← Sticker Wall AGENT · DETAIL

H Company's Banking and Insurance Workflow Automation Agent Generates Over €1M Monthly

Workflow: The agent receives task queues from banking or insurance back-office software daily, parses screenshots via a computer u

AGENT

Key Fields

FIELD STAMPS

IndustryFintech

RegionEurope（法国）

ScaleSME

ChannelOnline

🔧 Workflow

The agent receives task queues from banking or insurance back-office software daily, parses screenshots via a computer use model, identifies form fields, button positions, and data tables, and automatically executes clicks, inputs, and cross-system copy-verification operations. Upon completing each task, the agent logs operations, generates a structured result summary, and flags low-confidence steps for human reviewers to spot-check. Once confirmed by reviewers, the results are written back to the client's system, and the team iterates on prompts and screen operation rules based on error logs.

🛠 Setup Requirements

Requires enterprise-grade API integration capabilities, an understanding of banking or insurance software UI workflows, and a cloud server or local inference instance. The team must first obtain test accounts for the client's back-office software, map out at least three automatable repetitive processes, integrate the H Company API, and configure screen parsing modules, operation logs, and anomaly alerts. The setup cycle takes about a month and has a high technical barrier, making it suitable for teams or individual developers with enterprise software deployment experience. Throughout the process, data boundaries and permission scopes must be repeatedly confirmed with the client's IT department.

🧰 Toolchain

🔧 H Company API
🔧 Enterprise internal legacy software
🔧 Cloud servers
🔧 Operation log and alarm system
🔧 Computer vision screen parsing tool

💰 Revenue

1) Process outsourcing: Banking and insurance clients pay an automated outsourcing fee per task processed or on a monthly basis; 2) Project-based delivery: Deployment and system integration fees for computer use charged per process line; 3) Monthly maintenance: Monthly exception handling and rule maintenance fees charged per client; 4) Peer replication: Replicating and promoting proven solutions to more financial institutions (an opportunity item, with no figures yet on potential earnings). At the corporate level, H Company itself charges based on API call volume, while the replicators charge both ends via the aforementioned outsourcing and delivery models, generating monthly revenues around €150,000.

💸 Cost

Main costs are H Company API call fees and cloud server rentals, estimated at approximately €10,000 to €20,000 per month. APIs are billed per token, with a single complex financial process computer use call costing around €0.50 to €2.00. When batch-processing 10,000 tasks monthly, API costs account for about 10% to 15% of revenue, with an additional several thousand euros for cloud resources and O&M expenses.

⏱ Time Investment

2 hours per day for maintenance, exception handling, and review spot-checks

🚀 Getting Started

First, contact local European insurance brokerages or small banks to offer a two-week free trial, select three back-office processes with the highest repetitiveness, and run underwriting or claims pre-auditing tasks using the H Company API. Once successful, charge based on processing volume or project delivery while maintaining human review steps to build client trust. Beginners can first use public computer use demo environments to familiarize themselves with screen operations and prompt design before pitching to clients with real-world case studies.

🔑 Keys to Success

✅ Bind to highly repetitive financial processes
✅ Cover non-API legacy software using computer use
✅ Charge based on successfully processed volume
✅ Retain human reviewers as the final judge
✅ Stay synchronized with NVIDIA's open-source ecosystem
✅ Rapidly replicate to similar clients after workflow standardization

⚠️ 风险

⚠️ Frequent UI updates in financial software cause agent operations to fail
⚠️ Data compliance audits in banking and insurance may delay deployment
⚠️ Error rates of computer use models in complex form scenarios still require human fallbacks
⚠️ Client internal IT departments may resist external agents connecting to core back-office systems

📌 Real Cases

📌 A French insurance brokerage uses H Company agents to automatically process claims pre-audits, saving about €50,000 in labor costs per month
📌 H Company open-sourced its computer use model API in June 2026, opening it to enterprise product teams and developers with the goal of enabling agents to execute complete workflows on screens rather than just conversational Q&A
📌 H Company joined the international open-source AI alliance led by NVIDIA, promoting deployment cases of its agent models in banking and insurance vertical scenarios

Sources