Lone Working Policy for Trades Firms: Check-In Rules Drafted With AI

Include who may work alone and who may not, tasks that need two people, check-in times, a step-by-step response to a missed check-in, emergency contacts, phone or app tools, training and a review date. Draft it with AI from your own risk assessment, then have a competent person review it. None of this is legal advice.

The policy is the rulebook every lone worker follows; the risk assessment is where the reasons live. Regulator guidance I read says it will often be safe to work alone, but the law requires you to think about the risks first, and that the assessment should decide the level of supervision. So write the assessment, then the policy, and let an assistant turn one into the other.

Follow me on Instagram@sagnikteaches

What goes in the policy, section by section

A small firm's policy fits on two pages. Longer ones do not get read on the seat of a van. Every line should trace back to something in the assessment, and a line that cannot is either unnecessary or a sign the assessment is missing something.

Connect on LinkedInSagnik Bhattacharya

For the wider health and safety policy this sits inside, see writing a health and safety policy with AI and what to check. For a cleaning firm, the companion piece is the site-by-site lone worker risk assessment for cleaners, which shows the assessment form that a policy like this one relies on.

Subscribe on YouTube@sagnikteaches

Which jobs may be done alone, and which need a second person

The regulator's leaflet gives examples where at least one other person may need to be present, including working in a confined space, where someone is also needed for the rescue role, and working near exposed live electrical conductors. It adds that the level of supervision is a management decision based on the risk assessment. The tiers below are my suggestion for a trades firm and not regulator text, so let your own assessment decide where each task falls.

Two further rules belong in this section. A new starter is accompanied at first; the guidance says a lone worker may need company when new to a job, in training, doing a job with specific risks or facing new situations. And a worker who finds something unexpected, such as an unsafe structure or an aggressive customer, stops and calls the contact, because the policy should state the limits of what can be done alone.

What the contact needs on the day: a day sheet and a card

A contact who receives "arrived" with no address is no use at 12.50, when the ladder says to ring the customer. Each morning the contact needs a day sheet listing where every worker will be, and a card that sits by the phone with the ladder and the numbers. The sheet below is illustrative, with placeholders where a real address and number would go.

Booking notes often hide the facts that set the tier, and an assistant can read them and propose flags for a person to confirm. Suppose a note says: "Boiler lockout. Tenant says landlord hasn't been. Flat is empty most days, key under the mat, last engineer was shouted at." A good reading flags an empty property, a key arrangement worth raising with the agent and aggression on the previous visit, and proposes amber. What you fix is that the key detail should not travel further than it needs to, and that you, not the assistant, set the tier. A day sheet carries customer addresses, so keep it to the contact and the worker, and do not paste it into a tool whose data terms you have not checked.

Check-in rules people actually follow

Check-ins fail when they are fiddly. Every rule below is there because a check-in that takes a minute to compose is one that gets skipped on a busy Friday.

• One channel. Text or one phone call. Not "message or email or call, whichever."
• One word. "Arrived", "OK", "finished", "late". A worker can send it with a wet thumb.
• A named contact and a deputy. The contact must have the authority and the numbers to act. A person on holiday is not a contact.
• A rota for the contact. Pin who is on duty this week where everyone can see it.
• A "running late" message. It resets the clock, so lateness is not mistaken for danger.
• A "home safe" message. For evening call-outs and remote jobs, the guidance says to have a system to confirm a lone worker has returned to base or home once the work is done.

The intervals are a decision for your assessment, but a starting point looks like this. Treat the figures as suggestions to test, not rules.

Policy wording that works in practice: "On every job, text the contact 'arrived' when you arrive. On amber jobs, text 'OK' every 90 minutes and 'finished' when you leave. If you will be late for a check-in, text 'late' with a time. If the contact does not hear from you, they will follow the missed check-in steps."

The missed check-in ladder

A ladder turns "someone should do something" into named actions with times. The steps below are a starting point, and the intervals should be shorter on amber tasks and longer where an assessment justifies it.

Say in the policy who may skip steps. Escalate early when the worker is on an amber task, when the last message sounded unwell, when the property is empty or remote, or when the phone goes straight to voicemail. A false alarm costs ten minutes of embarrassment, and the leaflet's advice on emergency procedures is to establish them, train people in them and practise them regularly and realistically.

A six-person plumbing and heating firm, in numbers

All figures here are illustrative. A firm has four field engineers, an office manager and the owner. Each engineer does about five jobs a day, and two of those a day are amber: a loft, an empty flat or an evening call-out.

The firm's policy asks for three check-ins per engineer a day (arrival, midday, finish) plus one extra on each amber job, so 4 x (3 + 2) = 20 check-ins at most. If the contact spends 20 seconds reading and ticking each one, that is about 7 minutes a day, or around 140 minutes over a 21-day working month, which is just over two hours. A dedicated app that cost, say, $8 per user per month would be $32 for four engineers; the figure is only an illustration, and you should get real quotes. The comparison is not about money. It is about whether anybody reliably reads the messages.

Here is what a missed check-in looks like in practice.

It would be easy to call that a failure of the engineer. It is a failure of the rule. The policy now says that on amber tasks the phone stays on the person, set to vibrate for the check-in time, and the log records the incident with the change. That is the realistic mistake and how it shows up: a ladder that works, a trigger that was mildly embarrassing, and a rule that improved because someone wrote down what happened.

Phone and app options compared

The regulator's guidance lists several ways to keep in touch: supervisors visiting, pre-agreed regular contact by phone or radio, devices that raise an alarm manually or automatically, and a system to confirm the worker has returned. Small firms mostly choose from the four rows below.

The built-in features are described on Apple's Check In for Messages page and Google's emergency help for Android page. Both are personal safety features that alert your contacts, not monitored lone-worker services, so treat them as an extra layer, test them on the devices your people really carry, and remember that a phone that is off, flat or left in a van does nothing.

Questions to ask any dedicated service

• Does it work without signal, and what happens when the battery dies?
• Who responds to an alarm, a person or an automated text, and how fast?
• How are false alarms handled, and can the worker raise one silently?
• What does it cost per user, with what minimum term and cancellation terms?
• Where is location data stored, who can see it, and can tracking be off outside working hours?
• Is there a test mode, so drills do not trigger real responses?

Tracking staff location is personal data. Tell people what is collected and when, collect only what the policy needs, and ask your data-protection adviser before you switch it on. If staff use personal phones, get their agreement in writing to what is shared, and never ask anyone to share location outside working hours.

Drafting the policy with AI: prompt, output and fixes

An assistant is good at turning a risk assessment summary into clean two-page prose and at keeping the structure consistent. It is not good at knowing your firm, so give it the facts and forbid it from guessing. Use roles, not names, and leave phone numbers out.

Draft a lone working policy of no more than 800 words for a small trades
firm, in plain English with our usual spelling. Use "we" and "you", not legal language.

Firm facts:
- Trade: [for example, plumbing and heating], [number] field staff.
- Check-in method: text to a named contact; deputy named by role.
- Green, amber and red tasks: [paste the tier table from my assessment].
- Check-in intervals: [paste].
- Missed check-in ladder: [paste].
- Tools: personal or work phones; [app, if any].

Sections: scope, who may work alone, two-person tasks, before the visit,
check-in rules, missed check-in steps, emergencies, tools and privacy,
training, review and records.

Rules:
- Use only the facts above. Write [TBC] where something is missing and list
  every [TBC] at the end.
- Do not cite any law, regulation or guidance by name.
- Do not use "shall" or "must never" unless my facts say so.
- Write [emergency number] where an emergency number belongs.
- End with a list of the assumptions you made.

A section of a typical draft, illustrative, might read: "If the contact has not heard from you ten minutes after a check-in was due, they will call you once and send a text. If they still cannot reach you after 20 minutes, they will call again and ring the customer or site contact named on your job sheet. After 30 minutes the deputy will take over. If you cannot be reached after 60 minutes, or earlier if the signs are bad, we will call [emergency number] and give them your job address, access notes and vehicle registration."

That draft needs fixing in five places, and these are the checks I would run every time.

1. Legal citations. A draft may name an Act or a regulation. Delete it; you do not know it applies, and a wrong citation is worse than none.
2. Absolutes. "Staff must never work alone" is a promise you will break by Tuesday. Replace it with the tier table you actually use.
3. Invented tools. "All staff carry a personal alarm" when none does. Delete anything you do not provide.
4. Unrealistic intervals. A check-in every 15 minutes in a firm with one office manager will be ignored. Set intervals you can keep.
5. Stiff voice. "The employer shall ensure" becomes "we will". People follow a policy they can read aloud.

The same care applies to the paperwork around it. A roofing risk assessment and method statement with a competent person's sign-off shows the pattern for a higher-risk trade, and writing toolbox talks with AI and what to check covers delivering the policy to a crew in ten minutes.

Training, testing and records

The guidance says to test technical solutions and emergency procedures regularly so that lone workers can be reached, and to put training in place that is relevant to the work. The schedule below is a suggestion for a small firm.

A near-miss log makes the review honest. A late check-in that turns out to be a dead battery goes in the log with the fix. An app for logging near misses in a construction company compares ways to keep that record without paperwork.

Who reviews it: a competent person, your insurer and an adviser

A competent person must review the policy. That means someone with the training, experience and knowledge to judge the risks of your type of work, which may be a trained manager or an outside health and safety adviser. Give them the assessment and the policy together so they can check that each rule traces to a risk. Ask your insurer whether any policy conditions mention lone working, and ask an adviser how the rules apply where you operate, because this tutorial cannot.

• Reviewer and date. Written on the policy.
• Next review date. A fixed date plus the triggers: an incident, a new type of work, a new member of staff, new equipment or tools.
• Staff acknowledgement. A signed line showing each person has read it and attended the drill.
• Consultation. The guidance says to involve workers when considering risks and controls, so ask your engineers where signal fails and which jobs feel risky, and write down what they say.

Keep the signed policy in the staff handbook, and give the contact a one-page card with the ladder, the numbers and the information the emergency services will ask for. Writing an employee handbook with AI covers where a policy like this fits among the others.

Lone working policy: questions from small trades firms

Is it legal for my staff to work alone?

The guidance I read says it will often be safe to work alone, but the law requires you to think about and deal with the risks before people do so. It also says a separate lone-worker assessment is not required as long as those risks sit inside your general assessment. Ask your adviser how local rules apply to you.

How often should a lone worker check in?

The guidance leaves the interval to your risk assessment and lists pre-agreed regular contact as one option. A reasonable starting point is arrival, midday and finish for ordinary jobs, and every 60 to 90 minutes on higher-risk tasks. Shorten the gaps for riskier work, then adjust after you test the system.

Do I need a dedicated lone-worker app?

Not necessarily. Text check-ins and the built-in timers on iPhone and Android can cover a small firm with a named contact, provided you test them. A dedicated service earns its cost when nobody can reliably watch check-ins, or the work is high risk. Whatever you choose, test it where your people really work.

What about self-employed subcontractors on my jobs?

Legal positions differ, so ask your adviser. A practical approach is to apply the same check-in rules to subcontractors working on your jobs and to write that into the subcontract, so that nobody is working alone without a contact who is expecting to hear from them.

Further reads

• How to Write a Method Statement for a Small Build, With AI Help — Method statements that sit beside the lone working policy on a job.
• How to Draft Risk Assessments With AI and What a Human Must Check — The assessment this policy should be built from, drafted with AI.
• What to Check in an AI Tool's Privacy Policy and Terms — Check an AI tool's data terms before pasting job sheets or staff details.
• Simpro Form Builder: Draft Job Sheets and Safety Checklists With ChatGPT — Job sheets that carry address, access and risk flags to the contact.
• Tree Surgeon Risk Assessment Template: AI Drafts, You Sign Off — A high-risk trade's approach to AI drafts and competent sign-off.
• Apprentice Training Plan for Trades Firms: Sign-Offs and Reviews With AI — A skills matrix, four-stage sign-offs and a weekly review sheet for a small trades firm, with AI drafting the paperwork and the owner signing.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: a national workplace-safety regulator's lone working guidance leaflet and web pages; Apple Support on Check In for Messages; Google Android Help on Safety Check (read 1 November 2026). Anything beyond those sources is my suggestion for your assessment to confirm. Not legal advice.