“The zip is broken” does not yet define a repair
A handbag arrives with a note: “Replace the zip.” A later message says: “Only the piece I pull has come off; do not replace the entire zip.” Before quoting, clarify what the person is pointing to and which work they are requesting. AI could be trialled to help draft those questions, leaving component identification and the repair decision to the workshop.
This guide proposes a fictional text exercise: retain the original words, separate the component mentioned from a diagnosis and prepare a worksheet for review. It includes a hand-written reference for comparison. The pilot has not been run and no completed repair is presented.
Working vocabulary without turning a phrase into a diagnosis
YKK describes three main zip components: the interlocking elements or teeth, the moving slider and the fabric tape. Its Italian catalogue also distinguishes Slider and Puller. In this exercise, “slider” means the sliding body and “pull tab” means the part held by hand. Clarify this working vocabulary with the person bringing the bag; it does not establish which component has failed.
Do not automatically translate “the little piece”, “the fastening” or “it will not move” into a spare-part reference. Preserve the phrase and ask which part and opening it refers to. A bag may have a main opening and pockets: identify the stated area before grouping messages.
For real-world context, Valentina Rivera Madrid describes work on pull tabs, sliders and guides, and requests a description and a photo or video of the area. El Reparador de Zurbano lists work on zips, handles and linings. These describe their services; they do not demonstrate AI diagnosis or the effectiveness of this exercise.
Prepare a worksheet you can check against the note
I propose recording: bag identifier, source, stated area, exact wording, component explicitly mentioned, request, exclusion, outstanding question and workshop decision. This is a worksheet design for the exercise, not a required trade format or an existing software export.
Use only the fictional data below. Do not paste real names, phone numbers, conversations, orders or photographs without permission and suitable confidentiality and processing conditions. The cover is an AI-generated illustration depicting no real client or commission.
Fictional BOLSO-PRUEBA case: two openings and a changing request
Every document, object and message in this case is invented by the editor. There are no photographs of the bag in the case. D0 · Workshop instruction: “Prepare questions before quoting. Do not select parts, diagnose, repair or confirm price or delivery.”
D1 · Intake note, bag B1: “Main zip: replace zip. Inside pocket: the slider will not move.” D2 · Later clarification from the person, B1: “On the main opening, only the piece I pull has come off. Do not replace the entire zip. For the inside pocket, I only want a quote for now.”
D3 · Workshop note, B1: “Awaiting inspection; the main-opening component has not been identified and no replacement part has been selected.” D2 changes the request for the main opening, without turning the description into technical identification. D3 does not confirm a fault in the pocket either.
Manual reference: what to retain and what to ask
The table is the editor’s proposed reference for checking an answer, not AI output or a diagnosis. Questions are internal drafts and have not been sent.
| Area | Wording and scope retained | Question and working status |
|---|---|---|
| Main opening · D1, D2 and D3 | D1 requests zip replacement. D2 specifies “the piece I pull” and excludes the whole zip. D3 does not identify a component. | Which piece is being indicated? Inspect in person and clarify scope with the person; original request modified, no part or repair confirmed. |
| Inside pocket · D1 and D2 | “The slider will not move” is D1’s account. D2 requests only a quote for now. | What does the workshop observe on inspection, and what will it propose for quoting? Cause and solution pending; no work authorised. |
A trial prompt and a worthwhile check
Copy D0–D3 in full after this instruction and save the response separately from the original notes:
“Work only with BOLSO-PRUEBA. Prepare one row per stated area of B1. Include sources, exact quotations, components explicitly mentioned by each source, current request, exclusions, questions and pending status. Retain D1’s original request alongside D2’s clarification without executing it. Do not turn ‘the piece I pull’ into an identified pull tab or ‘the slider will not move’ into a diagnosed cause. Do not choose parts or invent authorisation, price, stock or delivery. Distinguish the person’s account from workshop observation. State what information is missing.”
Claude’s documentation recommends letting the model acknowledge uncertainty and asking for citations to check claims. These inform the exercise without guaranteeing compliance. No particular tool or ability to identify components in images has been tested here.
Check that both areas appear and each field points to D1, D2 or D3. Reject a worksheet stating “broken pull tab confirmed”, ordering a whole-zip replacement or turning “only a quote” into an accepted repair. Also check it has not erased the original request: retaining how the commission changed is useful.
Then try a separate variant: D4 · “The workshop identifies a missing pull tab on the main opening. Quote not yet issued; no repair decided or authorised.” The new worksheet should add that identification attributed to D4, retain D2’s account and infer no acceptance. D4 resolves neither the inside pocket nor replacement-part compatibility.
Measure the whole task before connecting anything
Prepare several fictional cases with a manual reference written first: an unidentified part, two openings, an exclusion, a changed request and a later workshop observation. Record omitted areas, altered quotes, invented diagnoses, expanded scope and questions that recur after a new source resolves them.
Compare source preparation, checking and correction time with writing the same worksheet by hand. Define invalidating errors beforehand; invented authorisation would be one. Keep failed cases too. These are proposed measures, with no demonstrated savings, accuracy or productivity results.
The professional inspects the item, identifies components, assesses options and agrees the quote and scope with the person. The worksheet organises that conversation; it certifies neither condition, authenticity, material suitability nor repair quality.
From a manual trial to a scoped integration
The manual trial means copying fictional text into an assistant and checking its proposal. Connecting automation requires investigating actual channels, permissions, identifiers for each bag and area, retention of original notes and a record of who changes scope. A new note should trigger pending review, not an automatic work order. Each connection and its error recovery would need validation.
Before developing, check whether your current system can already record areas, attach notes and distinguish a quote from approved work. Adaris develops custom software, CRM systems and AI automation case by case. A query inbox with these controls and agreed acceptance criteria could be assessed. This is a development proposal, not an existing proprietary leather-repair application or verified compatibility with your software.
Commercial support is in Spanish. To discuss scope, explain what information you receive, who reviews it and where the distinction between enquiry, quote and repair gets lost.
Related services
Sources
Does the request change before reaching the workbench?
Tell us how you receive notes and who approves scope. We can assess a scoped integration for your workshop. Commercial support is in Spanish.