Forklift Safety Checklist: AI-Built Daily Inspection Forms

A warehouse can build a daily forklift safety checklist with AI by giving it each truck's operator manual and data-plate details, asking for one short form per truck type, then having a competent person check every item against the manual and local rules. AI drafts the wording; a person decides what counts as safe.

The common failure is a single generic form pinned to every truck. A reach truck, an LPG counterbalance and a pallet jack fail in different ways, so the check has to differ, and the form has to say what happens to a truck that fails it.

Follow me on Instagram@sagnikteaches

Gather the paperwork before you write a single prompt

The quality of an AI-built checklist depends almost entirely on what you feed it. Without the manual, the model will write a plausible list based on forklifts in general, and plausible is not good enough for a machine that can lift two tonnes over somebody's head.

Connect on LinkedInSagnik Bhattacharya

For each truck, collect four things:

Subscribe on YouTube@sagnikteaches

1. The operator manual as a PDF or clear phone photos of the daily-check pages. Most manufacturers include a pre-use section; it is the backbone of your form.
2. The data plate details: make, model, serial number, rated capacity, power type, attachment fitted. Photograph the plate itself rather than copying it by hand.
3. The last service report and any thorough-examination report, so recurring problems (say, a slow tilt cylinder) can be added as a specific check.
4. Your site rules: speed limits, pedestrian routes, charging or gas-handling rules, anything that makes your warehouse different from the manual's assumptions.

Group the trucks by type, not by number. A warehouse with eleven trucks usually has three or four types, and each type gets one form. Add the truck's own identifier at the top as a field, so the form is one template used many times.

If the manual is a scan, check the AI actually read it. Ask it to list the headings of the daily-check section and compare against the page. If it skips one, the form will too.

What the daily check covers on each truck type

The table below is a starting map, not a finished form. It shows how the emphasis moves between truck types. Your manual may add or remove items, and it wins over this table every time.

Notice that "often missed" items are where AI earns its keep. A model fed the manual will list every item the manufacturer names; a busy supervisor writing from memory tends to skip the dull ones. The reverse is also true: AI can miss a site-specific hazard, which is why the competent-person review matters.

A prompt that turns the manual into a one-page form

Paste the manual's pre-use section (or attach it) and give a clear job. Here is a prompt I would use for an electric counterbalance truck. Replace the bracketed parts with your own details.

You are helping a warehouse manager draft a daily pre-use inspection form.

Truck type: electric counterbalance forklift, 2.5 tonne rated capacity, standard forks.
Source: the attached operator manual (pre-use inspection section). Use only checks
that appear in the manual, plus the site rules listed below. If the manual names a
check, keep its wording close to the manual's.

Site rules to include:
- Pedestrian aisles are marked yellow; trucks sound the horn at every blind junction.
- Seat belts must be worn at all times.
- Trucks are charged only in the charging bay.

Output a form that fits one A4 page:
1. A header block: truck ID, date, operator name, shift, hour-meter reading.
2. Checks in the order an operator would walk round the truck, grouped as:
   Walk-round (engine off), Cab and controls, Short test drive.
3. For each check: a short question answerable Yes/No/Defect, and a column for notes.
4. A "Do not use the truck if" box listing the checks that stop work.
5. A sign-off line for the operator and a line for the supervisor.

Do not invent figures such as wear limits or pressures. If the manual gives a limit,
quote it and cite the page; if it does not, write "see manual p. __" and flag it.

The final instruction is the important one. Models like to fill gaps with confident numbers, and a wear limit that sounds right but was never in your manual is worse than a blank.

A sample of the kind of output you would get back (illustrative, trimmed to a few lines):

WALK-ROUND (key out, parking brake on)
1. Forks: any cracks, bends or wear at the heel? Fork tips level with each other?  Y / N / Defect
2. Mast and carriage: chains evenly tensioned, no stiff or rusted links, no hydraulic leaks?  Y / N / Defect
3. Tyres: no deep cuts, chunks missing or flat spots; wheel nuts present?  Y / N / Defect
4. Battery: cover latched, restraint in place, cables and connectors undamaged?  Y / N / Defect
5. Overhead guard and load backrest: undamaged and secure?  Y / N / Defect

CAB AND CONTROLS (key on)
6. Charge indicator shows enough for the shift (see manual p. __)?  Y / N / Defect
7. Seat belt fastens and releases; seat secure?  Y / N / Defect
8. Horn, lights and reversing alarm work?  Y / N / Defect
...
DO NOT USE THE TRUCK IF: brakes or steering feel wrong, a hydraulic leak is visible,
a fork or chain is damaged, the seat belt is faulty, or the horn does not sound.

Two things to fix in output like this. First, the "see manual p. __" placeholders need a person to look up the page and write the real figure in. Second, "enough for the shift" is vague; your competent person should replace it with a number (for example a minimum charge level the manual or engineer recommends).

Make each line something an operator can tick in ten seconds

Operators complete forms in gloves, at the start of a shift, with a queue of pallets waiting. Wording decides whether the check is real. Ask the model to rewrite each item so it names the part, the fault and the action, and keep only one question per line.

Run this rewrite as a second prompt after the first draft, with the instruction "keep every safety item, split any line that checks two things, and keep the whole form to 25 lines or fewer". A form with 45 lines gets ticked by memory.

Also ask for a version in the languages your operators actually read. Translation of short, simple questions is a strong use of AI, but have a bilingual operator or supervisor read it, because a mistranslated "defect" can invert the meaning of a tick.

The defect-reporting flow that makes the form worth keeping

A checklist that finds a fault but has nowhere to send it is paperwork. The part most warehouses leave out is the flow after a "Defect" tick. Draw it first, then let AI write the wording for each step.

1. The operator stops. A defect on a "do not use" item means the truck does not move. The operator ties a red tag to the steering wheel or key, removes the key and tells the supervisor in person.
2. The supervisor classifies the defect in one of three levels: stop (unsafe now), fix this week (safe but worn) or monitor (cosmetic or under the limit). The form lists examples for each level so no one improvises.
3. The defect goes into a single log with date, truck ID, who found it, what it is, photo if possible, and the level. One log per site, not one per shift.
4. A named person owns the repair: the in-house engineer or the dealer. The log shows who and by when.
5. The truck returns only on a sign-off: the person who fixed it records what was done, and a supervisor checks it before removing the tag.
6. The log is reviewed weekly for repeats: the same truck, the same part, the same operator or shift.

Ask the AI to draft three things from this flow: the defect-report wording on the back of the form, a one-paragraph tagging rule for a laminated card at each charging point, and an email template to the dealer's service desk. Here is the dealer email it might give you (illustrative, with placeholders):

Subject: Truck 07 out of service - hydraulic leak at mast, request for visit

Hello [first name],

Truck 07 (reach truck, serial [serial number]) was found on the 06:10 pre-use check with
oil on the floor beneath the mast and a wet hose near the lift cylinder. It has been tagged
out and will stay out of service until it has been repaired and signed off.

Photos are attached. Last service was on [date]; hour meter reading is [hours].
Could you tell me the earliest date a technician can attend, and whether you need us
to arrange anything beforehand?

Thank you,
[your name]
[site name]

If your team already keeps a log for hazards, connect the two. The log structure in near-miss logging for a construction company adapts well to a defect register, and a wider policy is covered in writing a health and safety policy with AI.

An online gift shop's four-truck warehouse, start to finish

This is an illustration, not a case study. Say an online gift shop runs its own warehouse with about 900 pallet spaces and four trucks: two electric counterbalance trucks, one reach truck and one powered pallet truck. Until now it has used one A4 form that says "Forklift check: tick if OK" with eight lines, filled in on Monday mornings for whichever truck the manager remembered.

Week one: the paperwork. The manager photographs the data plates and the daily-check pages of the three manuals (the two electric trucks share one). The service reports show that the reach truck had a mast-chain adjustment twice in a year.

Week one, afternoon: three drafts. Using the prompt above, the manager asks for three forms, one per type. The model returns them with eighteen to twenty-four lines each. The reach truck draft includes the chain check twice, once on the walk-round and once under "mast raised". The manager removes the duplicate and adds, in the notes column, the extra reason for the second look.

Week two: the competent-person review. The dealer's technician reads all three, compares them with the manuals, and makes six changes: two placeholder limits replaced by real figures, one missing check added (the reach truck's operator restraint), a wording fix on the brake test, one item moved to a weekly form (a lubrication check), and one new line for a damaged fork-positioning sensor flagged on the last service. The manager records the technician's name and date at the top of each form as the version approved.

Week three: a trial. The forms are printed and clipped to each truck. Operators fill them in at the start of each shift. After five days, the log shows 20 truck-days of checks, three defects (a loose seat belt buckle, a worn tyre and a hydraulic hose weep), and the time per check averages four minutes, which comes to 80 minutes a week across four trucks for one shift. That is the whole cost in time.

The numbers worth watching. The shop sets two figures: completion rate (checks done out of trucks used, target 100 per cent) and defect-to-repair time (stop-level defects closed within one working day). In week three completion is 19 of 20; the missing one is a truck used by an agency worker who had not been told. The fix is a line in the induction, not a new form.

What the competent person must verify before the form goes live

The AI drafts; this review makes the form safe to rely on. Give the reviewer a short list so they check the same things every time.

1. Every manual-listed daily check appears on the form for that truck type. Take the manual's list and tick each one off against the form.
2. Every figure is real: wear limits, pressures, minimum charge, mast tilt, fork heel thickness. Replace any "see manual" placeholder and delete any number you cannot trace to the manual or a service engineer.
3. The stop conditions are right: brakes, steering, hydraulic leaks, damaged forks or chains, faulty seat belt and horn. No item that could cause a crash should be allowed as "fix this week".
4. Site rules match practice: charging bay rules, pedestrian aisles, gas handling, speed limits.
5. The form meets local requirements: how often the check must be done (some rules require an examination at least daily before the truck is placed in service, and after every shift where trucks run round the clock), how long the record is kept, and which equipment needs periodic thorough examination by a qualified person. These vary by place, so confirm them with your regulator, insurer or safety adviser and write the result on the form.
6. Operators are trained and authorised to use that truck type. A checklist does not replace operator training or licensing.

Record the reviewer's name, role, date and version on each form, and keep the previous version. If an incident ever happens, you want to show which form was in use and who approved it.

How an AI-built form fails in practice

These are the failures I would plan for, because they come from the way the tools behave and from the way people use paper.

• The invented limit. A draft says "replace tyres when worn below 50 per cent" and nobody notices. The prompt rule against inventing figures reduces this; the reviewer's number check removes the rest.
• Merged trucks. A warehouse with an old and a new reach truck assumes the same form fits. The newer one has a camera system and a seat-sensor interlock the older one lacks. Name each model in the header and have one form per model.
• The silent deletion. When you ask the AI to "shorten", it may drop a real check. Always compare the shortened form to the previous version, line by line, or ask the model to list what it removed.
• The ticking habit. A form that never shows a defect for three months is not proof of a perfect fleet. Look for patterns such as the same handwriting, identical times or every item ticked in under a minute. A supervisor who spot-checks one truck a week catches this fast.
• The shared drive trap. A phone form that nobody watches fills up with defect alerts that go unread. Assign the alert to a named person, and test it once a quarter by logging a fake defect.
• Stale versions. After the manual or the truck changes, a printed form on the wall keeps going. Put the version date in the footer and review all forms once a year, or whenever a truck is replaced.

Treat this as your first AI process with a safety consequence, and apply the same principle you would with any other: keep the model to drafting, keep a person on the decision, and keep a record of who decided. The wider case for that split is in AI tools for warehousing.

Turning the log into a weekly review with AI

After a month, the log itself becomes useful data. Export it as a spreadsheet and ask the model to find patterns. A prompt that works well:

Here is a month of forklift defect-log entries (columns: date, truck ID, shift,
item, level, description, days to close). Please:
1. Count defects by truck and by item.
2. List any item that appears three or more times on the same truck.
3. Say which defects took longer than two days to close, and who owned them.
4. Suggest, as questions for the maintenance engineer, anything that looks like a
   recurring cause (for example, repeated hose weeps on one truck).
Do not recommend repairs; only describe patterns and questions to ask.

A sample reply (illustrative): "Truck 03 shows tyre-wear defects on four dates; three were found by the night shift. Hydraulic weeps appear twice on truck 07, both on the mast hose. Two stop-level defects took four days to close, both waiting on parts." The manager's job now is to call the dealer about truck 07's hose and to ask why parts wait four days.

The same weekly slot suits other warehouse reviews too: tracing pick errors back to their cause uses a similar short review of a log to stop small problems growing.

Six smaller uses of the same method around the warehouse

Once you have the manual-to-form pattern, it applies to other equipment. Each takes about an hour the first time.

• Dock levellers and loading doors: a weekly form from the manufacturer's manual, with a "do not use" box for damaged lips or broken hold-backs.
• Pallet racking: a monthly walk-round form listing the damage types your racking supplier names (bent uprights, missing beam locks, overloaded bays). Racking inspection by a competent person remains a separate duty.
• Charging bays: a short daily check on ventilation, signage, spill kit and cable condition.
• Hand-held scanners and wearables: a weekly battery and damage check, so a dead scanner does not push pickers into unsafe shortcuts.
• Visitor and driver rules: a one-page card in plain language, generated from your site rules, for drivers arriving at the dock. Translate it for the languages your regular drivers use.
• Induction quiz: ten questions generated from your site rules and truck forms to check that new starters read them. The questions need the same competent-person check as the forms.

For slot rules at the dock itself, dock scheduling for a small warehouse covers reminders and booking windows that reduce the rush which causes shortcuts.

Keeping the form alive after launch

Forms decay. Put three dates in the diary on day one. After 30 days, read the log with the operators and ask which lines are never ticked "Defect", which are unclear and which are missing; trim and reword. After 90 days, send the revised forms for another competent-person check. After 12 months, or whenever a truck is bought or sold, redo the manual-to-form step for that type.

Keep one master file per truck type with the manual version, the form version, the reviewer and the date. That file is what you show an auditor, an insurer or a new safety manager.

If you would rather map your own trucks and paperwork with someone, that is the kind of job my AI implementation consultation covers, and you can reach me through the contact page.

Questions warehouse managers ask before switching to an AI-built form

Can AI write a forklift checklist without the manual?

It can write a generic one, but a generic list is the weak point. Truck-specific items such as mast chain wear limits, battery restraints or the gas-cylinder mount come from the operator manual. Upload the manual and the data plate details, then have a competent person confirm that every manual-listed check made it onto the form.

How long should the daily check take an operator?

A well-built form for a counterbalance or electric truck should take a few minutes, including the walk-round and the short test drive; time a handful of checks on your own site to set the norm. If it takes ten, the form is probably too long and operators will start ticking without looking. Trim wording, not safety items, and move occasional checks to the weekly form.

Should operators fill it in on paper or on a phone?

Either works if the record is kept, dated and tied to the truck and the operator. Paper suits a noisy dock with gloves and no signal. A phone form suits a site that wants photos of defects and automatic alerts to the supervisor. Pick the one operators will actually complete at the truck.

Does a ticked form prove the truck is safe?

No. It proves someone looked at the truck at a certain time and recorded what they saw. The form supports, but never replaces, planned servicing and statutory thorough examinations by a qualified engineer, and it only works if defects found really do take the truck out of service.

Who counts as a competent person?

Usually someone with the training, experience and authority to judge forklift condition and the local safety rules, such as a qualified truck engineer, the dealer's service technician or a trained safety officer. The exact definition depends on the rules where you operate, so confirm yours with your regulator or insurer.

Further reads

• Simpro Form Builder: Draft Job Sheets and Safety Checklists With ChatGPT — Another way to turn checklists into reusable forms.
• Cycle Count Template: AI Plans Counts by ABC Class — Another routine you can schedule once and keep tidy.
• Factory Training Videos With AI: Safety Clips in Several Languages — Short safety clips for operators who read other languages.
• Peak Season Ecommerce Checklist: Stock, Support and Shipping — Peak-season stock, support and shipping checks.
• AI for Warehouse Loading: Plan Pallets and Trucks From Orders — Planning loads so trucks are not overstressed.
• AI Robots for Warehouses: Do They Pay Off for a Small Operator? — Whether robots would pay off in a warehouse your size.
• 3PL Client Onboarding Checklist: AI-Drafted SOPs and SLAs — The 3PL's own onboarding checklist for a new client: what to collect, how AI drafts the client SOP and SLA, and who signs what before the first pallet arrives.
• Shift Handover Template: Let AI Turn Operator Notes Into a Report — A fillable shift handover template and an AI prompt that turns rough operator notes into a report, with safety items protected.
• Warehouse SOPs: AI Training Cards for Picking and Packing — One task per card: photograph the best operator, let AI draft the steps, translate the card, and have a supervisor test it on a new starter.
• How Is AI Used in Warehouses? Small-Warehouse Examples That Work — Four warehouse jobs a small operation can give to AI today, with numbers, and a plain test for when robots might pay.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: a published workplace-safety standard for powered industrial trucks (examination before a truck is placed in service, at least daily, and after each shift on round-the-clock use; defects reported at once), read 5 November 2026; operator-manual structure and inspection practice described generally; no vendor-specific figures used. Check your own truck manuals, local workplace-safety rules and insurer requirements.