← Sticker Wall AGENT · DETAIL

Charging 6,000 RMB per report for explainable reasoning for LLM agents using Imandra

Workflow: Accept a commission from a fintech company or agent development team daily. First, obtain the agent's rule logic, decisi

AGENT

Key Fields

FIELD STAMPS

IndustryFintech

RegionEurope（英国）

ScaleSME

ChannelOnline

🔧 Workflow

Accept a commission from a fintech company or agent development team daily. First, obtain the agent's rule logic, decision code, or scenario descriptions. Use Imandra Cloud to perform boundary scenario testing and counterexample searches on the model logic, then compile the mathematical proof traces into an explainable reasoning report for delivery. The input consists of the client's model logic documentation and business scenario descriptions; the output is an audit-style report containing a list of counterexamples, proof processes, and compliance conclusions. The delivery cycle for a single report is about 2 to 3 days. Humans are responsible for interpreting the results and writing conclusions understandable to business stakeholders, while AI handles the exhaustive search for counterexamples.

🛠 Setup Requirements

Requires a solid foundation in functional programming and mathematical logic. First, systematically complete the tutorials on the official Imandra documentation site, mastering OCaml or ReasonML syntax and formal modeling concepts. Then, build your own report template library and financial scenario case library. Tools required include an Imandra Cloud account, a GitHub repository for delivery, and Feishu or Tencent Docs for collaborative delivery, plus a standard service contract template. The first month requires approximately 100 hours of study to replicate official examples and financial algorithm verification cases.

🧰 Toolchain

🔧 Imandra Cloud
🔧 GitHub
🔧 OCaml or ReasonML development environment
🔧 Feishu Docs

💰 Revenue

① Fintech companies and agent development teams pay per order (primary revenue): Charging 3,000 to 6,000 RMB per explainable reasoning verification report, with complex system-level verification potentially reaching over 10,000 RMB. At a stable stage, delivering 6 to 8 reports per month yields a monthly income of approximately 20,000 to 40,000 RMB (this revenue share is not verified independently). ② Long-term audit framework contracts: Signing monthly framework contracts with 1 to 2 banks or agent vendors to lock in verification capacity; contract values and number of clients are undisclosed. ③ Licensing of counterexample libraries and formal modeling templates: Licensing counterexample lists and modeling templates to small and medium-sized agent teams; pricing and volume are undisclosed. ④ Opportunity: Annual agent compliance audit subscriptions for securities firms; annual subscription fees and market share are currently undefined.

💸 Cost

Primarily Imandra cloud computing subscriptions and personal cloud server costs, totaling approximately 500 to 1,500 RMB per month. Additional one-time expenses include study materials, community conferences, and contract templates. The business model is asset-light.

⏱ Time Investment

During the learning phase, 3 to 4 hours daily are spent on official documentation and replicating cases. During the active service phase, 5 to 6 hours daily are spent on formal modeling, counterexample testing, and report writing, with half a day per week reserved for community content creation to acquire clients.

🚀 Getting Started

The first step for beginners is to follow the introductory examples on the official Imandra documentation site, fully replicate official financial algorithm and autonomous system verification cases, and publish breakdown articles in Chinese on platforms like Juejin, Zhihu, and WeChat Official Accounts to establish a reputation as a formal verification expert. Then, distribute free sample reports in fintech and agent developer communities to secure the first paid pilot client.

🔑 Keys to Success

✅ Master the methodology of Imandra formal verification and automated reasoning rather than just calling the interface.
✅ Reports must include a list of counterexamples and reproducible mathematical proof traces for clients to value the audit.
✅ Establish professional trust by consistently outputting breakdown content in vertical technical communities before soliciting orders.
✅ Focus on a specific niche within financial trading, risk management, or compliance to build a reusable case library.
✅ Use standard contract templates to define delivery boundaries and prevent clients from treating verification reports as unlimited liability guarantees.

⚠️ 风险

⚠️ High technical barrier and steep learning curve, leading to a high dropout rate.
⚠️ Clients are concentrated in foreign financial circles and top-tier fintech companies, resulting in high acquisition and communication costs.
⚠️ Changes in platform pricing or service policies may increase cloud usage costs.
⚠️ Potential liability disputes if report conclusions are misused by clients, leading to financial losses.

📌 Real Cases

📌 Imandra's official website states that its tool suite, based on automated reasoning and formal verification, serves clients in finance and autonomous systems writing mission-critical code, ensuring algorithm safety, explainability, and fairness, with 'Reasoning as a Service' as its core positioning.
📌 In an October 2025 evaluation, opentools.ai positioned Imandra as a neuro-symbolic automated reasoning platform for finance and safety-critical systems, emphasizing its verification, compliance, and resilience capabilities.
📌 In June 2026, at the China International Finance Exhibition, Longshine Technology released a financial agent product system. The dense deployment of domestic financial agents implies that the outsourcing demand for explainability and compliance verification is expanding accordingly.

Sources