Agentar 2.0

Ant Digital's "Business Intelligence Super Factory" core platform for the era of intelligence allows enterprises to build, deploy and manage intelligence in batches like Lego.

Enterprise-grade AI platform

https://www.antdigital.com/

Visit Website

In-depth Report

Agentar 2.0 is the core platform of the "Business Intelligence Super Factory" released by Ant Digital (an enterprise-level technology company under Ant Group) at the 2026 World Artificial Intelligence Conference (WAIC). It will be officially unveiled on July 19, 2026. It packages Ant's self-developed large models, agent development frameworks and industry experience accumulated in finance, energy, overseas and other fields, allowing enterprises to build, deploy and manage agents in batches "like building Lego". The goal is to advance AI from "tool efficiency improvement" to "organizational productivity upgrade". The platform is the first batch of pre-configured nearly 200 position-level digital expert templates and hundreds of subscribing Skills-level tools, focusing on plug-and-play, low threshold, and standardization. Its greatest confidence comes from the actual results in the financial industry, but to replicate it on a large scale across thousands of industries and small and medium-sized businesses, it still has to overcome the hurdles of cost, ROI and trust.
Ant Digital Technology (Ant Digital Technology) is the technology sector of Ant Group for corporate customers. It has been out of the spotlight for a long time and mainly serves high-threshold and strong risk control industries such as finance, energy, and medical care. Therefore, ordinary users rarely directly perceive it, but almost everyone in the financial circle knows it. According to public information, almost all state-owned banks and joint-stock banks and more than 60% of local commercial banks in China are using Ant Digital’s intelligent capabilities. The Agentar platform itself is not new. It has been online for one year and two months and has implemented more than 300 professional agents in banks, securities, insurance and other institutions, covering core financial scenarios such as wealth management, intelligent risk control, investment research and operations. This version 2.0 is the first time that Ant Digital has systematically positioned itself as a hosting platform for the "Business Intelligence Super Factory", upgrading from single-point agent development to a complete "model-inference-operating system-application" production line. At the launch site, Ant Digital CEO Zhao Wenbiao, CTO Yan Ying, and Ant Group CEO and Ant Digital Chairman Han Xinyi all appeared. The endorsement at the market level is also very direct: According to IDC reports, the overall market size of China's financial large models, agent applications and services will reach 2.436 billion yuan in 2025, and Ant Digital ranks first with a 13.3% share; in China's 1.75 billion yuan privatized market for intelligent agent development platforms, Ant Digital ranks among the top non-cloud vendors.
The core selling point of Agentar 2.0 is “out-of-the-box digital employees.” Enterprises do not have to train models or write agents from scratch, but can directly choose from nearly 200 position-level digital expert templates, such as strategy, finance, supply chain, legal affairs, ESG and other roles, and then add hundreds of subscribing Skills-level tools. The platform also has a built-in MCP service plaza and pluggable industry know-how components. By combining positions and business needs, enterprises can build an AI team with professional knowledge, business processes and tool calling capabilities. The typical picture given by Ant Digital is: a new energy company wants to go overseas to build a factory. The management only needs to set goals, and then a team composed of hundreds of agents including strategy, finance, supply chain, legal affairs, ESG, etc. will execute the entire process independently. In their words, enterprises "for the first time truly have an AI team working around the clock." Supporting this experience are several self-developed bases. The base model LingDT focuses on "extreme Token efficiency" and is used to save money and speed up AI deployment for enterprises; the Agentix agent operating system is responsible for the full process scheduling of agents, allowing them to work collaboratively "off the production line"; the large model service platform DTMaaS cooperates with Agentar 2.0 on the delivery and tool chain sides to achieve "plug and play". This production line, which connects models, reasoning, operating systems and applications, is Ant Digital’s answer to the demand that “what companies want is not parts, but complete vehicles.” It is worth mentioning that Agentar has been connected to the Alipay AI open platform and has become its core development engine, which allows Ant Digital to further move from serving large financial institutions to becoming the base for AI upgrades for tens of millions of small and medium-sized merchants.
As a platform-level product for enterprises (B-side), Agentar 2.0 currently does not have a public standardized price list, and its business model is closer to a combination of "platform subscription + industry solutions". Digital expert templates and Skills-level tools are provided in a "subscription-ready" form, and enterprises can activate them on demand. Ant Digital emphasizes that this modular and standardized approach is cheaper and faster to get started than the traditional service-heavy model of "professionals stationed on site and project-based coaching". Its business logic is based on the inclusive path of Alipay's payment code back then: first, it polished its capabilities to the financial level in the most difficult financial industry, then quickly copied this mature solution to energy, travel, consumption and other industries, and finally reached a large number of small and medium-sized merchants through the Alipay ecosystem. In other words, Ant Digital wants to sell "intelligent productivity" as an infrastructure like water and electricity.
Judging from implementation cases, positive feedback focuses on efficiency and cost. The Bank of Ningbo relied on Agentar to build an "intelligent decision-making pipeline", and the accuracy of complex question and answer increased from 68% to 91%; the energy company Linyang Zhiwei built a "power trading pipeline", reducing labor costs by 60%, and increasing the speed of analysis and strategy generation by more than 20 times; the photovoltaic operation and maintenance intelligent agent jointly developed by GCL New Energy and Ant Digital compressed key exception identification and feedback time to less than 1 minute. There are also many voices from the long-tail industry. Tang Qian, the founder of Hangzhou PetTalk, said that Agentar's agent development framework has helped a lot in product design, research and development, and promotion, greatly reducing early investment costs and increasing the speed of product release. They turned the experience of senior pet doctors into a "digital clone" that allows owners to get basic treatment suggestions by taking pictures and asking questions in the early morning. The doctor will call back for confirmation the next day. Frontline doctors reported that it can also help young doctors quickly narrow down dozens of examinations to three or five. However, it should be noted that most of these cases come from the official accounts of Ant Digital or partners, and are of a "model room" nature, lacking independent third-party horizontal evaluation and scattered feedback from large-scale real users. There is currently a lack of public verification as to whether ordinary small and medium-sized enterprises can reproduce these figures.
Many media have put Agentar 2.0 into the larger narrative of "AI Super Factory". The South China Morning Post reported that not only Ant, but also technology giants such as Tencent, Alibaba, and Baidu are promoting AI solutions for real enterprise scenarios. The competition for domestic enterprise-level AI agents has been fully upgraded; Lenovo has also launched an overall "AI factory" plan in an attempt to turn computing power into a hydropower-type inclusive infrastructure. Industry consensus is that in the first half of 2026, the global AI narrative is shifting from "how smart the model is" to "whether the intelligent agent can actually work and enter the enterprise process." IDC predicts there will be 2.2 billion active digital workers by 2030. Against this background, Ant Digital is widely regarded as an "invisible trump card" by virtue of its deep involvement in the financial industry - being able to take the first place in the market in the financial hard core of high complexity, high security requirements, and high knowledge density is considered to be a strong proof that its capabilities can be transferred to other industries.
The first risk is "concept first". "Business Intelligence Super Factory" is a very imaginative marketing framework, but the released information is more about vision, number of templates and sample cases. It lacks open trial of the platform, public pricing and independent benchmark testing. It is difficult for outsiders to judge the actual delivery quality and versatility. The second level is ecological binding and autonomy. The entire solution is highly dependent on Ant's self-developed LingDT, Agentix, DTMaaS and Ant's blockchain system, and is also deeply coupled with the Alipay ecosystem. This is convenient for customers who are already in the Ant system, but it also means strong vendor lock-in, and the freedom of enterprise migration and multi-cloud selection is questionable. The third level is the debate over the necessity of “blockchain trusted identity”. Ant Digital advocates that "AI creates intelligence, and blockchain creates trust." It gives each Agent a trusted identity and puts decisions and fund flows on the chain. The direction sounds right, but for most companies, whether blockchain is necessary for agent accountability and data traceability, and whether it will bring additional complexity and cost, still needs to be tested in real scenarios. The fourth level is competition and implementation costs. Tencent, Alibaba, Baidu, Byte and even Lenovo are all vying for the cake of enterprise-level intelligence, and there are more and more homogeneous narratives. What small and medium-sized businesses are most concerned about is still real ROI: Plug and play sounds great, but truly embedding digital experts into their own business processes and accumulating exclusive data often still requires considerable configuration and operational investment.
Agentar 2.0 is most suitable for medium and large enterprises that are already within the Ant Digital system or have strong needs for financial-level security and compliance, especially in industries with strong supervision and high knowledge density such as banking, securities, insurance, and energy - these are the areas where Ant Digital is best at and has the most solid models. For organizations that want to deploy agents in batches but don’t want to build a complete AI infrastructure themselves, it provides a relatively worry-free “whole vehicle” route. For small and medium-sized merchants, they can first start with the lightweight capabilities of Alipay's AI open platform, verify it with specific and clear-bounded single job scenarios, and then evaluate whether it is worth investing in depth. Teams that pursue independent and controllable technology stacks, want to avoid vendor lock-in, or require flexible selection of multiple models and multiple clouds should include the costs of ecological binding and compare them horizontally with solutions such as Tencent, Alibaba Cloud, and Baidu. As an alternative, you can pay attention to Baidu Qianfan, Alibaba Cloud Bailian, Tencent and various open source agent frameworks (such as Dify, Coze, etc.), which also focus on enterprise-level agent platforms.
Agentar 2.0 is a key upgrade for Ant Digits to commercialize and platform financial-level intelligence capabilities. The "super factory" idea is in line with the industry trend of "letting agents really work in enterprises" in 2026, and the actual results in the financial field are also strong enough. The real point is not the number of templates at the press conference, but whether it can copy this set of capabilities at a low cost and reproducibly to thousands of industries outside of finance, as well as small and medium-sized businesses. The direction is right and the foundation is strong, but from "model room" to "standard configuration", real, independent and verifiable implementation results must speak for themselves.

User Reviews

Stephen_Phillips_2021

—

The accuracy rate of Bank of Ningbo's complex questions and answers has soared from 68% to 91%. This data is quite impressive. You must know that the accuracy requirements in financial scenarios are abnormal. One wrong answer may be a compliance incident or complaint. Being able to achieve this level stably in real business shows that it has indeed entered the core workflow, rather than stopping at the surface of customer service copywriting.

JosephPerez_88

—

The nearly 200 position-level digital expert templates sound very intimidating, but in actual use, you still need to adjust them twice. The template is only the starting point, not the end point.

m6irf

—

Having been engaged in B-side delivery for so many years, I most resonate with what Yan Ying said: "What companies want is not parts, but complete vehicles." In the past, when building an agent for a client, I had to save the model, vector library, and tool chain myself, and it would take two months to just tune it up. Agentar 2.0 unifies the entire production line from models, reasoning, operating systems to applications. In theory, it can indeed save a lot of trouble. However, when it comes to customer sites, data access and permissions still have to be peeled off. Don't think too foolishly about it.

Brody665_eth

—

I have reservations about the blockchain issuing a set of trusted identities to each Agent. The direction is right, but most companies don't need to confirm the rights and go to the chain at all. It feels like Ant wants to squeeze in its ten years of blockchain wealth and sell it.

DragonDropPerry

—

To be honest, what impressed me most was not the number of templates, but the Trusted AI 5 rating of the Academy of Information and Communications Technology. What financial institutions fear most is model illusion. Being able to get the highest rating shows that it has really worked hard on its trusted links. Otherwise, the bank's core business would not dare to be touched by AI.

清风_6

—

Ant Digital is really strong in the financial circle. It covers almost all major state-owned banks and 60% of city commercial banks use it. Most people can’t dig this moat.

orangepanda834

—

LingDT focuses on token efficiency, which is very important to enterprises. A large part of the high cost of deploying AI is the inference overhead.

Gabriel616

—

Ant Digital wants to replicate the inclusive payment method used in the past and let roadside shops also use AI employees. The idea is beautiful, but what banks want is independence, trustworthiness, and deep customization. What pet stores want is cheapness, ease of use, and ready-to-use. These are two completely different businesses. The ability to serve large customers cannot be directly equated with the ability to serve small and medium-sized businesses. Whether they can serve these two bowls of water at the same time depends on the subsequent execution.

Jeffrey_Foster

—

We are engaged in electricity trading, and we were a little moved when we saw the case of Lin Yangzhiwei. Manpower is reduced by 60% and strategy generation is 20 times faster. If this is true, then the ROI is too impressive. I’m going to talk to their sales team about the implementation details.

贾红_1

—

I opened a gym in passing and used their AI to do membership operations. I was worried at first whether it was just another gimmick. As a result, the conversion of new customers and card renewals did increase by more than 10% in the few months of trial operation, and the overall performance of the store also improved. This kind of AI that can truly cut into the daily operation process and directly see the changes in data is what our store is willing to pay for.

Zachary.Evans_Plus

—

It is essentially the Chinese version of Copilot Studio. Microsoft's set has already generated $37 billion in ARR. Whether Ant can replicate this path in China is worthy of long-term tracking.

薛强

—

The case of digital clones in pet medical care really touches the pain point. If a pet has an accident in the early morning, you can take a photo and get advice on how to deal with it. The doctor will return for a visit the next day. This kind of long-tail industry needs this kind of AI the most.

JerryRivera_Plus

—

It’s too eco-friendly. LingDT, Agentix, DTMaaS and the blockchain are all self-developed by Ant and are deeply tied to Alipay. Once you get on board and want to change suppliers, it’s basically impossible. When selecting risky enterprises locked by such manufacturers, you really need to carefully consider them in advance. Don’t try to save trouble for a while and end up being stuck. The cost of migration can be painful.

73s14bq2n_n

—

Ant Afu Abao is the C-side ace, and Ant Mathematics is the one who quietly does big things. This time he finally came to the forefront.

许轩军

—

Enterprise-level intelligence is a real thing this year. Tencent, Alibaba, Baidu, and Huawei Pangu are all competing for this piece of cake. The narrative is becoming more and more homogeneous. Ant’s differentiation is the signature of financial-level credibility. In the end, it still depends on who can implement large-scale implementation first. Just having a benchmark project is not enough.

LandonBaile_y

—

Green Tea Restaurant uses AI to build 7x24 kitchen quality control and site selection models. This implementation scenario is quite down-to-earth and much better than those AIs that can only write PPT.

JudithCox_202430

—

I think the logic of the data flywheel is that the more merchants it serves, the more industry components and skills it accumulates, the lower the cost of new customer acquisition, and the more you use it, the more you understand the business. This is where the imagination of platform business lies.

Sara.Richardson

—

The press conference was full of hype, but unfortunately neither trial nor pricing was open. It is purely a B-side game, and the threshold for evaluation is still high for small and medium-sized enterprises to get started.

CeylanKıraç

—

The whole vehicle plan sounds great, but the cost of implementation is the last word. If you want digital experts to really understand your business, you have to feed them exclusive data, configure processes, and do alignment. Officials will not write down this part of the investment and cycle in PPT. Only when you actually work on the project will you know how deep the water is.

高昊磊

—

The financial inference model Agentar-Fin-R1 has taken first place in several financial benchmarks and can surpass open source models with the same number of parameters. The path of vertical deep cultivation is indeed better than general large models in the professional field.

EugeneHernandez_2024

—

There are 200 templates, hundreds of Skills, and an MCP service plaza. The ecosystem is quite complete, but I don’t know if non-ant system customers will be acclimated to it.

MariaKim

—

Standardized modularity, low cost, these words are the key to impressing a small company with a limited budget like ours. We really can't afford the heavy delivery project system.

purplesnake428

—

After watching the announcement on the spot, the metaphor of "building an assembly line like building Lego" is indeed vivid, but Lego also needs instructions. What small and medium-sized enterprises lack is the person who can do it.

Agentar 2.0

In-depth Report

User Reviews

Similar Products