AI for Shoe Repair Shops: Tickets, Pickup Reminders and Custom Work

An AI agent for a shoe repair shop works best as a ticket-aware assistant. It reads each repair ticket, answers status questions from that record, sends pickup reminders from approved templates, and logs custom work notes for the bench. It should not set prices, promise dates the ticket does not support, or release or dispose of unclaimed items without a person deciding.

Key takeaways

• Give the agent a ticket record it can read, and let it answer status questions only from that record.
• Write ticket standards first: intake fields, status names, promised dates and who may change them.
• Send pickup reminders from approved templates, with staff approval for anything involving money, exceptions or disputes.
• Keep custom work notes (materials, stitching, heel and sole choices, customer preferences) attached to the ticket, not buried in chat history.
• Decide the unclaimed-item process on paper, with local legal advice, before any reminder runs automatically.

Why a shoe repair counter depends on the ticket

A shoe repair shop runs on its tickets. A ticket links a customer name, a phone number, a pair of shoes with its condition at intake, the work requested, the price agreed at the counter and an expected pickup date. When a customer calls to ask whether their boots are ready, the honest answer lives in that ticket, and someone has to find it, read it and reply. On a busy day that person is also at the machine, at the bench or handing a pair across the counter. Messages wait, reminders get skipped and small promises made in one line of a ticket are forgotten.

An AI agent with access to the ticket record can take the repetitive part. It can answer whether a pair is ready, send the pickup reminder when the work is marked complete and record what a customer said about a custom request. It does not replace the bench or the counter. Its usefulness depends on how clean the tickets are. An agent reading a vague ticket will give a vague answer, or a confident wrong one, and a wrong answer about pickup is the message that causes a wasted trip across town.

Ticket standards to set before any automation

Automating messages on top of inconsistent tickets multiplies the inconsistency. Before connecting anything, write a short standard that every ticket must meet. Keep it small enough that staff will actually fill it in at the counter, during a busy hour, without a manual.

• One customer record per person, with a phone number confirmed at intake and a preferred contact method.
• A status list with a few fixed states, such as received, in work, waiting on customer, ready for pickup and collected, plus a clear rule for moving between them.
• Intake condition notes: existing damage, stains, missing hardware or parts, and photographs when the shop takes them.
• A promised date field that only the counter can set, plus a reason field whenever that date changes.
• Custom work fields for materials, color, last or measurements where relevant, heel or sole choices, and anything the customer asked the shop to avoid.
• The agreed amount, whether it is final, and whether any deposit is recorded as paid or unpaid.
• One named owner for each open ticket, so the agent knows who to escalate to when it cannot answer.

Once these fields exist, the agent can answer from facts. Without them, it has to guess. Intake photographs deserve a special note: they protect the shop when a customer later says a stain was already there, and they give the agent something concrete to describe when a customer asks about the state of their shoes.

What the agent should remember, and what it should not

For a shoe repair agent, memory means the ticket history plus a small set of standing preferences: the customer's preferred channel, their language, whether they accept texts or prefer calls, and notes such as 'always check with me before replacing stitching.' Keep these inside the customer record, where staff can see and correct them.

Do not let the agent carry impressions it cannot verify, such as a guess about how much a customer is willing to spend or a promise remembered from a conversation that never reached the ticket. Guidance on context engineering for agents generally favors giving a model only the information it needs for the current task and retrieving that information from a controlled source. For a repair shop, this means the agent reads the live ticket each time it acts, rather than relying on a summary written days earlier.

Prompt injection is a real risk as well. A customer message can contain text written to change the assistant's behavior, such as a request to waive a fee or release a pair without a ticket. The OWASP prompt injection guidance is worth reading before any agent acts on text from outside the shop. Treat customer messages as data to be read, never as instructions that change prices, dates or custody of an item.

Pickup reminders: a plan you can run this month

A pickup reminder is the easiest message to automate and the easiest to get wrong. The plan below keeps the agent in the safe part of the work.

• Step one: list every status that should trigger a message and write the exact template for each. Keep the wording plain, with the shop name, the ticket reference the customer recognizes, pickup hours and a contact number.
• Step two: make the trigger a change in ticket status, not a calendar guess. A reminder should go out after the status moves to ready for pickup, never on a fixed date that may come before the work is finished.
• Step three: set quiet hours and a maximum frequency so a customer receives one reminder at a sensible interval rather than a daily message.
• Step four: list what the agent must not answer, including price disputes, refunds, warranty questions about a repair, and requests to hold or release an item. Those go to a named staff member with the ticket attached.
• Step five: route every reply that mentions a date, an amount or an exception to a human queue, and log every outgoing message on the ticket.
• Step six: review a sample of sent messages each week for the first month, and tighten templates wherever customers misread them.

Handled this way, the agent reduces the number of status calls the counter has to answer, while staff keep control of anything that touches money or the shop's obligations.

Custom work notes and approvals

Custom work is where memory matters most and where errors cost the most. A customer who asks for a specific stitch, an unusual heel height or a leather that must match an older pair expects the shop to remember. An agent can store those requests on the ticket, repeat them back before pickup and flag any change requested after the bench has started. The jewelry store approach to repairs and custom orders follows the same discipline of keeping specifications attached to the job rather than to a conversation.

The agent should not approve the change itself. If a customer writes after work has begun to ask for a different heel, the agent records the request, tells the customer that a staff member will confirm, and places the question in the owner's queue beside the original specification and photographs.

Illustrative scenario: a customer asks for a second pair of riding boots to match a first pair, and asks that the original sole profile be kept. The agent pulls the earlier ticket, attaches the sole note and the previous photographs to the new ticket, and asks the customer to confirm the leather color by photo. The bench lead confirms the match before any work starts. The agent does not estimate the cost of the change, and the counter gives the final figure.

Unclaimed items: write the rule before the reminder

Every shop with a pickup counter eventually holds items nobody collects. Automating reminders without a written unclaimed-item process creates two risks. The agent may send messages that imply a deadline the shop has not actually set, and staff may discover too late that an item was disposed of while a customer was still expecting it.

Write the rule with a local legal reference or adviser, because notice requirements and holding periods vary by place. Then make the agent follow it:

• Define when an item becomes unclaimed, measured from the ready date recorded on the ticket.
• Define the reminder sequence and its wording, approved by the owner.
• Define what the shop may do with an unclaimed item, and who must approve each option.
• Require human sign-off before any item is marked as disposed, sold or donated.
• Keep a log of every notice sent and every decision made, attached to the ticket.

The agent can prepare the list of items approaching the end of their holding period and draft the notices. It should not make the final call. Its job is to make sure the decision is made on time, with the full record in front of the person who makes it. Customer records also contain personal data, so keep only what the workflow needs and follow the data protection guidance that applies where you operate, such as the material published by the ANPD in Brazil.

How Be Human handles this, and where to begin

Be Human designs and operates this kind of agent as a managed service. A first engagement defines one useful role for the shop, usually a counter assistant for status and pickup messages. It also defines the ticket memory the role needs, the tools it may use, the approval boundaries and a success criterion the shop can observe, such as the share of status questions answered from the record without staff time. Ongoing operations can include monitoring, periodic evaluations of the messages sent, integration maintenance, permission reviews and controlled changes to templates, so the shop's standards keep up with how the work actually runs. Service is remote, in English or Portuguese.

The practical starting point is small: choose the message that costs the counter the most time, write the ticket standard it depends on, and set the approval rule for anything involving money or disposal.

If you want help choosing that first task, the free Be Human AI diagnostic is built for exactly that decision.

Frequently asked questions

Can an AI agent answer repair status questions for a shoe repair shop? Yes, provided it reads the live ticket each time and answers only from the fields the shop has standardized, such as status and ready date. If the ticket is missing a status or a date, the agent should say a staff member will confirm rather than estimate. Status answers are among the safest tasks to automate because they require reading facts, not making decisions.

What should a shoe repair shop automate first? Start with pickup reminders triggered by a change to the ready for pickup status, since the rule is simple and the time saved at the counter is easy to notice. Write the ticket standard and the list of forbidden replies before you connect any messaging channel. Expand to custom work notes once the reminders run cleanly for several weeks.

Who decides when an unclaimed item can be disposed of? The shop owner or a named manager decides, after checking the written unclaimed-item rule and any local requirements. The agent can identify which items have reached the end of their holding period and draft the notices, but it must not mark anything as disposed, sold or donated. Every decision should be logged on the ticket with the person's name and the date.

Does the agent need to store customer photos and personal details? It needs only what the workflow requires, which usually means name, a confirmed contact method, the ticket history and intake photographs of the shoes. Keep sensitive or unnecessary personal information out of the agent's memory, restrict who can view and edit it, and review those permissions periodically. Follow the data protection rules that apply in your jurisdiction, and document what the agent stores and why.

Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.