AI for Locksmith Businesses: Lockout Calls and Key Cut Records
An AI agent can take lockout calls and keep key cut records, but identity checks and key release stay with a technician. Here is what to log.

An AI agent for a locksmith business can answer lockout calls at any hour, collect the facts a dispatcher needs, and keep clean records of key cuts and service visits. It should not decide whether a caller may enter a property, receive a key, or hear a key code read back. Identity checks and every release of access stay with a trained technician or a named human approver.
Key takeaways
- A lockout agent gathers location, service type, urgency and callback details, then hands the job to dispatch.
- Every lockout and key cut should leave a record with a timestamp, a job reference, the technician and the approval used.
- Key codes and cut specifications are sensitive. Store them in a restricted area and never read them back to an unverified caller.
- Identity verification, proof of authority to open a property or vehicle, and the decision to release a key remain with a technician.
- Start with one call type and one log format, and widen the agent's role only after the logs show the handoffs work.
What a lockout call agent should handle
Lockout calls arrive at inconvenient hours from people who are stressed and often standing outside a door or beside a car. The useful job for an agent is intake: confirm the service type, capture the location, sort the request by urgency, and tell the caller clearly what happens next. The agent can also ask a fixed set of safety questions, such as whether anyone is inside, whether a child or an elderly person is involved, or whether there is an injury. Any yes answer sends the call to a human at once, with the transcript attached.
The agent should not commit to a final cost for work it has not seen, promise an arrival time the dispatch board cannot support, or advise a caller on forcing a lock. Those boundaries belong in the agent's written instructions, not only in someone's memory. A caller may also try to persuade the agent that they are the owner or that an exception applies. Prompt injection, where text in a conversation tries to override the agent's rules, is a known risk, so the agent's permissions should not depend on what a caller says.
The lockout call log, field by field
A lockout record is only useful if the same fields are filled in every time. The following structure works for most residential, commercial and automotive calls:
- Call time, channel and callback number, stored as provided
- Service type: residential, commercial, automotive or safe
- Location as a street address or landmark, plus any access notes the caller gives
- Urgency flag and any safety concern that triggered a human handoff
- The caller's stated relationship to the property or vehicle, recorded as a claim
- Dispatch decision: technician assigned, estimated window or callback scheduled
- Outcome: opened, not opened, referred or cancelled, with the technician's note
The phrase that matters most is recorded as a claim. When a caller says they own the car, the log should say exactly that, and it should not mark the statement as verified. Six months later, the difference between a claim and a checked fact is often the difference between a clean review and an argument.
Key cut records: what to store and who can see them
Key cut records help with repeat customers, fleet and property accounts, and questions about who received which key. They are also the most sensitive part of the business, because a cut specification can be used to make a copy of a key. Each field needs its own access rule.
- Job reference and date, which office staff can see
- Customer account and property or vehicle identifier, limited to dispatch and accounts
- Key blank type and cut specification, limited to technicians who cut keys
- Any key code or bitting data, held in a restricted store and never returned in chat or voice replies
- Release record: who collected the key, how their identity was checked, and who approved the release
The agent can write the record, attach the photo or note the technician uploads, and remind the technician to complete the release section. Its role should not include reading key codes back, even to other staff. Access should follow least privilege, the principle that each role sees only what its job requires, as described in OWASP's authorization guidance. If a role does not need the cut specification to do its work, it should not have it.
Which identity checks stay with the technician
This is the boundary that matters most. Deciding whether the person at the door has the right to open a home, a car or a safe depends on documents, context and judgment that an agent cannot verify from a phone conversation. Requirements for proof of ownership or authority differ by location and by property type, so the business should set its policy with its own advisers or licensing body. The agent should follow that policy rather than improvise one.
These decisions stay with a human:
- Checking identification and proof of ownership or authority at the site
- Deciding whether to open a property, vehicle or safe for a given caller
- Releasing any cut key or physical key to a person
- Approving any exception, such as opening for a third party or after hours without documents
- Any situation involving a suspected break-in, a shared property or a dispute between household members
The agent can prepare the checklist the technician uses on site, so the check is faster and consistent from job to job. It should never be the step that writes verified unless a technician has entered that result. This follows the broader principle in the NIST AI Risk Management Framework, which treats human oversight and clear accountability as part of managing AI systems, not as an afterthought.
A handoff plan you can review before launch
Before an agent takes live calls, the business should be able to read the plan end to end. A workable version has six steps:
- Step 1: The agent answers, confirms the service type and asks the safety questions.
- Step 2: Any safety concern transfers the call to a human immediately, with the transcript.
- Step 3: The agent creates the log entry and marks every caller statement as a claim.
- Step 4: Dispatch assigns a technician and sends the on-site verification checklist.
- Step 5: The technician records the verification result, the outcome and any release decision.
- Step 6: The agent tells the caller the window it is allowed to give and schedules a follow-up if the job stays open.
If a step has no owner, the plan is not ready. Each handoff should name the person or role that receives it and what they must do within the working day.
An illustrative scenario
Illustrative scenario, not a client case: A caller says they are locked out of a rental van and that the rental company has sent them. The agent records the service type as automotive, marks the renter status as a claim, and notes that the van's rental paperwork has not been seen. Dispatch sends a technician with the verification checklist, which asks for the rental agreement and a call to the rental company's authorization line. The technician declines to open the van until the authorization arrives. The agent tells the caller the status, lists the documents to have ready, and books the follow-up. The log shows who decided, when, and why, which is the record the business needs if the job is questioned later.
How Be Human runs this in practice
Be Human's managed approach starts with one defined role, such as lockout intake for a single service area, and specifies the memory and tools that role needs: the dispatch board, the job log, and a read-only view of the customer account. Approval boundaries are written down before launch, including who approves any release and which situations go straight to a human. The success criterion is observable, for example whether every lockout call has a complete log and a recorded handoff decision. After launch, monitoring, recurring evaluations, permission reviews and incident response keep the agent inside those boundaries, and any change to its scope is made under human control. For the dispatch side of this work, see Superintelligent Agents for Locksmiths: Dispatch, Key Codes and Calls. Businesses that also cut keys at a counter may find AI for Hardware Stores: Special Orders, Key Cuts and Pickups a useful comparison.
Where to start with a locksmith agent
Choose the call type with the clearest rules and the lowest risk, usually intake and logging for residential lockouts in one service area. Write the log fields, the safety questions and the handoff owners before any configuration begins. Review the logs weekly for the first month, looking for missing fields, claims recorded as facts and handoffs that stalled. Expand to key cut records only after access rules for codes and specifications are tested and written down. If you are not sure which task to hand to an agent first, take the free Be Human AI diagnostic to find the first task worth handing over.
Frequently asked questions
Can an AI agent unlock a car or open a door for a caller? No, and it should not be designed to. The agent can gather the details, schedule dispatch and explain next steps, but the physical opening and the decision to open depend on verification done by a technician on site. Where the business's policy permits, the agent can prepare the checklist the technician needs.
Should the agent read key codes back to customers who call about a copy? No. Key codes and cut specifications should be held in a restricted store with access limited to technicians who cut keys. A caller asking for a code should be routed to a human approver, and the attempt should be logged as a request. Restricting read-back is one of the simplest and most important permission settings in the system.
What should go into a lockout log if the caller cannot prove ownership? The log should record what the caller said, marked clearly as a claim, along with the service type, location, urgency and any safety concern. It should also record that verification was not complete and what the caller was told about the next step. A log that separates claims from verified facts is far easier to review later.
How do we know the agent is working safely after launch? Check the logs on a regular schedule for missing fields, claims recorded as facts, and handoffs that were not completed. Review the agent's permissions whenever the dispatch tools or staff roles change, and treat any incident as a reason to tighten the boundary before widening it again. Be Human's managed operations include these reviews as an ongoing part of the service.
Your next step
See where an AI agent fits in your company first.
Take the free Be Human AI diagnostic. A few questions about your work show which task to hand to an agent first, what it needs to know and where a person approves.
Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.


