uses artificial intelligence, not human professionals, as the primary operating layer for delivering a professional service — and it sells the finished outcome to its customer, not a software licence. The category is easy to conflate with adjacent models such as AI copilots or traditional business process outsourcing (BPO), but one distinction anchors the thesis: AI-native services (AINS) companies sell outcomes, not software. They take on the risk of delivering a service and build their own processes and intellectual property to deliver it themselves, end to end, with no human required in the seat.

This differs from a copilot company, whose software assists a human professional who remains responsible for the output, and it differs from a traditional services or BPO firm, whose delivery scales with headcount rather than with AI. An AI-native services firm's marginal cost of delivery is inference — the cost of running its AI models — rather than labour hours, which is why its economics can converge toward software-like gross margins even though the company is, in substance, delivering a service.

The market landscape: from thesis to a well-funded wave

The AI-native services (AINS) thesis has moved from a contrarian idea to a well-funded wave in a short period. Since 2024, General Catalyst, Sequoia, Emergence Capital, a16z and Lightspeed have collectively backed more than 25 AI roll-up platforms spanning legal, insurance, IT managed services, accounting and healthcare. Each of these platforms is underwritten to reach venture-scale outcomes — not the more modest returns multiples typically expected of a traditional private-equity roll-up.

The autopilot thesis — the idea that AI can run a professional workflow without a human in the loop — was uninvestable for as long as AI models sat below the human competence bar on the relevant tasks. That changed in roughly eighteen months, as frontier AI models crossed into expert territory across coding, mathematics and computer-use benchmarks at the same time. On SWE-bench Verified, a coding benchmark, frontier model accuracy rose from below 30% to more than 84% in about eighteen months. On OS World, a computer-use benchmark that TEN13 considers the closest proxy for the admin-style work AI-native services firms sell, frontier accuracy rose from around 5% to more than 85% over the same period. TEN13 treats 80% accuracy as "the 80% line" — the point at which a model's performance on a task is close enough to expert territory that it can be meaningfully compared to a human professional in the seat.

(Source: Epoch AI, "Frontier performance across benchmarks" (CC-BY), epoch.ai. The full step-chart plots named frontier models — including Qwen2.5-72B, o3-mini (high), Claude 3.7 Sonnet, GPT-5 (high), GPT-5.2 Pro, GPT-5.5 Pro (x/high) and Claude Opus 4.7 (max) — against four benchmarks (SWE-bench Verified, FrontierMath Tiers 1–3, FrontierMath Tier 4, and OS World) from October 2024 onward; the individual accuracy values for each model/benchmark point on that chart could not be read reliably from the PDF and are listed at the end of this article rather than estimated.)

Despite this jump in raw model capability, adoption of AI agents remains heavily concentrated in one profession. As of 2026, 49.7% of AI agent tool calls are made within software engineering — software engineers adopted AI agents first, and every other professional category, including legal, insurance, healthcare and accounting, sits in single-digit percentages of adoption. TEN13 reads this as evidence that the work AI-native services companies target has barely begun to be automated, leaving substantial headroom for future adoption. (Source: Sequoia Capital, Agent Tool-Call Adoption, 2026.)

TEN13's framework for why now splits professional work into two layers of increasing judgement. Layer 1, execution — filing, drafting, coding, logging, matching and reporting, the repeatable, high-volume tasks a professional performs every day — is now solved and autonomous: models can run these tasks end to end with no human directing them. Layer 2, pattern application — translating a known problem into a working answer using a domain playbook, which is the judgement layer of most professional services — is where the frontier is now crossing the bar, and it is the layer where AI-native services (AINS) companies are built to operate. (Framework: Sequoia Capital, "Services: The New Software," 2026; TEN13.)

Who is investing in AI-native services

As of July 2026, General Catalyst, Emergence Capital and Lightspeed lead the field with five or more tracked AI-native services investments each, with Y Combinator, Andreessen Horowitz (a16z), Bain Capital and Sequoia close behind. TEN13 attributes the United States' position as the leading indicator in this category to three factors: depth of venture capital, density of AI talent, and the scale of existing professional-services infrastructure that AI-native services companies can displace or acquire.

Bubble-chart legend from the source: "Most active" covers investors with 5–6 tracked investments; "Multiple" covers 3–4; "Two investments" covers 2; TEN13 is highlighted separately as the report's publisher. Source: TEN13 Research, July 2026.

The market map: 35 companies across ten categories

TEN13 tracks 35 venture-backed companies operating across ten AI-native services categories, spanning insurance, healthcare revenue-cycle management, IT services, customer service, legal services, property services, accounting, professional services, construction and software. This map is illustrative and non-exhaustive; TEN13's own portfolio holdings — Arintra and Source — are highlighted separately below.

Total: 35 companies across 10 categories. Source: illustrative; TEN13 Research, Crunchbase, investor portfolio pages, July 2026; non-exhaustive.

The full company roster by category, as tracked by TEN13:

Insurance (7)

Reserv, Pace, Gyde, Harper, Strala, Shepherd, Vivere

Healthcare RCM (6)

Arintra (TEN13 portfolio), Candid Health, Anterior, Amperos Health, Autonomize AI, Prosper AI

IT Services (6)

Distyl AI, 8090, Shield, Titan, Mechanical Orchard, Treeline

Customer Service (4)

Wonderful, Giga, Crescendo, Smith.ai

Legal Services (3)

Norm Law, Manifest, Crosby Legal

Property Services (3)

EliseAI, Long Lake, Dwelly

Accounting (2)

Modus, Hanover Park

Professional Services (2)

Source (TEN13 portfolio), Multiplier

Construction (1)

E3Tech

Software (1)

Beacon

Companies defining the category

The AI-native services category is taking two dominant shapes, according to TEN13's analysis: "autopilot-native" firms built from inception to run a workflow without a human in the seat, and "AI-enabled roll-ups" that acquire traditional service-delivery businesses and then deploy AI across the acquired workflows. The six companies below are illustrative examples of each shape.

Sources: Emergence Capital, "The AI-Native Services Playbook" (2026); Sequoia Capital, "Services: The New Software" (2026); company reporting.

Copilot vs autopilot: why the distinction matters

The distinction between a "copilot" and an "autopilot" AI company is the central dividing line in TEN13's AI-native services thesis, because it determines whether a company sells software or sells a delivered outcome. A copilot is software that assists a human professional: it suggests, drafts and flags, while the human stays in the seat, owns the output and applies judgement. The productivity gain from a copilot is real, but it is bounded by the professional's own time and attention, because a human is still required to review and approve every unit of work.

An autopilot — the model that defines AI-native services (AINS) — is an AI-native operating system that executes a piece of work end to end, delivers the outcome, and closes the loop with no human in the seat. The customer buys a delivered outcome, not a tool, and every improvement in the underlying AI model makes the service faster, cheaper and more defensible, because the company captures that improvement directly rather than passing it to a user who still has to do the work.

Framework: TEN13, adapted from Emergence Capital and Sequoia Capital (2026).

Why this is possible now: the capability curve

AI-native services (AINS) companies are only possible now because frontier AI models have crossed from below the human competence bar into expert territory, in roughly eighteen months, across the specific tasks that professional services work requires. On SWE-bench Verified, a coding benchmark, frontier model accuracy rose from under 30% to more than 84% over that period. On OS World, a computer-use benchmark — which TEN13 regards as the closest proxy for the admin-style work AI-native services firms actually sell — frontier accuracy rose from around 5% to more than 85% over the same roughly eighteen-month window.

TEN13 calls 80% accuracy "the 80% line": once a model clears that bar on a given task, its performance is close enough to expert territory that it becomes reasonable to compare it directly to a human professional doing the same task. (Source: Epoch AI, "Frontier performance across benchmarks" (CC-BY), epoch.ai. As noted above, the underlying step-chart tracks named frontier models against SWE-bench Verified, FrontierMath Tiers 1–3, FrontierMath Tier 4 and OS World from October 2024 onward; individual data points on that chart could not be read with confidence and are listed at the end of this article.)

Adoption has not yet caught up with capability. As of 2026, 49.7% of AI agent tool calls occur within software engineering alone, while every other professional category TEN13 tracks — legal, insurance, healthcare, accounting — sits in single-digit percentages of adoption. Software engineers adopted AI agents first because their work was easiest to benchmark and verify; TEN13's reading is that the professional-services work AI-native services companies target has barely begun to be automated, leaving significant headroom ahead. (Source: Sequoia Capital, Agent Tool-Call Adoption, 2026.)

Three ways AI enters a service business

TEN13 identifies three distinct ways that AI can enter a service business, distinguished by how much of the delivery workflow AI actually owns and by the gross margin each produces. As AI moves from being a tool a professional uses to being the operating layer that runs the business, delivery decouples from headcount and gross margin converges toward software-like economics.

Source: TEN13 analysis; Emergence Capital; company disclosures (2026).

The playbook: buy distribution, upgrade delivery

TEN13's playbook for building an AI-native services company is to buy distribution first, then upgrade delivery with AI — the reverse sequence of the traditional SaaS playbook. The SaaS playbook builds a product, then a sales and marketing team, then acquires customers one by one, spending a dollar of customer acquisition cost (CAC) for every dollar of growth. AI-native services (AINS) companies run a different sequence: they either sell "service as software" direct to customers, or they acquire traditional service providers to inherit distribution and a customer base overnight. A bookkeeping firm, a BPO, or a property manager is cheap to acquire because its earnings are capped by headcount; deploying AI across the inherited workflows breaks that cap. TEN13 summarises the approach in four words: buy distribution, upgrade delivery.

The entry point into an existing customer relationship is typically a vendor swap. When a customer already pays a third party for a service, displacing that vendor is a procurement decision, not an organisational change, which makes it far easier to execute than convincing a company to change how it works internally. Replacing one outside vendor with a better, cheaper AI-native option is the path of least resistance. TEN13 frames the outsourced task the customer already pays for as "the wedge" that already has a budget attached to it, while the work the customer still does in-house is "the long-term TAM" (total addressable market) the AI-native services company can eventually capture.

TEN13 defines an "AI rollup" as a company that systematically acquires service-delivery capability across a vertical, using AI as the shared operating layer to compress cost and expand margin across every business it acquires. General Catalyst has pioneered this model most visibly, with Eudia (legal), Crescendo (customer service), Titan MSP (IT), Long Lake (multi-vertical) and Dwelly (property) in its portfolio.

Sources: General Catalyst portfolio; Sequoia Capital (2026); TEN13 analysis.

What is an AI rollup, and how does it differ from a PE rollup?

An AI rollup is a company that systematically acquires service-delivery businesses across a single vertical and uses AI as a shared operating layer to compress the cost of delivering the service and expand gross margin across every acquisition. This differs fundamentally from a traditional private-equity (PE) rollup, which captures margin mainly through procurement leverage, headcount cuts and financial engineering — a PE rollup can succeed on multiple arbitrage alone, buying a business cheaply and selling it at a higher multiple without materially touching how the business operates day to day.

An AI rollup cannot take that shortcut. Its entire margin case depends on actually replacing the delivery layer inside each business it acquires — deep systems integration, executed under time pressure, across every new acquisition — which TEN13 notes is far harder to repeat reliably across many deals than it is to prove once in a single flagship acquisition. Where a PE rollup's moat is largely financial (scale, leverage, negotiating power), an AI rollup's moat compounds through an operating system that becomes more accurate with every acquisition's proprietary workflow data — a compounding advantage that accelerates with scale rather than merely repeating it.

The compounding data flywheel

The core moat in an AI-native services (AINS) business is a data flywheel that compounds with every customer engagement, and it is structurally different from the data advantage a copilot vendor can build. A copilot vendor gathers only thin, workflow-level data, and its customers typically resist letting an external software provider train its models on their data. An AI-native services firm, by contrast, is the service provider itself: it manages the entire workflow and the data that workflow generates, and it improves using direct customer data every day. As Jake Saper of Emergence Capital puts it:

"If you're not building this flywheel from day one, you're just a services company that uses AI tools."

(Source: Emergence Capital, "The AI-Native Services Playbook," March 2026.)

The underlying foundation model that an AI-native services company uses is a commodity input available to every competitor; the proprietary signal the company generates case by case, engagement by engagement, is what TEN13 considers the irreplaceable advantage. As concrete examples, TEN13 cites Harper's proprietary data on which insurance placements succeed, and Crosby Legal's proprietary signal on which contract language creates disputes — data that exists in no publicly available training set. A firm that starts by owning Layer 1 (execution) builds the flywheel that lets it credibly move up to own Layer 2 (pattern application) as well: the frontier of what the company can competently own shifts outward with every engagement it completes.

Business model characteristics

AI-native services (AINS) companies have gross margins that converge toward software economics even though they are, in substance, delivering a service — because their cost structure and their scalability curve both differ fundamentally from traditional services businesses. Traditional professional-services businesses carry high cost of goods sold (COGS) in the form of professional labour, which typically produces gross margins of 20–40%. Traditional SaaS businesses carry low COGS but heavy sales, marketing and product spend instead. AI-native services companies incur inference and technology cost with a leaner human-in-the-loop team, which lands their overall margins on a trajectory that converges toward software-like economics. Crescendo, for example, already runs 60–65% gross margins across its acquired call centres.

Scalability follows a fundamentally different curve as well. Traditional services businesses scale headcount roughly in line with revenue; AI-native services companies scale non-linearly, because AI takes on an increasing share of the production work as the company grows. Emergence Capital flags revenue per employee as the critical indicator that separates a genuine AI-native business from what it calls "Mirage PMF" — a services company that is simply adding more human staff and calling the result "AI."

TEN13 describes a shared four-step architecture common to successful AI-native services companies:

1. Pick a service category with a high intelligence-to-judgement ratio and existing outsourcing.
2. Enter at the most repeatable task, where accuracy is easiest to prove.
3. Build domain credibility before scaling distribution.
4. Invest in the data flywheel from day one.

TEN13 summarises the underlying logic as: the outsourced task is the wedge; the insourced work is the long-term total addressable market (TAM). As Julien Bek of Sequoia Capital puts it:

"The next $1T company will be a software company masquerading as a services firm."

(Sources: Emergence Capital, 2026; Crescendo company reporting; Sequoia Capital, "Services: The New Software," March 2026.)

Risks and open questions

TEN13 identifies six open questions where the AI-native services (AINS) thesis could fall down, and pairs each risk with the counter-argument where one exists in its analysis.

Source: TEN13 analysis, July 2026.

Why Australia, and why now

TEN13's view is that Australia is early to AI-native services (AINS), and that this is precisely the opportunity: the services pool available to be repriced dwarfs the software pool by roughly ten to one, and TEN13 expects it to be repriced slowly and unevenly at first, but permanently, by founders who understand that the business of the next decade is doing professional work at a fraction of the historical cost and a multiple of the historical margin.

TEN13 believes the founders who will define Australian AI-native services will share a specific set of qualities: genuine domain credibility in a chosen vertical, an architecture that treats AI as the operating layer for the business rather than as a bolt-on feature, proprietary access to workflow data, and the discipline to prioritise AI-native delivery over bespoke, services-style revenue. TEN13's framing is direct: the window is open now, and the category will consolidate before it closes. As Y Combinator put it in its Summer 2026 Request for Startups:

"AI-native companies that don't sell software, they sell the service. Instead of giving you a tool, they just do the work."

TEN13's position in AI-native services

TEN13 is a Brisbane-based Australian venture capital syndicate founded in 2019 by Stew Glynn and Steve Baxter. TEN13 invests first cheques of A$300,000 to A$2 million from pre-seed through to Series A and beyond in technology companies, and has invested more than A$125 million across 54 companies to date. Within AI-native services specifically, TEN13's own tracked portfolio includes Arintra, in the healthcare revenue-cycle management category, and Source, in the professional services category — both named in TEN13's market map of 35 venture-backed AI-native services companies.

Arintra is TEN13's portfolio example of the AI-native services thesis in practice: it operates in healthcare revenue-cycle management (RCM), a category TEN13 tracks as one of the most active in the AI-native services market map, alongside companies such as Candid Health, Anterior, Amperos Health, Autonomize AI and Prosper AI. TEN13 is also one of the more active investors tracked in the category overall, with two tracked AI-native services investments — Arintra and Source — placing it alongside larger investors such as Lux Capital, Thrive Capital, Peak XV, Coatue and SV Angel on TEN13's own investor bubble chart.

TEN13's stated position is that it is actively investing in AI-native services companies, both as a participant backing founders in the category and as a firm publishing research to help define it. Investors can join the TEN13 network by signing up at hub.ten13.vc, and founders building an AI-native services company can reach TEN13 by submitting a pitch or emailing investments@ten13.vc.

Frequently asked questions

What is an AI-native services company?

An AI-native services company is a business that uses AI, rather than human professionals, as the primary operating layer to deliver a professional service end-to-end — such as legal work, insurance underwriting, or healthcare administration — and it sells the delivered outcome to its customer, not a software licence.

What is an AI rollup?

An AI rollup is a company that systematically acquires service-delivery businesses across a single vertical and uses AI as a shared operating layer across every acquisition to compress the cost of delivery and expand gross margin.

How is an AI rollup different from a private equity rollup?

An AI rollup differs from a private equity (PE) rollup because its margin case depends on actually replacing the delivery layer inside each acquired business with AI, through systems integration under time pressure, whereas a PE rollup can succeed on multiple arbitrage alone — buying businesses cheaply and selling at a higher multiple without materially changing how they operate.

What gross margins do AI-native services businesses achieve?

AI-native services businesses achieve gross margins in the range of 60–80%+, according to TEN13's analysis, compared with 45–50% for "AI-augmented" services businesses and only a modest lift for "AI-assisted" businesses that use AI purely as a productivity tool; Crescendo, for example, already runs 60–65% gross margins across its acquired call centres.

Which VCs are investing in AI-native services?

The most active venture capital investors in AI-native services, as tracked by TEN13 as of July 2026, are General Catalyst (6 tracked investments), Lightspeed (5) and Emergence Capital (5), followed by Y Combinator (4), Andreessen Horowitz (4), BCV (3), Sequoia Capital (3), and several investors with two tracked investments each, including TEN13.

Is Australia behind on AI-native services?

Australia is early to AI-native services rather than behind, according to TEN13's thesis, which frames Australia's current lack of scaled AI-native services companies as an open opportunity: the services pool available to be repriced by AI is roughly ten times the size of the software pool, and TEN13 expects the category to consolidate before the window closes.

What is the difference between a copilot and an autopilot in AI-native services?

A copilot is software that assists a human professional, who stays in the seat, owns the output and applies judgement; an autopilot is an AI-native operating system that executes the work end-to-end and delivers the outcome with no human in the seat, which is the model that defines an AI-native services company.

What does "service as software" mean?

"Service as software" describes the AI-native services business model in which a company delivers a professional service through an AI system and sells that service directly, the way a SaaS company sells software, rather than delivering the service through billable human labour.

Sources

• WIPO Global Innovation Index (2024)
• The Business Research Company, Professional Services Market (2024)
• OpenAI announcement; press reporting (May 2026), re: DeployCo
• TEN13 Research (July 2026)
• Crunchbase (July 2026)
• Investor portfolio pages (July 2026)
• Emergence Capital, "The AI-Native Services Playbook" (March 2026)
• Sequoia Capital, "Services: The New Software" (March 2026)
• Company reporting (Crescendo, RFA/Titan MSP, and others cited above)
• Epoch AI, "Frontier performance across benchmarks" (CC-BY), epoch.ai
• Sequoia Capital, Agent Tool-Call Adoption (2026)
• General Catalyst portfolio (2026)
• TEN13 analysis (July 2026)
• Y Combinator, Summer 2026 Request for Startups
• TEN13: hub.ten13.vc · investments@ten13.vc

This article is for informational purposes only it is not investment advice.