Gas Engineer Software: What AI Can Draft and What You Must Sign

Yes, AI can help a gas engineer with the admin around certificates and job records: customer emails, quote wording, reminders, job summaries and plain-English explanations. But the tests, readings, appliance outcomes, defects and the signature on the certificate must be completed and signed by the registered engineer personally, from what they found on site.

The dividing line is simple to state and easy to cross by accident: if a sentence says what was measured, tested or judged safe, it comes from you. If a sentence explains, schedules or tidies, software can draft it. Most gas engineer software now offers both kinds of feature in one screen, so the skill is knowing which button you are pressing.

Follow me on Instagram@sagnikteaches

What the signed record is actually evidence of

A gas safety record or service certificate is a statement by a registered engineer that specific checks were done on specific appliances on a specific date, with specific results. Published guidance on landlord gas safety records lists the minimum a record must contain: a description and the location of each appliance or flue checked; the name, registration number and signature of the engineer who did the check; the date of the check; the property address; the landlord's (or agent's) name and address; any safety-related defect found and the remedial action taken; and confirmation that the required examinations were made. An electronic record is acceptable when it is secure, can be printed and uniquely identifies the registered engineer, for example by an electronic or scanned signature. The same guidance says the record should be issued once every required check is complete, even when faults are found, rather than held back until repairs are done. Those are the details a regulator, a tenant or an insurer will rely on after an incident, so none of them can be reconstructed afterwards by software that was not there.

Connect on LinkedInSagnik Bhattacharya

That is why the split below goes by what each output claims. A certificate claims to be first-hand evidence. A customer email claims nothing of the sort. Check your registration body's current guidance for the exact content your record type needs, because forms and requirements change and an AI tool will happily describe an old version with great confidence.

Subscribe on YouTube@sagnikteaches

A sorting table for a normal job

Walk through a standard annual service visit and mark each step. The right-hand column is the rule I would apply.

Two points stand out. Defect and warning wording is "draft only" because it sits on the border: the observation is yours, but the customer-friendly explanation is wording. And any step where software proposes a value, such as a pre-selected "pass", should be treated as a prompt to look, not an answer.

What gas engineer software offers, and where AI shows up

Job management apps for trades usually bundle scheduling, customer records, quotes, invoices and forms. AI features tend to appear in three places: writing help (rewriting a note or composing a message), quote and invoice drafting from a short description, and phone or message handling. Pricing and feature sets move quickly, so treat the following as pointers to verify rather than a buying guide.

• ServiceM8 lists an AI writing helper and AI quote and invoice drafting on every plan, with the free plan capped at 10 AI uses a day, per its USD pricing page in November 2026. Its dollar list pricing runs on job count: $0, $29, $79, $149 and $349 a month for roughly 30, 50, 150, 500 and 1,500+ jobs, with unlimited users on paid plans, and each new job card, quotes included, counts as a job. The $79 plan is described as including forms and certificates. Check whether the form library has the record types you issue; do not assume it does.
• Jobber sells quote follow-ups and auto-drafted quotes from its Connect tier upwards, not on the entry plan, and has an AI receptionist add-on. Jobber is a general field service tool, so check how it handles gas-specific record forms, if at all, before relying on it.
• Dedicated gas certificate apps are built around the forms themselves and often show up as vendor sign-up pages when you search. Evaluate them on whether your registered record types are there, whether signatures are captured clearly, and how records are stored and retrieved.

For a comparison written from a heating firm's point of view, read Heating Engineer Software Compared; for two cheaper options head to head, Powered Now vs Tradify for a sole-trader heating engineer. The questions in this tutorial apply whichever you choose.

Test a new tool with a record that never leaves your desk

Before any real customer sees output, run a dummy job. Invent an address, an appliance and a fault, complete the whole flow in the software, and read the certificate it produces next to your paper pad. Look for three things. Does every field you must record appear and accept your own value? Does the software pre-fill or suggest any result? Can you export or print a clear copy, and can you find the record again in a year? If the tool pushes AI-generated text into a field that is supposed to hold a reading or an outcome, that is a reason to reject it, not to be impressed.

Also test what happens when you leave a field blank. A good form refuses to issue the record. A poor one issues it with a default. You want the first.

Example 1: customer explanation from your own findings

This is the best first use of AI because the risk is low and the saving is visible. After a landlord's boiler service you have recorded the results yourself. You then ask for a plain-English summary using only those results.

Write a short, friendly email to a tenant explaining the results of today's boiler service.
Use only the findings below. Do not add safety advice beyond what I list, and do not say the appliance is safe unless I have written that.
Findings: Appliance: wall-mounted combi boiler, kitchen. Outcome as recorded by me: safe to use. Advisory: condensate pipe external run unlagged, may freeze in cold weather. Next service due: November next year.
Tone: clear, no jargon. Under 120 words. Sign off as [your name].

A sample reply (illustrative): "Hello [first name], thanks for letting me in today. I serviced the boiler in the kitchen and recorded it as safe to use. One thing to keep an eye on: the outside part of the condensate pipe is not insulated, so in very cold weather it could freeze and the boiler might stop. Your landlord has the details of the fix. The next service is due in November next year. Kind regards, [your name]". What you check: it said only what you gave it. If the draft had added "no further action required", you would delete that line, because you did not record it.

Example 2: tightening a defect note without changing it

Engineers write terse defect notes. A tidy rewrite is fine as long as the meaning is untouched. Before: "flue joint not sealed, cf. spillage test fail. Turned off, tag on, cust advised". After the AI rewrites it for the customer-facing page: "The flue joint was not properly sealed and the spillage test failed, so I turned the appliance off and attached a warning label. I explained this to you at the visit." Compare the two line by line. Every fact in the second version must exist in the first. A good habit is to put the original beside the rewrite in your own review and delete any word you cannot match.

The decision to turn the appliance off, the label and the conversation are your actions, and the record of them is yours to complete in the proper form. The rewrite is only for the customer's copy of the explanation. If the AI drifts into new advice, such as "do not use any gas appliance in the property", remove it unless you meant it, because it is a safety statement.

Example 3: a quote for remedial work

Quotes are admin, but accuracy still matters because a wrong part number costs money and a missing manufacturer requirement can cost far more. Give the AI your own line items and prices and ask for presentation, not content.

Format this as a customer quote with a short plain-English introduction and numbered lines.
Do not change quantities or prices. Do not add items. Add a line saying the price is valid for 30 days.
Items: 1) Replace expansion vessel, supply and fit: $185. 2) System flush and inhibitor: $140. 3) Replace room thermostat with programmable model, supply and fit: $95.
Total: $420.

Sum check, which you do and the AI does not: 185 + 140 + 95 = 420. If the AI returns a different total, your numbers rule. Then verify parts against the boiler manufacturer's instructions yourself.

Example 4: a landlord portfolio reminder run

A small firm services 60 rental properties for three letting agents. Each year the office needs to send an advance notice to each landlord or agent with the property list and the earliest available dates. Drafting that message is tedious and error-prone by hand. Using the software's reminder or messaging feature, or a spreadsheet and an AI helper, the office can produce 60 messages in a sitting. The safe pattern: the system calculates due dates from the last service date, AI writes the message template once, and a person reviews a sample of five before the batch goes out. Check the due dates because a wrong interval is a missed service, and a missed service on a rented property is a serious problem for the landlord. For reminder structure, see AI service plan reminders for heating and cooling customers.

Example 5: photos for the handover pack

On installations, a photo set of the work (flue terminal, pipework, data plate) goes into the handover pack. AI can sort photos into labelled groups and suggest captions, which saves an evening of renaming files. It cannot judge whether the flue terminal position is acceptable or read a faded data plate reliably. Treat captions as drafts: open each picture and confirm. The checklist idea is covered in the boiler installation checklist, where AI sorts the photos and you sign the records.

Example 6: the call that says "I can smell gas"

Many gas firms are tempted by an AI receptionist or website chatbot to catch missed calls. It can book annual services well. It must not be the thing that handles a suspected leak. Set the script so that any mention of a gas smell, a carbon monoxide alarm, headaches, or a fault that sounds urgent triggers a fixed message telling the caller to follow the national gas emergency procedure and to call the emergency number, and then alerts you or your on-call engineer immediately. Write that message yourself and have the tool read it verbatim; do not let a model improvise safety advice.

Test it the way you test a form. Ring your own line and say, in six different ways, that you can smell gas: "there's a funny smell by the cooker", "my CO alarm keeps beeping", "I think there's a leak". Count how many trigger the fixed message. If two out of six slip through to ordinary booking, the setup is not ready, whatever the vendor's demo showed. The same caution applies to the website form; what the AI chatbot and form should ask a heating engineer's customers sets out the fields to include.

Handling an apprentice or a subcontractor

If a trainee or a subcontracted engineer works under your business, check what each person may record and sign under their own registration. AI drafting makes it easy for an office worker to produce a complete-looking certificate from a trainee's brief text message. Do not let it. The registered engineer who carried out and verified the work completes the record. The office may format the customer's copy and send it, but the content and signature come first from the person who did the checks. Build that into your software permissions too: if the app lets you assign who can issue a record, limit it to registered engineers, and review the audit trail occasionally to confirm nobody else is issuing certificates.

A mistake that shows up: the pre-filled pass

Here is a realistic failure. A firm adopts an app whose form defaults every appliance outcome to "safe" so that the engineer only changes exceptions. On a busy day an engineer rushes a fourth job, forgets to reopen the appliance screen, and the record goes out with "safe" against an appliance he never tested. Nobody notices until the landlord's tenant reports a problem months later. Nothing in that story involved AI, yet adding AI makes it easier to repeat: a draft that looks complete invites a quick tap on "issue". The defence is structural. Choose forms with no default outcomes, require a reading before the outcome field opens, and train staff to treat any pre-populated field as unfinished.

Writing a short AI rule for your firm

Put the boundary in writing so apprentices, office staff and subcontractors follow it. A one-page policy is enough. A workable draft:

1. AI may draft customer messages, quotes, reminders, job summaries and invoice descriptions.
2. AI may not enter, suggest or infer any reading, test result, appliance outcome or defect classification.
3. Certificates and records are completed and signed only by the registered engineer who did the work, after the work.
4. Any AI-written text that mentions safety is checked against the engineer's recorded findings before it is sent.
5. No customer names, addresses or phone numbers go into a general chatbot; use the job software's own features or anonymise first.
6. The engineer remains responsible for everything issued under their name.

Rule 5 matters because address and contact details are personal data and different AI plans treat inputs differently. Check the vendor's current data terms for your plan before pasting anything, or use the AI inside your job software where the vendor already holds that data under your account. If you want to review a vendor's terms in more detail, what to check in an AI vendor's data processing agreement lists the questions.

Measuring whether it is worth the monthly fee

Do a simple sum before signing up. Say you issue 12 records a week and spend eight minutes per job on customer messages, quote wording and invoice text. If AI cuts that to three minutes, you save 60 minutes a week, about 52 hours a year. At an internal value of $40 an hour that is roughly $2,080 a year, set against a software plan that might cost $29 to $149 a month, so $348 to $1,788 a year. The saving is not guaranteed, and the exact plan you need depends on features, not only job count, so test with a trial before committing. The figures are illustrative; swap in your own volumes.

The cheapest saving often has nothing to do with AI: a form that remembers each customer's appliances saves more time than any drafting feature. Judge software on forms first, AI second.

A monthly spot-check of issued records

• Pull five issued records at random and confirm every reading and outcome matches your site notes or instrument log.
• Read three AI-drafted customer messages from the month and confirm none adds a safety claim you did not record.
• Confirm records are backed up and retrievable by address and date.
• Check that no default values appear on forms.
• Re-read your registration body's guidance if a notice has been issued; update your rule sheet if needed.

If you want help deciding which tasks in your week belong to software and which to you, that is what my AI implementation consultation is for; you can start from the contact page. And for the wider picture of AI in a trade business, the AI hub collects the other tutorials in this series.

Gas engineers ask about AI and certificates

Can AI fill in a gas safety record for me?

It should not. The record states what you tested and found, so every reading, appliance outcome and defect must come from you on site. AI can tidy a customer-facing summary afterwards, but a record that was pre-filled or inferred by software is a record you cannot stand behind.

Is it safe to type customer addresses into an AI tool?

Only if the tool's terms for your plan suit it. Address, phone number and access notes are personal data. Prefer the AI features built into your job software, where the vendor already handles that data, or remove names and addresses before pasting into a general chatbot.

Which tasks are safest to hand to AI first?

Start with customer emails, quote wording, appointment reminders and plain-English explanations of what you found. Those carry no test results. Leave anything that records a reading, a pass or fail, or an unsafe-situation decision out of the AI loop entirely.

Does gas engineer software replace paper pads?

Often, but check what your registration body and your landlord customers accept as a record, and that the software gives you a clear, secure copy that can be printed and a way to keep records for the period required; landlord guidance asks for records to be kept for at least two years. Paper remains acceptable where it is complete and legible.

Further reads

• Is ServiceM8 AI Worth It for a Small Heating Firm? — What ServiceM8's AI helpers do for a small heating firm.
• Electrical Installation Certificate: What AI Can Draft and You Must Sign — The same draft-versus-sign split for electrical certificates.
• Landlord Compliance Checklist: Track Gas Safety and EICRs With AI — How letting agents track gas safety record due dates.
• Boiler Installation Quote Template: Can AI Write the Quote and Cover Email? — Quote wording and a cover email for a boiler job.
• Is Checkatrade's AI-Powered TradeMore App Right for a Heating Engineer? — Whether a newer app is worth switching to.
• How Much Do Missed Calls Cost a Heating and Cooling Business? — What unanswered calls cost a heating business.
• EPC Assessor Software: What to Check First, and Where AI Helps — Check scheme approval, approved-list status and update speed before you pay, then use AI for client emails and wording, never for lodgement.
• EV Charger Installer App or Software: AI and Features Beyond Quotes — A neutral checklist for EV charger installers: survey forms, certificates, handover packs, scheduling, what a general job app can do and where AI helps.
• Backflow Testing Software for Plumbers: Reminders and Reports With AI — A small plumbing firm's route to annual backflow test reminders and customer notes: register fields, reminder ladder, cover-note prompts and filing checks.
• Emergency Lighting Certificate With AI: Test Log to Signed Paperwork — A worked emergency lighting test log, from site readings to a signed certificate: what AI can draft and what the competent person records and signs.
• Fire Protection Software: Inspection Reports and Reminders With AI — A clean asset register, formula-driven due dates, a reminder sequence and report-wording prompts, with findings and compliance judgement left to the technician.
• PAT Testing Software: Where AI Helps With Reports and Reminders — Test the reminder and report claims in PAT software, then let AI draft customer summaries and chasers while the tester keeps every result and decision.
• Best AI for Plumbing Questions: A Ten-Question Test for Plumbers — A ten-question test plumbers can run on ChatGPT, Claude and Gemini, with a scoring sheet and the checks to make before acting on any answer.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: official landlord gas safety record guidance (record contents, electronic records, when to issue, retention), read 1 November 2026; ServiceM8 USD pricing page and Jobber pricing page, read 1 November 2026. The registration body's own site could not be read; check its current guidance and your unsafe situations procedure.