---
title: "How can AI applications find sustainable monetization opportunities?"
type: "News"
locale: "en"
url: "https://longbridge.com/en/news/290968681.md"
description: "Consumer AI subscriptions face fragility due to rising token costs and low user willingness to pay. Sustainable monetization lies in vertical AI platforms that integrate deeply into high-value workflows, such as legal (Harvey) and finance (Farther), or transaction infrastructure (Adyen). These platforms transform industry-specific knowledge and processes into durable assets, focusing on quantifiable business results rather than model parameters, thereby creating competitive advantages as basic models become commoditized."
datetime: "2026-06-26T12:43:24.000Z"
locales:
  - [zh-CN](https://longbridge.com/zh-CN/news/290968681.md)
  - [en](https://longbridge.com/en/news/290968681.md)
  - [zh-HK](https://longbridge.com/zh-HK/news/290968681.md)
generator: "portal-rs"
---

# How can AI applications find sustainable monetization opportunities?

The consumer AI subscription model is facing a dilemma: token costs continue to rise, while users' willingness to pay is struggling to keep pace. This structural tension makes this business model extremely fragile. More sustainable AI commercialization may occur in scenarios with high-value users, deep workflow integration, and direct linkage to real business results. This is precisely the economic premise for the establishment of vertical AI tracks. The 9th installment of "Agentic Economy" will dissect the cases of Harvey, Farther, and Adyen to see how they built competitive advantages in the context of commoditizing basic models. It will also address two more difficult questions: Will these advantages still hold true when token subsidies fade? And what does it mean when OpenAI and Anthropic begin sending engineering teams to enterprise clients? As the marginal effect of parameter competition diminishes, once scarce intelligence is rapidly evolving into commoditized public infrastructure. This is causing the business models of lightweight applications that rely on simply calling third-party APIs and lack the ability to embed into specific scenarios to collapse at an accelerated pace. However, commoditization is never the end. Every time an emerging technology becomes widespread, it shifts value opportunities from those who possess the technology to those who can truly implement it. This pattern has spurred the rapid rise of vertical AI platforms. Vertical AI platforms refer to AI application layers that delve into specific industries, deeply encapsulate the capabilities of general-purpose models, and rearrange business processes around specific workflows. Through self-built evaluation systems (Eval) and multi-agent architectures, these platforms are reducing the underlying basic models to readily replaceable computing components, thereby firmly locking the industry's core workflow assets within the system. Its core essence lies in eliminating friction in business processes, transforming complex professional work into a sustainable, accumulative systemic asset. To understand the emergence of this new track, it's crucial to clarify that what businesses and professionals pay for is never the scale of the underlying model's parameters, but rather its ability to deeply embed itself within internal workflows, forming a data loop and driving actual revenue. This is why high-paid legal experts, financial advisors serving high-net-worth clients, and leading merchants with massive transaction volumes are becoming the focus of competition for the next generation of vertical AI platforms. These individuals wield budget decision-making power, bear compliance responsibilities, and are driven by clear business results. Whether it's helping a lawyer earning a thousand dollars an hour save ten hours, or assisting a wealth advisor in increasing assets under management and optimizing after-tax profits, the business value they create can be directly quantified. This precise binding with high-value production entities is the economic foundation upon which vertical AI can succeed. Currently, this exploration is mainly unfolding in two directions. Firstly, AI is being used to reorganize professional workflows, significantly reducing operating costs that were previously only borne by large institutions. In the legal and wealth management sector, high-barrier tasks such as compliance, risk control, and professional delivery are being systematically handled through technology platforms, allowing professionals to complete higher-density work with fewer resources. Secondly, there is the reconstruction of transaction infrastructure, reshaping the connection between merchants and agents. In agentic commerce, while the front-end intent interception and interaction are controlled by AI labs, the final transaction conversion still occurs within the merchant's infrastructure. Adyen Agentic acts as a universal translator, helping merchants connect once and participate in various AI shopping platforms across protocols without rebuilding the system for each new protocol. These three cases have different entry points, but all involve transforming core capabilities that were previously difficult to standardize in the industry into sustainably usable assets through systemic evolution. Harvey has accumulated legal judgment and industry knowledge; Farther has accumulated client relationships and tax optimization capabilities; Adyen has accumulated merchant product data, protocol adaptation, and settlement capabilities. This is precisely what Microsoft CEO Satya Nadella referred to as Token Capital: the long-term value of AI comes not only from the execution of single tasks, but also from the structured retention of human judgment, knowledge, and workflows within the system, forming self-iterable assets through continuous interaction. 190 Million ARR and 460 Million Computing Fees: Harvey's Unsustainable Scale Game Harvey is one of the highest-valued and fastest-growing examples in the current vertical AI wave. The potential and dilemmas of this logic are most concentrated in Harvey. This legal platform, which doesn't possess any generic model, boosted its ARR from $100 million to $190 million and reached a valuation of $11 billion within five months (August 2025 to January 2026) by deeply customizing the core workflows of law firms. This demonstrates that vertical platforms don't need to participate in the underlying struggles of generic models; as long as they truly understand industry tasks and reconstruct the daily work scenarios of high-value users, they can build strong commercialization capabilities. However, behind the impressive financial figures lies an ever-expanding computing power bill. Public data shows that Harvey's monthly token usage has grown from approximately 1 trillion to 12-13 trillion. Estimated at $3 per million tokens, its annualized theoretical inference cost is as high as $468 million. Even though this cost is currently temporarily reduced through discounts from major companies and technologies like Prompt Caching, the cost structure, constrained by external factors, means that once subsidies narrow, the bills will immediately rebound. Under this financial pressure, ARR growth is extremely difficult to translate into real cash flow, and instead constantly faces the risk of a backlash from scale. This reflects the cost paradox that native AI applications cannot avoid: the more popular a product, the higher its inference costs. Traditional SaaS has almost zero marginal cost, but in scenarios like law, with its long contexts and high inference density, every complex task consumes real computing power. Therefore, developing in-house models has shifted from a technological option to an inevitable choice driven by cost. Currently, Harvey is advancing its post-training strategy for proprietary models, collaborating deeply with Applied Compute to fine-tune open-source base models (such as GLM-5.1) specifically for the legal industry. According to the latest technical disclosures from both parties, the post-trained proprietary model achieved a rubric pass rate of 0.913 in Harvey's self-built Legal Agent Benchmark (LAB), improving from 0.853 to 0.853, surpassing GPT-5.5 xhigh and approaching Opus 4.8 Max. Cost reduction is equally significant. By replacing the evaluation model with a state-of-the-art large model and batch-processing multiple evaluation criteria, evaluation costs have been reduced by 40 to 100 times. This means Harvey can continuously iterate its evaluation cycle at a lower cost, and its proprietary evaluation system itself has become an accumulative competitive asset. Even more noteworthy are the changes behind the performance improvements. Several key behaviors, including output completeness, numerical accuracy, document traceability, and illusion suppression, have seen quantifiable improvements. During training, the number of times the model calls tools has continuously decreased, but each call is more precise, resulting in a decrease in total token consumption. In other words, what the model learns is how to work effectively in a specific tool environment, and these behavioral patterns accumulated through numerous legal tasks are more difficult to replicate externally than the model parameters themselves. Harvey's case illustrates that the competitive foundation of vertical AI platforms is extending deeper. While workflow design and client relationships are important, the post-training capabilities and control of open-source models, proprietary evaluation systems and data generation capabilities, multi-agent architectures, and inference cost optimization are becoming new sources of differentiation. Farther's Deorganization: Breaking the Binding of Traditional Large Brokerages to Advisors If Harvey compressed delivery costs within large professional service institutions, the wealth management platform Farther demonstrates how to help core talent break free from the organizational pull of traditional giants. Farther is a technology platform for independent advisors (RIAs), specifically recruiting wealth advisors who have left giants like Morgan Stanley, Merrill Lynch, UBS, and Goldman Sachs. In traditional full-service brokerage systems, advisors often bear low commission rates and heavy back-office administrative burdens. Farther's approach is to directly recruit advisors, integrating the back-office capabilities previously monopolized by large institutions into a unified platform: in addition to high commission rates, tax loss mitigation, direct indexing, private equity market access, compliance auditing, and document management are all built-in. Official data shows that its tax-intelligent algorithms alone can bring clients a 1% to 3% improvement in after-tax investment returns. This model has already received strong validation from the capital market. In May 2026, Farther completed a $150 million Series D funding round led by General Atlantic, officially joining the ranks of unicorns. Currently, its assets under management have exceeded $23 billion, including a star private banking team recently poached from Goldman Sachs' private wealth division, managing $1.5 billion in assets. The continued influx of independent wealth advisors indicates that the systemic ties that traditional large brokerages rely on are becoming ineffective, and independent practice is no longer a marginal option for a select few. Harvey focuses on improving the efficiency of professional delivery within law firms; Farther, on the other hand, built an independent platform from scratch, allowing advisors to obtain equal or even stronger back-office capabilities without relying on traditional large brokerages. While their entry points differ, both are redefining the way professional services are produced. Supported by this platform, complex investment tools previously limited to the ultra-high-net-worth (UHNW) departments of large institutions, such as direct indexing and private equity markets, can now be easily accessed by independent advisors, greatly expanding the scope of their professional work. Traditional SaaS can only handle shallow process automation such as recording and storage, and cannot handle complex execution such as decision-making and coordination. AI-native systems based on a multi-agent architecture are naturally suited to handle the gray areas between administrative execution and non-standard logical judgment, such as compliance review, personalized document writing, and asset allocation advice. These tasks, which previously required an entire back-office team, are being rapidly absorbed by the system. The Underrated Merchant Side: The Closed Loop of Agentic Commerce Discussions about Agentic Commerce remain heated, but current public attention is almost entirely focused on the consumer side—how AI assistants replace users in searching for products, comparing prices, and automatically placing orders. In contrast, the actual feedback from the merchant side is much more lukewarm. Walmart's conversion rate on its AI-native checkout (Instant Checkout) is currently only one-third that of the traditional click-to-redirect model; and the proportion of merchants truly integrating Shopify's AI checkout system will remain limited even in 2026. There is a clear gap between AI-activated demand and the actual completion of a transaction. This gap arises because agent transactions are a complex system. Understanding user intent is only the first step; converting demand into revenue requires end-to-end support including inventory verification, tax calculation, fraud prevention and risk control, fulfillment, and fund settlement. These capabilities are currently locked into merchants' local systems. At the same time, multiple smart agent payment protocols, such as UCP, ACP, AP2, Agent Pay, and Visa Tokenization, coexist and are incompatible with each other. Merchants have neither the incentive to adapt to each one nor the cost of fragmented technology. Adyen has launched Adyen Agentic to address this, using a three-layer modular API to cover different stages of the transaction chain: Agentic Feed: Responsible for standardizing and distributing merchants' product catalogs, pricing, and real-time inventory data to mainstream AI platforms; Agentic Cart: Integrates merchants' existing checkout, tax, fulfillment, and order management systems into the conversational commerce underlying layer; Agentic Payments: Handles identity verification, network risk control, and multi-channel fund settlement in agent-driven transactions. Once merchants integrate, Adyen can translate across different AI agent platforms and protocols, eliminating the need to rebuild the underlying system every time the market landscape changes. In the intelligent agent business ecosystem, while front-end AI labs and dialogue interfaces may capture user intent and traffic, substantial value conversion, transaction implementation, and financial closure still heavily rely on merchant-side infrastructure. Compared to the fiercely competitive front-end entry points, systematic integration services on the merchant side have a greater chance of becoming stable, chargeable underlying infrastructure. The Hidden Concerns of Vertical Platforms: The Penetration of Model Factories and the Restructuring of Token Costs As the market for low-priced general-purpose tools becomes saturated, the business logic of large model platforms relying on low-price subscriptions is gradually showing its fragility. When generalized functions like summarizing web pages and drafting emails are easily replaced, vertical platforms must focus on high-value customers who prioritize business results. However, the higher the value of an application, the more complex the competitive environment becomes. One source of pressure comes from the proactive expansion of model vendors' business boundaries. OpenAI and Anthropic are no longer content with simply being API wholesalers; instead, they are directly engaging with core customers through their frontline engineering teams (FDE). In April 2026, OpenAI partnered with Customers Bank, a $26 billion company, with its engineering team working on-site to develop intelligent agents for loan approval and account opening using local data. Anthropic, on the other hand, partnered with financial IT giant FIS, embedding its FDE team into its internal systems to develop anti-money laundering tools, leveraging FIS's extensive network of banks to directly reach the deepest levels of the banking business. This on-site collaboration model indicates that major model manufacturers are leveraging infrastructure channels to directly learn and replicate the internal business processes of high-barrier industries. Another pressure comes from the unsustainable token pricing logic. Currently, most cutting-edge foundational model tokens are actually sold at a loss after subsidies. With the high-frequency calls of enterprise-level multi-agent architectures, once subsidies from major manufacturers subside, vertical platforms that rely entirely on external cutting-edge API interfaces will find their computing power costs unsustainable. This pressure will be further amplified as inference demands increase. When hundreds of agents operating 24/7 interact frequently in the background, the demand for computing power will grow exponentially. However, constrained by the extremely long manufacturing cycles of machines like ASML lithography machines, the underlying hardware supply chain simply cannot keep up. For most daily operations, using cutting-edge large models to handle all tasks is itself a serious resource misallocation. This is precisely why Harvey had to collaborate with Applied Compute to establish a dedicated test set, a private evaluation system, and a manual annotation pipeline. Vertical platforms are not just about product development; they are undertaking highly complex cost engineering: precisely calculating the token loss for each task, clarifying which intermediate steps can be offloaded to low-cost open-source small models, which critical decisions must utilize high-priced flagship models, and where manual review should be initiated. In this context, a simple, aesthetically pleasing workflow interface is no longer sufficient to provide a sustainable competitive advantage. Mastering backend cost engineering is the key to the long-term survival of vertical AI platforms. Conclusion: Market Scarcity Returns to the Top of the Industry Chain When general-purpose large models become as readily available as water, electricity, and gas, the value of AI applications begins to concentrate at the very top and bottom of the industry chain. At this stage, the industry's scarcity attributes haven't disappeared: the upstream remains the core that cannot be standardized by algorithms, such as customer trust, complex non-standard judgments, and unstructured knowledge hidden in practitioners' experience; the downstream is the merchant network, carrying product data, compliance links, and payment channels. The essence of vertical platforms lies in transforming the professional experience of these high-value entities into sustainably accumulated token capital. This also determines that the competitive logic of the application layer is returning to pragmatism. The "scale narrative" that supported the software industry's rapid growth over the past decade is beginning to fail in the face of rigid constraints on computing power expenditure and physical supply chains. In this new cycle, the survival of application layer companies depends on a sophisticated cost and computing power arbitrage. With the decline of model price wars and limitations in computing resources, application platforms must find the optimal solution between cost and performance, rather than continuing to rely on capital injections. While large model vendors possess greater computing resources and frontline engineering teams, the most unique competitive advantage for agile vertical platforms and independent professionals remains the ability to transform their accumulated tacit expertise into system assets that basic model vendors cannot replicate. Avoiding broad competition for traffic and prioritizing the commercial closed loop for high-value production entities is the logic behind the long-term survival of vertical AI in the era of large model commoditization.

### Related Stocks

- [ADYEY.US](https://longbridge.com/en/quote/ADYEY.US.md)
- [OpenAI.NA](https://longbridge.com/en/quote/OpenAI.NA.md)
- [MSFT.US](https://longbridge.com/en/quote/MSFT.US.md)

## Related News & Research

- [Adyen names Nicole Olbe EMEA president, effective Sept. 1, 2026](https://longbridge.com/en/news/296448305.md)
- [Adyen Shares Jump as Earnings Rise and Growth Accelerates](https://longbridge.com/en/news/295766146.md)
- [3 Next-Gen AI Stocks That 3G Capital Bought in Q2](https://longbridge.com/en/news/296496173.md)
- [CRCL: Stablecoins and Arc infrastructure are driving global payments innovation and regulatory adoption](https://longbridge.com/en/news/296377931.md)
- [MRX: Record revenue and profit growth driven by infrastructure-led, recurring earnings and strong risk management](https://longbridge.com/en/news/296372054.md)

---
> **Disclaimer: This article is for reference only and does not constitute any investment advice.**