Skip to content
be-human.ai
← Back to the journal

AI by industry

AI for a Crane Rental Company: Lift Scheduling and Inspection Logs

An AI agent can hold crane lift schedules, operator certification dates and inspection logs, then flag gaps for safety staff to review before a lift.

Be Human typographic cover with the title: AI for a Crane Rental Company: Lift Scheduling and Inspection Logs
Conceptual illustration created with AI · Be Human

An AI agent for a crane rental company keeps the connected records of the fleet in one place: which crane is booked for which job, which operator is certified on which crane class, and when each inspection and certificate expires. It does not decide whether a lift is safe. It checks each booking against those records, flags gaps to the people accountable for safety, and leaves every approval with them.

Key takeaways

  • Start with lift scheduling conflicts and certificate gaps, because those are the records that stop a job when they are wrong.
  • Keep the first agent read-heavy: it reads schedules, logs and expiry dates, and it drafts flags instead of changing records.
  • Give every flag a named owner, a deadline and a resolution state so that nothing waits unseen in a shared inbox.
  • Write down who can override a flag and who must approve a schedule change before the agent is connected to anything.
  • Measure success by how early a gap reaches a named person before a lift, not by how many messages the agent produces.

Why do lift schedules break at the joints between records?

A crane rental operation depends on several records that rarely live in the same place: reservations, crane configuration, site requirements, operator assignments, transport, inspection logs and certificates. Dispatch may work from a board or a scheduling tool, safety may keep inspection records in folders of PDFs, and HR may track certifications in a spreadsheet. Each record can be correct on its own and the operation can still fail at the joints. A booking is moved to a different crane whose latest inspection log is not attached to the job. An operator is certified for one crane class but assigned to a neighboring model with different capacity charts. A periodic inspection falls due in the middle of a long booking and nobody connects the two dates.

These gaps are not caused by carelessness. They appear because no one person is responsible for checking the whole chain before each lift, and because the checks that do exist happen at different times. An agent is useful here because it can run the same check on every booking, at the same point in the workflow, and report what it relied on.

What should the agent hold in memory?

The agent needs a defined set of sources, and it should read them from systems of record rather than from notes that staff have typed into a chat. The minimum useful set is usually four:

  • The fleet register: each crane, its configuration, the capacity charts on file, and its current status.
  • Operator records: certifications, the crane classes each one covers, and the expiry date of each.
  • Inspection logs: pre-use checks, periodic inspections, and open defects with their closure notes.
  • Bookings and site requirements: dates, the load, site constraints, and the competent person named for the site.

Retrieval is the practical way to keep this memory current. The agent looks up the latest record when it checks a booking instead of relying on a summary it wrote last week. Google Cloud's reference architecture for retrieval-augmented applications describes this pattern in general terms, and it is a useful starting point when you discuss the design with your IT provider. Keep the sources small at first. A narrow memory that is accurate is more valuable than a broad one that is partly stale.

The agent is only as current as the feeds that update its sources, so monitoring the feeds is part of the job.

How do you plan the first agent role?

The first role should be narrow enough that a safety lead can review every rule it applies. A workable plan looks like this:

  • Choose one role: a pre-lift compliance check on bookings for the coming week, for one branch or one crane class.
  • Write the rules in plain language, such as which certificate must be valid on the lift date and which inspection must be closed. Have the safety lead approve each rule in writing.
  • Connect read access only to the scheduling tool, the fleet register, the operator records and the inspection log. Leave billing, payroll and customer contact out of scope.
  • Define the output format: a flag that states the booking reference, the gap type, the record the agent relied on, the person who must act and the date by which they must act.
  • Set the approval boundary before connecting anything: the agent never reassigns an operator, changes a booking status or closes an inspection finding.
  • Run the check against past bookings where the outcome is already known, and compare the agent's flags with what staff caught and missed.
  • Widen the scope only after the review shows that the flags are accurate, specific and timely enough to act on.

This sequence is deliberately conservative. Anthropic's guidance on building effective agents makes a similar point: a simple, well-bounded workflow is often a better first step than an open-ended autonomous system.

Illustrative scenario: an operator certificate expires during a job

The following scenario is illustrative and does not describe a real client. A two-week job is booked for a mobile crane with a named operator. The operator's certificate for that crane class expires on the fifth day of the booking. When the agent reviews the booking the day before the job starts, it notices that the expiry falls inside the booking window. It sends a flag to the dispatcher and the safety lead with the certificate record, the booking reference, and the expiry date. It does not reassign the operator or alter the schedule. The safety lead decides whether to renew the certificate before that day, move the operator to a different job, or assign another certified operator. The flag records what was decided and by whom, so the next review can see the full trail.

Which gaps should hold a lift and which should only warn?

The thresholds belong to your safety policy and to the rules that apply where you operate. The agent applies them. It does not write them. A simple decision guide helps the safety lead agree the logic before it is built:

  • If a required certificate is missing or expired on the lift date, the lift is held until a named person clears it in writing.
  • If an inspection log contains an open defect that affects the load or the rigging, the crane is marked unavailable for that booking and the safety lead is notified.
  • If an expiry falls inside the booking window but not on the lift date, the agent raises a warning and assigns a renewal owner with a deadline.
  • If a required record is absent from the system of record, the agent flags a data gap and does not assume compliance.
  • If the agent cannot classify a situation with the rules it has been given, it escalates to a person and does not resolve it itself.

The last rule matters most. An agent that guesses at an unfamiliar case creates the appearance of control without the substance. A clear escalation path keeps accountability where it belongs.

How should permissions and approvals be set?

Permissions should follow the principle of least privilege. The agent receives read access to the sources it needs, write access only to its own flag queue, and no authority over the systems that make the decisions. The OWASP authorization guidance is a useful reference for checking that access decisions are enforced consistently and not only in the interface. If the scheduling tool supports scoped tokens, use them so that each connection can read one data set and nothing else. Review these permissions when people change roles, when a new crane class is added, and when a branch opens or closes.

The NIST AI Risk Management Framework describes the broader discipline of mapping, measuring and managing AI risk, and it is worth reading alongside your existing safety management system. The aim is not to add a new bureaucracy. The aim is to be able to show, for any flag, which record was used, who saw it, and what happened next.

How does Be Human run this in managed operations?

Be Human designs the first role with your safety and operations leads. We define the memory sources, the read and write permissions, the approval boundaries and the success criterion, then we operate the agent under those rules. Managed operations include monitoring whether each feed stays current, recurring evaluations against bookings your staff have already reviewed, permission reviews when roles change, and incident response when a gap was missed and needs a root-cause review. Changes to the rules go through the same approval path as the original design. The service is remote and available in English and Portuguese. If you want to find the first task worth handing to an agent, take the free Be Human AI diagnostic.

Frequently asked questions

Can an AI agent decide whether a crane lift is safe? No. The agent checks bookings against records and applies rules that your safety lead has approved, then flags what it finds. The decision to proceed, hold or override stays with a named person who is accountable for the lift. The agent's flags exist to make that decision faster and better informed.

Do we need to replace our inspection software before we start? Usually not. The first role reads from the systems you already use, and it needs a reliable export or a read connection rather than a new platform. If a source has no usable interface, the first step is to document that gap and decide whether the workflow can wait. Forcing a migration is rarely the right opening move.

How do we keep operator certification data accurate? Accuracy depends on ownership. Assign one person to maintain the operator records and one to review expiry flags each week, and make the agent flag any record that is missing or inconsistent rather than filling it in. The agent can point out where the data is weak, but people must correct it at the source.

How do we know the agent is working? Compare its flags with a sample of bookings that staff have already reviewed, and track whether gaps reached the right person before the lift date. Accuracy matters, but timeliness and clear ownership matter just as much. If the flags are accurate and nobody acts on them, the design needs to change, not the model.

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.

Less busywork. More room for you.

Bring us the work you wish you had time for.

Book a call to try a personalized AI agent free for 7 days ↗Or take the free AI diagnostic →