Skip to main content

嘉谏人工智能科技

JJ AI TECH

Turn field work into traceable delivery

Practical workflow software for maritime service, industrial maintenance, facilities and asset-intensive operations.

We design software around the work that actually happens on site: the task, the equipment, the photo, the measurement, the exception, the review, the signature and the report that has to reach the customer.

What we do

We turn a process into a record that holds up later

JJ AI TECH designs practical, human-reviewed workflow software for field service, maintenance, facilities and asset-intensive operations.

Most of the operations we work with already have competent people and correct procedures. What is missing is a structure that keeps the evidence attached to the right object — this vessel, this unit, this door, this zone, this inspection point — so that progress can be checked, a reviewer can approve, and the record is still readable a year later.

We are not a general AI outsourcing shop. We work on a narrow, repeatable pattern: organise field execution into a traceable delivery chain.

  • Real objects — vessel, equipment, door, zone, asset, work order
  • Real roles — engineer, supervisor, reviewer, approver, customer
  • Real evidence — photo, measurement, test value, form, signature
  • Real output — a report someone is willing to sign and send

The workflow

One chain, eight steps

Every industry below is a variation of the same chain. That is why the same team can move between a heat exchanger overhaul, a pipeline service visit and a cleaning round without starting from zero.

  1. 01Task
  2. 02Capture
  3. 03Verify
  4. 04Approve
  5. 05Sign
  6. 06Report
  7. 07Deliver
  8. 08Follow Through

1 Task — created or assigned against a real object.

2 Capture — photos, measurements, forms, signatures, barcodes and site notes taken where the work happens.

3 Verify — missing items, contradictions and out-of-range values are surfaced.

4 Approve — a supervisor or specialist reviews and accepts or returns it.

5 Sign — the customer confirms or signs.

6 Report — a branded report is generated from the same record, not retyped.

7 Deliver — the report is sent or handed over.

8 Follow Through — spare parts, quotations, rectification, cost and the next task continue from the same record.

Where AI sits. AI is used to organise, cross-check, draft and expose what is missing. Engineering, quality, safety, commercial, inventory, pricing and educational decisions are made by authorised people. Every AI output keeps a human review step and a stop condition.

Priority industries

Where the pattern is strongest

Four operating contexts where we have the clearest evidence and the most specific design experience.

Maritime Equipment Service

Operating context

Service engineers work on board, in yards, in workshops and back on board again, often across several vessels and several units in parallel.

Common friction

Photos, measurements, pressure tests and open problems end up in phone galleries and group chats, so a project manager cannot tell which unit or which node is still unfinished.

Workflow solution

Project → vessel → equipment → standard node → photo, measurement and test record → problem closure → report and archive.

Human decision boundary

Engineers, review staff, class surveyors and safety-critical control systems keep their decisions. The software organises the record.

Evidence status

Delivered — internal record (PCIM, CPIM, AP marine door)

Industrial Field Service & Maintenance

Operating context

Mobile engineers attend equipment on customer sites, then have to produce a report the customer will accept.

Common friction

Paperwork is written after hours, signatures are collected on loose sheets, and reports reach the customer days later.

Workflow solution

Capture the site condition, the work done and the customer signature on the phone, then generate and send the service report from the same record.

Human decision boundary

Engineering judgement, safety sign-off and commercial follow-up stay with authorised people.

Evidence status

Delivered — internal record (UK pipeline service company)

Facilities, Property & Multi-site Operations

Operating context

Cleaning, security and inspection rounds run across many zones, shifts and buildings.

Common friction

Managers cannot verify that a round happened; evidence is scattered across chat groups and nothing accumulates over time.

Workflow solution

Zone-based task capture, central review, and a reviewed daily service report that can be shared with the client or residents.

Human decision boundary

Managers decide what is acceptable, what is escalated and what is published.

Evidence status

Delivered — internal record (property / security / cleaning inspection platform)

Warehouse, Intralogistics & Quotation Handover

Operating context

Field work discovers a spare part need; the warehouse, the approver and the quotation sit in different systems.

Common friction

The handover between “we found a problem” and “here is a priced, approved quotation” is manual and slow.

Workflow solution

Stock and location records, inbound/outbound and picking exceptions, and a controlled path from field finding to approval and quotation.

Human decision boundary

Pricing, stock write-offs and commercial terms are set by authorised staff.

Evidence status

Contributed (WMS, warehouse and quotation management work)

Best-evidenced industry

Maritime equipment service

Three separate products, three different equipment families, one method: the project is the container, the equipment is the managed object, the standard node or check item is the unit of work, and photos, measurements, tests and problems are the evidence.

PCIM — plate heat exchanger service

Multi-stage service across on-board inspection, removal, logistics, workshop unpacking, cleaning, re-plating and pressure testing, then reinstallation on board.

  • Project → heat exchanger → ten standard nodes
  • Point-by-point photos, A-values and plate data
  • Multiple pressure test rounds kept in sequence
  • Node data carried into the project report

CPIM — cathodic protection and auxiliary systems

ICCP, AEN and MGPS, where every vessel has a different component configuration and pre-inspection, spares, service and testing easily lose their link to each other.

  • System and component records per vessel
  • Pre-inspection separated from actual replacement
  • Spare parts, service and re-test values in one chain
  • Inspection, service and sign-off reports

AP Marine Door — installation and commissioning

Many doors with similar numbering, template-driven checks, and Not OK findings that have to be rectified and re-verified before hand-over.

  • Installation check and function check as two stages
  • Not OK problems with owner, status and evidence
  • Rectification, re-verification or return
  • Open problems block the final report

Boundary. These are three vertical products that share a method. A single sign-on, single-database, unified ship-owner maintenance portal is a direction we want to build — it is described on this site as future direction, not as an existing platform.

Cross-industry

The same chain, different objects

Industrial equipment after-sales

Installed base, service visit, findings, signature, report, spare part follow-up.

Pipeline and utility field service

Site photos, service form, customer signature, report sent from the field.

Property, security, cleaning and inspection rounds

Zone task, evidence capture, supervisor review, reviewed daily report.

Warehouse, spare parts and quotation

Exception found, part identified, approval, quotation, traceable handover.

Project-based learning

Teacher-defined goal, student group work, process evidence, teacher-led assessment.

Evidence

How we label what we claim

We publish the status of each piece of work instead of implying that everything is a finished, referenced product. Client names, logos, vessel names, screenshots and outcome figures are only published with written permission.

Delivered — internal record

Built and used in a real project. Client name, screenshots and outcome figures are not published because written permission has not been given.

Prototype

Working software or a working flow that has not completed acceptance in a target production environment.

Contributed

JJ AI TECH contributed part of the work inside a wider project owned by another party.

Concept — validation required

A designed direction based on observed operational patterns. Not yet built or proven.

How we work

Map, Define, Prototype, Validate, Scale

1. Map

Walk the real process with the people who do it. Identify the objects, roles, decision points and the evidence that already exists.

2. Define

Agree the workflow, the record structure, the review roles and the report that has to come out of it.

3. Prototype

Build the smallest version that a real team can run on real work.

4. Validate

Run it against real cases. Check what breaks, what is missing and what a reviewer actually needs to see.

5. Scale

Extend to more sites, more equipment types and more roles once the workflow holds.

We do not publish fixed delivery dates or fixed prices on this site. Scope, timeline and commercial terms are agreed per engagement.

Start from one workflow, not a platform

Tell us the process, the role that owns it and what currently goes wrong. If it is not a fit, we will say so.

沪ICP备2026001288号-1
工业和信息化部备案管理系统