AINS Might Have Staying Power

Recently, I came across the abbreviation AINS for AI Native Services. The first place I saw it was in a piece by Jake Saper called The Death of Deloitte which argued that “service businesses that leverage both AI and humans to deliver holistic solutions to clients are poised to outgun and outpace the services behemoths that’ve dominated for the last 50 years.” Sarah Tavel’s piece Sell the Work, while it didn’t use the AINS terminology, is also a key part of this discussion. More recently, Emergence published the AINS Playbook, a list of some key observations on how to build a successful AINS business.

Emergence has always had a special resonance in my mind: Jason Green, who helped found the firm and is a legendary SaaS investor, was an early source of advice (and valuable criticism) for what we are building at Angular Ventures, engaging constructively with us when many others would not. Jake Saper and others at the firm have been friends and are continuing the tradition set by Jason. More importantly, Emergence is the firm that first publicly tied its fortune to SaaS, correctly identifying it as early as 2003 as a crucial trend in software and going on to back Salesforce, Zoom, and Veeva. Veeva—in particular—is one of the most impressive venture bets ever. So when the Emergence team hitches their fortune to another four-letter acronym, I pay attention. I think the AINS concept and term has staying power. It is the right formulation to capture the essence of what is currently happening in the market. Occasionally, I hear people try to use the term “Services as Software” (SaS), which is confusing and doesn't reflect the reality of the shift as accurately.

The AINS Playbook. Emergence’s AINS Playbook is a great place to start. I’d encourage every founder to read it—even if you don’t think your company is technically an AINS company. While Emergence argues that many SaaS companies have the option of becoming AINS companies, the majority of SaaS companies are already coming under pressure to move in that direction. Some are proactively experimenting with outcome-based pricing, which is the essence of AINS in the first place.

I want to specifically highlight three of Emergence’s points as particularly salient:

1. Mirage PMF: They warn against early false signals of product-market fit caused by a willingness to invest in customer-specific deployments that do not scale. We see this constantly in our dealflow; understanding which companies might have long-term compounding advantages is challenging. A large low-margin business is still a low-margin business.

2. Proprietary Data Moats: They highlight the importance of proprietary data access and generation as a potential moat. “When you become the infrastructure layer and not just the service layer, when your customer's entire data history and operational workflow runs through your system, the magnitude of that data and the depth of that integration make switching enormously painful. This is a qualitatively different moat than a simple API integration.”

3. Efficiency KPIs: They suggest revenue per employee and "Human Review Time" (HuRT) per unit of revenue/value as crucial KPIs to assess the degree to which a company is able to scale AINS-style profits.

The Human Last Mile. One framework I’m particularly interested in is the Human Last Mile. This may be the critical dividing line between AI-native SaaS (AINSaaS?) and AI-native Services (AINS). Both proprietary data and highly integrated operational workflows are critical moats for both AINSaaS and AINS companies, but AINS companies go a step further. By selling outcomes, they provide the "last mile" to their customers.

The last mile is not just the outcome itself (the legal contract draft), but the crucial intangibles that make that outcome valuable in a business context. These intangibles—trust, brand, peace of mind, liability, context—have historically been what set service companies apart. This is where they made their money, and this is where AINS companies (or SaaS companies adopting AINS-style strategies) can leverage AI to capture outstandingly high margins.

As a young analyst on Wall Street covering Akamai, I remember having that business explained to me by an industry expert: Akamai is paid $\$1$ for delivering content. It’s paid $\$99$ for guaranteeing that content gets delivered and taking on the liability in case it doesn’t. That is the difference between a technology and a service.

What drives value in AINS? We are just at the beginning of the AINS era. There will be many attempts, many investments, many big successes, and many small outcomes. One of the reasons AI is so exciting is the sudden explosion in service businesses that it will enable. The most popular heuristic for “which AINS company is valuable” thus far has been the salary pyramid. Nine VCs out of ten will sit you down and explain earnestly that they are focused on the “top of the salary pyramid” where human salaries are highest and, thus, the ability to pay and the value of replaced labor is greatest. This is why so much attention has been paid by the VC industry to legal (Harvey, Legora) and medical (Open Evidence). This makes sense, but it feels incomplete. We’re still early enough in the AINS evolution that new frameworks are emerging. Here are a few I’m thinking about:

• The Stamp-to-Prep Ratio. The value of an AI-Native Service is often gated by the "Human Last Mile"—the final moment where a human expert provides the "stamp of approval" required for a customer to accept the result. The scalability of an AINS business depends on the ratio between Preparation Time (the grunt work) and Stamping Time (the expert judgment). If the prep time is high and the stamping time is low, AI provides massive leverage by "ramming" significantly more high-quality drafts through the same human bottleneck. Conversely, if the expert's review (the stamp) is the time-intensive part, the AI’s ability to scale the business is structurally limited.

• The Digital-to-Physical Ratio. Beyond human leverage, the attractiveness of an AINS business is dictated by the extent to which a service can be delivered via bits rather than atoms. In heavy industries like industrial waste or metal inspection, the "physical" represents a structural drag on margins through logistics, manual sampling, and on-site labor. An elite AI-native service seeks to "digitalize the atom" by using remote sensors, computer vision, or IoT data to pull the core problem into the digital realm where AI can process it at near-zero marginal cost. If a business requires "boots on the ground" for every unit of revenue, it remains a traditional services firm with an AI veneer. If it can use AI to transform a physical inspection into a remote data-validation task, it achieves the scalability of software while capturing the high-value budgets of the physical world.

• Brand locus. The ability of an AINS company to create a brand, disrupt incumbent brands, and positively associate that brand with an AI-native company as opposed to a specific employee is crucial. Does the brand that confers peace of mind reside with a named person (your specific CPA at HR Block) or can it be subsumed into the larger brand itself? Similarly, how entrenched are the incumbent brands in the minds of customers and how positive or negative is it to associate automation with the outcome?

• The Exception Handling Ratio. Value in services is often created in how one handles the edge cases rather than the "happy path." An AINS company’s efficiency is determined by the ratio of the cost to handle an exception versus the value that exception handling provides to the customer. If every non-standard request triggers a collapse in margins due to expensive manual intervention, the business won't scale. A truly scalable AINS business uses AI to triage and route exceptions so effectively that the marginal cost of a "weird" request stays low—either because outliers are easy to automate or because they don't impact the customer's core needs.

• Human Training Loop Closure. In an open-loop system, humans fix the AI's mistakes, but that data never makes it back into the model to improve the service for the next customer. In a closed-loop AINS business, the "human-in-the-loop" isn't just a cost center; they are a data-labeling engine. Every time a human corrects a draft or adjusts a parameter, the system learns. The most valuable AINS companies will be those where the human review process is architected to train the underlying models, creating a compounding advantage that competitors—who might just be using off-the-shelf LLMs—cannot match.

• Cross-Customer Data Fungibility. The final framework I'm considering is how much the data learned from Customer A can be used to improve the service for Customer B. In some highly regulated or siloed industries, data must remain strictly isolated. However, in industries where "patterns of work" are fungible across the entire market, an AINS provider can achieve a "network effect of intelligence." The more customers they serve, the smarter the service gets for everyone, creating a winner-take-all dynamic where the incumbent’s AI becomes functionally superior to any newcomer's model simply because it has seen more "work."

We’re still in the early innings of defining AINS and figuring out how to build these businesses. If you are building an AINS business in Europe or Israel, or if you are thinking about AINS in new ways that can help me refine my framework—please reach out!

Gil Dibner

Could the future of software be fluid
How do we get the best of AI without losing the soul of software?

The future belongs to young missionary teams
Why it makes more sense betting on youth in the current moment

The AI-native enterprise playbook
Ten real-time observations on a rapidly evolving playing field

No more painting by numbers
It’s the end of the “SaaS playbook.

ENTERPRISE/TECH NEWS

Agentic Software - Oracle introduced a suite of “agentic applications” embedded across enterprise workflows, designed to proactively execute tasks across functions such as HR and finance. These systems move beyond responding to inputs; instead, they initiate and coordinate actions within existing processes. This reflects a shift in enterprise software from systems of record to systems of action, in which software is responsible for executing work rather than merely supporting it. The role of the application layer is expanding from interface to operator. “With Fusion Agentic Applications, we are moving enterprise software beyond passive systems of record… to applications that can reason, decide, and act.”

Category Repricing - Software stocks declined following Anthropic’s latest release, with capital rotating toward semiconductors. The market reaction reflects growing recognition that advances in AI are affecting differentiation at the application layer. As more functionality can be replicated or bypassed by AI systems, traditional SaaS offerings face pressure on margins and positioning. Public markets are beginning to reflect this shift, with capital reallocating toward the infrastructure layer underpinning AI capabilities.

HOW TO STARTUP

Trust Currency - PrometAI argues that early-stage success is shifting from attention to trust. Rather than prioritising growth speed or visibility, the focus is on consistency, credibility, and delivering real user value. Trust is built through repeated proof points — product quality, transparency, and follow-through — rather than one-off launches or short-lived traction. “The new currency is trust, not attention.” Startups are assessed less on initial traction and more on their ability to sustain and compound value over time.

HOW TO VENTURE

Partner Pathways- Ilya Strebulaev and Blake Jackson analyze a dataset of over 12,000 VC professionals across three decades to answer a simple question: what predicts who becomes a partner. By tracking individuals from junior entry points through their careers, the study identifies patterns in career outcomes. The findings show that progression is strongly linked to early access to investment responsibility and to exposure to decision-making. Those who join at more senior levels, rather than as analysts, are significantly more likely to reach partner. Venture is not a meritocracy in the abstract; it is a system where access to ownership and decision rights compounds over time.

Augmented Judgment-  Anna Atanasova outlines how AI is being integrated into venture workflows: not as a decision-maker, but as a force multiplier on everything before the decision. Sourcing, market mapping, memo drafting — all increasingly automated. The more interesting shift is where this leaves the investor. As information becomes abundant and cheap to process, differentiation moves away from access and towards interpretation. The edge is no longer who sees the most, but who knows what to ignore. “While the data is processed by AI, the critical thinking remains entirely ours.” AI is compressing the front end of venture — but in doing so, it is exposing something that was always true: judgment was never about processing information, but about forming conviction under uncertainty.

Angular Ventures mentioned on CTech piece The European VC gap in Israeli tech and why it matters now. "Structural gaps, not strategy, explain why European investors are missing from Israel's tech ecosystem. The small group of European funds that do appear share one thing in common: they invested in local infrastructure. Cardumen Capital, headquartered in Madrid, operates a full Tel Aviv office with an Israeli partner and has made 32 Israeli investments. Angular Ventures, nominally London-based, was founded by Gil Dibner, who splits time between London and Tel Aviv.”

Groundcover brings AI-Native observability to production analysis, running natively in customer clouds.

CruxOCM
Pipeline Hydraulics Engineer (Vancouver, Remote)

Reco
Solutions Engineer (US, Remote)

Sourcix
Backend Engineer (Tel Aviv)