Four service types we provide
Following the Ministry of Industry and Information Technology's definition of an AI application service provider, enterprise AI services fall into four categories: consulting and planning, delivery and implementation, operations management, and security governance. We cover all four — not just selling software, but taking responsibility from "should we do this, and where do we start" all the way to "how do we keep it working".
1. Consulting and planning: decide what to do first
Many companies stall at step one: they know they need AI but not which process to start with. We do three things:
- Scenario diagnosis: break the business process into concrete steps, mark repeatability, structure and data availability, and produce a scenario list
- Prioritisation: rank by business value times feasibility and recommend one or two starting scenarios
- Plan and acceptance criteria: define the technical route, deployment model and measurable, traceable acceptance metrics
Duration: usually one to two weeks. Output: scenario diagnosis report, implementation plan, acceptance criteria.
2. Delivery and implementation: make it work inside the business
Software
- AI application delivery: agents and RAG, workflow automation
- Data and system integration: ERP, WMS, MES, OA, WeCom/DingTalk
- Private deployment: models and data run inside the customer network
Hardware
- Robot workstations: sorting, loading, inspection, quality control
- Edge compute and visual inference
- Line integration and acceptance: PLC, MES, safety circuits
Delivery standard: every scenario must pass a quality gate before go-live — agreed test set, fixed metric definitions, reproducible results. No gate, no launch.
3. Operations management: go-live is only the beginning
The most common outcome is that a system is abandoned two weeks after launch. We treat operations as part of the service:
- Adoption coaching: regular follow-ups to remove real usage barriers
- Issue handling: 30-minute response for critical issues, 2 hours for major ones
- Iteration: update knowledge bases, prompts and configuration based on usage data
- Value review: quarterly or annual value reports with metric changes and next steps
4. Security governance: data stays inside the company
For customer lists, cost structures and financial data, security is a precondition, not an add-on:
- Private deployment: data never leaves the customer network
- Least privilege: role-based accounts, no shared logins
- Audit trails: data export, configuration and model changes logged for at least six months
- Post-delivery cleanup: temporary data, accounts and environments removed with written records
Delivery method: six steps
Diagnose → Plan → Contract → Pilot → Scale → Annual operation
Each step has defined outputs and exit criteria.
| Stage | Output | Exit criteria |
|---|---|---|
| Diagnose | Scenario diagnosis report | Customer confirms scenarios and priority |
| Plan | Implementation plan | Customer confirms plan and acceptance metrics |
| Contract | Contract, project charter | Scope, roles, fees and schedule agreed |
| Pilot | Working system, evaluation report | Agreed metrics met |
| Scale | Rollout plan, implementation package | Pilot results accepted by customer |
| Operate | Operations log, value report | Annual review of renewal |
What we don't do
- No general model training — we handle application-level selection, fine-tuning, prompt engineering and evaluation
- No compute or network infrastructure — we plan and adapt the customer-side environment, not build it
- No engineering construction — civil works, cabling and mechanical installation are handled by ecosystem partners
- We do not promise unverifiable results or use vague claims like "significant improvement"
Fee structure
Four fee types, quoted per project:
- Plan fee: per stage, by project type and depth
- Talent fee: a percentage of annual salary, agreed per case
- On-site fee: per month or person-day
- Annual fee: continuous service, coordination and planning
Why we can deliver
Confidence comes from our in-house foundation at Hangzhou Shenggong Intelligent Robotics (embodied-AI R&D centre): robot platforms and motion control, arm foundation model, industrial drones and AI product definition; FDE (forward-deployed engineers) bring it to customer sites.
We built the tools, we validated the plans, we can fix what breaks.