With the rapid iteration of AI technology, Agent products are springing up. From demo to real application, from proof of concept to commercialization, we have witnessed the rapid evolution of this track. In the face of the many Agent products emerging in the market, how to objectively assess their actual value and development potential has become an issue worthy of in-depth discussion.
Through in-depth evaluation of six representative Agent products, this paper tries to establish a relatively objective evaluation system and analyze the characteristics and positioning of different products.
Agent valuation: a three-dimensional analytical framework
There is a simple but effective analytical framework that we can use when assessing the value of Agent products:
Product Value = Execution Capability × Credibility × Frequency of Use
Each of these three dimensions corresponds to:
- implementation capacity:: Stability of the product in fulfilling user tasks and delivering usable results
- degree of credibility: Whether users are willing to entrust important tasks to it and whether the process is transparent and controllable
- frequency of use: Whether the product can be quickly called up when users need it, integrated into the daily workflow
A scoring system of 0-3 points is used for each dimension, and a total score of 8 or more can be considered a quality product with market competitiveness.
In-depth analysis of six mainstream Agent products
Selection of Assessment Subjects
For this review, we chose six representative products that have actual users in both the B- and C-end: Manus, Buckle Space, Lovart, Flowith Neo, Skywork, and Super Maggie.
| Diethylammonium chloride | Positioning type | Main Scenarios | Featured Functions |
|---|---|---|---|
| Manus | general purpose | Task automation | One sentence on automatic dismantling of execution |
| Button Space | general purpose | Multi-scenario collaboration | MCP Systematic Integration |
| Lovart | vertical | Design Creation | End-to-end design delivery |
| Flowith Neo | general purpose | complex inference | Visualize the chain of thought |
| Skywork | vertical | office document | In-depth research report generation |
| Super Maggie. | vertical | Enterprise OA | Process automation |
Detailed Product Analysis
Manus: concept-first explorers
Manus' biggest contribution is to show the market a new paradigm for Agents - moving from conversational interaction to task-based execution. The user only needs to describe the requirements in a single sentence, and the system automatically disassembles, plans and executes the complete process.
performance in practical use:
- The execution success rate is around 20%, there is still room for optimization
- Relatively independent portal with limited integration into the daily work environment
- Occasional link breaks in the middle of the process, affecting the user experience
Scoring: 1 point for ability to execute, 2 points for credibility, 1 point for frequency of use, total 4 points
Link:manus.im
Buckle space: a generalized program for complete architecture
Buckle Space excels in technical architecture, realizing a complete chain of MCP invocation, task orchestration, and result delivery. It has a high degree of engineering and is able to handle all kinds of anomalies in a stable manner.
Core strengths:
- Link integrity to support complex task scheduling
- MCP systematic integration with strong tool calling capability
- Transparent process with traceability at every step
Scoring: 3 points for ability to execute, 2 points for credibility, 2 points for frequency of use, total 12 points
Link:coze.co.uk
Lovart: professional assistant in the field of design
Lovart is outstanding in the vertical field, and can truly achieve "delivery-level" design output. Users only need to put forward the requirements, and the system will automatically handle style selection, color matching, layout design and other professional aspects.
Practical application examples:
- Public main visual design: generate a complete brand visual program at one time
- MBTI suite production: serialized design output in a unified style
- Marketing material production: end-to-end delivery from requirements to finished product
Scoring: 3 points for ability to execute, 3 points for credibility, 2 points for frequency of use, total 18 points
Link:lovart.ai
Flowith Neo: Unique Visual Interaction
Neo is unique in its interaction design, visualizing the AI's reasoning process as visual nodes where users can see each thinking step. Its concurrent processing capability is outstanding, capable of executing multiple subtasks at the same time.
Technical Highlights:
- Supports highly concurrent task execution
- Ability to handle long contexts
- Complete transparency of the reasoning process
Scoring: 3 points for ability to execute, 3 points for credibility, 1 point for frequency of use, total 9 points
Link:flowith.io
Skywork: the professional player in the office scene
Skywork specializes in office document generation, especially research reports and PPT creation. Its best feature is the traceability function - every conclusion has a clear data source.
Test Case: Kingsoft Office Stock Analysis Report
- Automated access to securities websites, annual reports and other data sources
- Generate complete reports with financial analysis and industry comparisons
- Sources are cited in each paragraph to support fact-checking
- Final output of a ready-to-use PPT file
Scoring: 3 points for ability to execute, 3 points for credibility, 2 points for frequency of use, total 18 points
Link:tiangong.cn
Super Magee: The OA Assistant with Deep Integration
McGee represents another class of Agent - the embedded system assistant. Instead of focusing on content generation, it focuses on automating processes within the organization.
Core Functions:
- Invoice recognition and automatic archiving
- Intelligent Approval Process Judgment
- Automated processing of enterprise reports
- Smart Fill for Travel Requests
Scoring: 3 points for ability to execute, 2 points for credibility, 3 points for frequency of use, total 18 points
Link:letsmagic.co.uk
Specialized Agent vs Generalized Agent: Different Development Paths
From the results of the evaluation, the three products with the highest scores (Lovart, Skywork, and Super Magee) are all specialized Agents in verticals.This reflects some of the characteristics of the current market.
The Core Advantages of Specialized Agent
Depth over breadth
Specialized Agents have accumulated a large amount of Know-How in specific areas, which includes not only the use of tools at the technical level, but more importantly, an in-depth understanding of industry standards, user expectations, and quality requirements.
Skywork, for example, not only knows how to generate PPTs, but also understands the logical structure of business reports, data presentation, credibility requirements and other specialized knowledge.
Greater deliverability
Specialized Agents often provide "out-of-the-box" results, rather than just materials or drafts; Lovart generates designs that can be used directly for commercial purposes, and Skywork's reports can be submitted directly to clients.
Growth Opportunities for General Purpose Agents
While generalized Agents face challenges in some areas, they also have unique value:
- adaptable: Ability to address diverse mission requirements
- learning ability: can be continuously optimized through user feedback
- Platform value: Opportunity to become a unified portal for multiple vertical scenarios
Trust Mechanisms: Important Considerations for Agent Commercialization
As a number of Agent products began to try to commercialize, the willingness of users to pay has become an important indicator to test the value of the product.
Payment model analysis
The current dominant payment model is the point system:
| offerings | Payment Standard | Cost per mission | quality requirements |
|---|---|---|---|
| Lovart | $10 ≈ 1000 points | ~300 points ($3) | Needs to be a one-time success |
| Skywork | $10 ≈ 10,000 points | ~500 points ($0.5) | Need for high quality output |
| Flowith | $10 ≈ 10,000 points | variable pricing | Need for reliable implementation |
Key elements of confidence-building
In a paid environment, each task execution is equivalent to a "trust bet" by the user. Building user trust requires attention in the following areas:
The importance of interpretability
Great Agent products have the following characteristics:
- Transparent visibility of the implementation process
- Support for mid-course interventions and corrections
- Results are traceable and verifiable
This explains why Skywork's traceability and Buckle Space's process visualization are so important - they allow users to "hand over control" and still feel secure.
Traffic Entry and User Mind: Key Elements for Product Success
Even with excellent execution and credibility, Agent products may still face bottlenecks if they fail to solve the problem of being "found" and "remembered".
Status of entry competition
On the desktop side, we can already see fierce competition between various products vying for the document processing portal. When a user drags and drops a document, multiple AI assistants may pop up with processing options at the same time.
Mind Occupation Strategy
Scene Binding
Successful Agent products tend to form strong bindings with specific usage scenarios:
- Think Lovart when you need to design
- Think of Skywork when you need to write a report
- Think of Magee when you need to handle OA processes
sensorless access
The strategy employed by SuperMagic is noteworthy - it doesn't require users to actively invoke it, but rather appears naturally in the business process. This "silent" approach represents an interesting development.
Implications for Product Development
For developers of Agent products, there is a balance to be struck between:
- Product Capacity Building and User Reach
- Functional integrity and ease of use
- Depth of specialization and breadth of application
Conclusion: Trends and Opportunities for Agent Products
By analyzing six mainstream Agent products in depth, we can observe several interesting trends:
- Specialization and generalization go hand in hand: Different types of Agent products each have their own value space, with specialized products having clear advantages in specific areas and generalized products performing better in terms of adaptability
- Deliverability as a core competency: Users are increasingly focused on the quality of the actual output of Agent products, not just the presentation
- Growing importance of trust mechanisms: As the commercialization process advances, transparency, controllability and consistency become key elements in building user trust
- There is a clear trend of portal integration: How to find the right place in the user's workflow is becoming an important factor in the success of the product
The future Agent market is likely to show a diversified development trend, with differently positioned products each finding its own user group and value space. The key is to accurately understand user needs and find the best balance between product capability, user experience and business model.
