From a working prototype
to a measured customer pilot.
EquipBridge AI helps equipment manufacturers and overseas distributors keep equipment details, bilingual knowledge and service follow-up in one workflow. This guide separates the current website from the proposed commercial product.
A public demo and individual saved workspaces. Drafts use reviewed bilingual excerpts; an external generative language model is not connected. Example equipment and service cases are fictional.
Download the complete project handbook (PDF)
A three-minute demonstration
- 0:00–0:30 · Open the queue. Enter the public demo, select “Part number confirmation” and point out the model, serial number and manual revision.
- 0:30–1:10 · Check the evidence. Select “Draft from sources”. Open a source to see its bilingual excerpt, equipment model, revision and page.
- 1:10–1:50 · Record the review. Read both response languages. Select “Review”, confirm the checks and approve. This records an internal decision; no external message is sent.
- 1:50–2:15 · Close the loop. Select “Resolve ticket”, enter a fictional confirmed outcome and show the activity history.
- 2:15–2:50 · Show an evidence gap. In the source assistant, change the revision to “UNMATCHED”. Ask which manual applies. Show that no matching source is returned.
- 2:50–3:00 · Explain the next step. A manufacturer-approved corpus, domain-engineer evaluation and a bounded customer pilot come before commercial performance claims.
How to use a saved workspace
- Open “My workspace” and sign in. Records are scoped to your account.
- Add reviewed English and Chinese excerpts with a model, manual revision and page. Optional original files may be PDF or TXT up to 5 MB.
- Create a ticket, confirm equipment details, retrieve a draft and complete the review.
- Export ticket and knowledge metadata from “Export my records”. Original source files remain available from their individual source cards.
A 90-day delivery roadmap
This is a planning schedule, not a record of completed work. Day 90 is a checkpoint within the existing 16-week validation programme; longer pilot observation continues beyond it.
| Window | Deliverable | Lead | Evidence to retain |
|---|---|---|---|
| Days 1–15 | Interview qualified exporters; select one equipment family and a domain reviewer. | Project lead | Interview notes, workflow baseline, candidate pilot scope. |
| Days 16–35 | Define shared roles, document applicability and reviewed import requirements. | Product lead / security lead | Role matrix, permission tests, document inventory. |
| Days 36–55 | Connect and evaluate a model with approved data and terminology. | AI lead | Held-out bilingual evaluation, source checks, error records. |
| Days 56–70 | Finish agreed pilot controls and run end-to-end acceptance. | Product lead / security lead | Import, export, access and recovery test results. |
| Days 71–90 | Begin a bounded pilot, review usage and revise the offer. | Project lead / delivery team | Eligible ticket counts, user feedback, early cost and value evidence. |
What a pilot must test
- Source applicability: report whether answers use the correct model and manual revision; do not infer applicability from a filename.
- Bilingual fidelity: preserve part numbers, fault codes, units and meaning in both languages.
- Engineering accountability: technical replies require the agreed reviewer; missing evidence triggers clarification.
- Workflow value: compare active handling minutes, first meaningful response time and repeat contacts against an agreed baseline.
- Willingness to pay: record actual buyer responses to a scoped commercial offer. Proposed prices and financial projections remain assumptions until tested.
- Delivery effort: record document preparation, onboarding, support time and model usage before expanding pilot volume.
Minimum readiness gates
Before customer data or live technical advice enters a pilot, agree the participants, approved source corpus, allowed data, engineering reviewer, access roles, hosting/model recipients and recovery process. Evaluate routine questions alongside missing identifiers, mismatched revisions and unsupported requests. Investigate any critical technical error before affected use resumes.
Metric definitions
| Metric | Definition |
|---|---|
| Source applicability rate | Reviewed answers using applicable model/revision evidence ÷ answers evaluated. Report both counts. |
| Draft acceptance | Drafts accepted after review ÷ drafts reviewed. Separate minor edits from technical corrections. |
| Active handling time | Engineer working minutes per eligible ticket. Keep waiting time separate. |
| First meaningful response | Elapsed time until a useful, reviewed response; automated acknowledgements do not count. |
| Repeat contact rate | Eligible closed tickets followed by the same issue within seven days ÷ eligible closed tickets. |
| Onboarding effort | Document preparation, configuration and training hours per manufacturer. |
Start the conversation
Contact the project team at caobuo218315@126.com, +86 176 8475 1441 or +61 401 254 388. Prepare one equipment family, a representative service workflow and the documents you are authorised to share.
Try the public demo