Describe each option by what goes where, at what installed thickness, with the manufacturer's own rating, and make no prediction about bills. Give the AI your survey facts and a rule banning savings, payback and percentage language, ask for one paragraph per option with trade-offs, and run a claims check before the proposal goes out.
The customer will ask "how much will I save?" within five minutes, and the honest answer is that it depends on things you cannot see from the loft hatch: the climate, the size and construction of the house, how the household lives, and the heating system and fuel. A proposal that states a figure turns that honest answer into a promise, and an AI draft will state one unprompted. The proposal's job is narrower and more useful. It says exactly what you will install, where, how thick, what it needs, what the trade-offs are and what it costs.
What a savings-free option paragraph contains
Each option gets one paragraph with the same seven parts in the same order, so a customer can lay three options side by side and compare like with like. The parts are: where it goes; what goes in, by product name; the installed thickness; the manufacturer's rating with its unit; what it needs from the customer or the building; the trade-offs; and a pointer to the price line. Nothing in that list describes an outcome, because every item is something you do, measure or read off a datasheet.
The AI's default is to write the outcome instead. Here is one loft option as a first draft tends to read, followed by the same option after the rule is applied. The figures here and throughout are illustrative.
First draft: Our premium loft insulation upgrade will slash your heating bills and typically pays for itself within a few years, keeping your family warm all winter.
Rewritten: Option A, loft top-up. We add 220 mm of mineral wool rolls, in two layers, over the 80 mm already in the loft, across the whole 54 m² floor. The manufacturer's datasheet gives a thermal resistance of 5.5 m²K/W for the 220 mm we add. Baffles at the eaves (quote line 2) keep the roof's ventilation openings clear. Anything stored in the loft will need to sit on a raised platform or come out, because boarding laid directly over the rolls would squash them. Price: quote line 1.
The first version contains three claims nobody can prove for a particular house: slashed bills, payback and warmth. The second contains a thickness you will fit, an area you measured, a figure printed by the manufacturer and two practical facts the customer needs before saying yes. It is also longer and less exciting, which is the right trade for a document someone may keep for years.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowR-value in plain English, and where your figures come from
Customers meet R-value on a datasheet or a rival's quote and assume it predicts savings. It does not. R-value measures thermal resistance, meaning how strongly a layer resists heat flowing through it, and a higher number means more resistance. It depends on the type of material, its thickness and its density, and when layers are stacked their values add together. That makes it a property of the material at a stated thickness, in the way a coat's warmth rating describes the coat and not how warm you will feel in a gale.
Four limits belong in your own explanation, because they stop the number being over-read.
- Compression. Insulation squashed under boarding, or forced into a space thinner than it was made for, does not deliver its rated value. Laying denser material on top of a light, easily compressed layer does the same.
- Framing. Heat also travels through joists and studs, so a roof or wall as a whole behaves differently from the insulation alone. This is called thermal bridging, and it is one reason a loft top-up should cover the tops of the joists.
- Air movement. Gaps, leaks around the hatch and around light fittings let heat and moisture bypass the layer. Sealing them before insulation goes over the top is easier than finding them afterwards.
- The household. Climate, the size and construction of the house, the insulation already there, how airtight the house is, living habits, the heating system and the fuel all shape what insulation changes in practice. None of them is on a datasheet.
Units trip people up. A metric datasheet gives thermal resistance (often still labelled the R-value) in m²K/W, while an imperial one gives the R-value in imperial units, and they are different numbers for the same layer: multiply the metric figure by about 5.7 to get the imperial one. Copy the figure with its unit exactly as printed, and never ask the AI to supply or convert one. If a datasheet lists only conductivity, you can cross-check the thickness, because thermal resistance is the thickness in metres divided by the conductivity.
A quick sum with illustrative figures: 0.220 m divided by 0.040 W/mK gives 5.5 m²K/W, and 5.5 × 5.7 is about 31 in imperial terms. Use the sum to catch typing errors, not to replace the datasheet, because the manufacturer's declared figure is the one you quote.
Feed it survey facts and nothing else
The draft is only as honest as the facts you give it. Use a survey sheet with fixed fields, fill it in on site, and tell the AI that any empty field becomes "TBC on site". This one belongs to the example house used from here on: a 1960s bungalow with a 54 m² loft. Its last line, the customer's own words, is there so the proposal can answer what they asked for without promising it.
SURVEY SHEET (illustrative)
Property: 1960s bungalow, owner-occupied
Loft floor area: 54 m2, measured on the survey
Existing insulation: about 80 mm of old mineral wool, flattened near the hatch
Joist depth: 100 mm
Eaves: ventilation openings on both sides, tight at the edges
Hatch: 600 x 600 mm, no seal
Ceiling light fittings: 6 downlights, rating unknown
Extractor fans: bathroom fan duct ends in the loft
Stored items: boxes covering about 8 m2
Water tank and pipes: tank present, pipes exposed
Customer's words: "warmer in the evenings, and somewhere to keep the boxes"
Datasheets on file: mineral wool rolls, loose-fill (installed and settled thickness), closed-cell foamThe prompt does two jobs beyond the wording rules. It makes the AI tag every number with where it came from, which turns a vague worry (is that figure real?) into a mechanical check (is there a tag?). And it keeps ratings out of the AI's hands.
Draft a proposal section for the insulation options below, using only the survey sheet and the datasheet figures I paste. One paragraph per option, in this order: where it goes, what goes in (product name), installed thickness, manufacturer's rating with its unit, what it needs from the customer or the building, trade-offs, price line reference.
Rules:
1. Tag every number with its source in square brackets: [survey], [datasheet] or [quote]. A number you cannot tag does not go in.
2. Never state or imply savings, payback, percentage reductions, comfort outcomes or bill changes.
3. Copy ratings with their units exactly as pasted. Never convert or estimate one.
4. Where information is missing, write TBC on site.
5. Do not describe fire, health or safety performance.
6. If you mention the customer's own words, quote them and do not promise they will be met.Four options for one loft, drafted and edited
The main example runs the survey sheet through the prompt for four options. The ratings are illustrative stand-ins for datasheet figures, and the prices are invented to show the shape of the comparison.
| Option | What goes where | Rating (illustrative) | Needs from the customer | Trade-off | Price |
|---|---|---|---|---|---|
| A | Mineral wool rolls, 220 mm added over the existing 80 mm, loft floor | 5.5 m²K/W for the added layer | Stored items moved; clear access to the eaves | Lowest cost of the four; stored items need a raised platform | $1,080 |
| B | Blown loose-fill over the existing layer, 250 mm installed and 220 mm minimum settled | 5.6 m²K/W at settled thickness | Same as A; hatch sealed first | Fills around pipes and awkward corners; depth must be even and clear of the eaves openings | $1,190 |
| C | Option A plus a raised boarding platform over 9 m² | As A | Choice of the boarded zone before work starts | Keeps stored boxes off the rolls; adds a step and cost | $1,458 |
| D | Closed-cell foam, 100 mm, sprayed to the underside of the roof | 4.0 m²K/W for 100 mm | Roof condition checked first; loft emptied | Higher rating per millimetre; covers the roof timbers, so later inspection and removal are harder | $4,200 |
Notice the rating column. Foam gives 4.0 from 100 mm, while the rolls give 5.5 from 220 mm, which matches the general rule that closed-cell foam offers a higher rating where space is limited. The table does not say which option is best, because the customer's priorities (evening comfort, storage, budget) differ and the proposal cannot know how any option will feel to live with.
The prices trace back to quote lines: A is 54 m² at $20, and C is A plus $378 for the platform (9 m² at $42). Keeping those sums visible lets the customer see what moves the total when they change their mind.
What the AI wrote for options A and D, and what I changed
Here is the tagged output for option A, close to how a first draft comes back.
Option A, loft top-up. We add 220 mm [quote] of mineral wool rolls in two layers over the existing 80 mm [survey], across the 54 m² [survey] loft floor. The datasheet rating for the 220 mm we add is 5.5 m²K/W [datasheet]. You told us you want the house "warmer in the evenings". We cannot promise a result, because what you notice depends on the house, the heating and how you use it. Stored boxes (about 8 m² [survey]) must move or sit on a raised platform. Price: quote line 1.
That one needs only light edits: delete the tags before the proposal goes out, and check that "two layers" matches how your fitters really lay it. Option D is where the draft drifts, and the first version is worth seeing because it is typical.
First draft: Option D, closed-cell foam. 100 mm [quote] sprayed to the underside of the roof gives a perfect seal and the highest rating available (4.0 m²K/W [datasheet]), eliminating draughts and protecting your roof timbers.
Four phrases fail. "Perfect seal" is an absolute, "highest rating available" is a comparison nobody checked, "eliminating draughts" promises an outcome, and "protecting your roof timbers" is both an outcome and a claim that could be wrong for a roof you have not fully inspected. The draft also left out the trade-off that matters most, because nothing on the survey sheet mentioned it. My version reads:
Option D, closed-cell foam. We spray 100 mm [quote] to the underside of the roof, rated at 4.0 m²K/W for 100 mm [datasheet]. Once sprayed, the foam covers the roof timbers, which makes later inspection and removal harder, so ask your insurer or surveyor whether that matters to you. The roof's ventilation arrangements may need to change with this option; we confirm that on site (TBC on site). Price: quote line 4.
The lesson is that the AI can only write honest trade-offs you have supplied. Keep a short trade-off library in the same project as the survey sheet, one entry per option you offer, written once by someone who fits it every week. For example: rolls compress under boarding and need baffles at the eaves; loose-fill needs an even depth and settles, so the quote states installed and settled thickness; foam covers the timbers it is sprayed onto. Update the library when a job teaches you something, and the next draft inherits it.
The claim check, in five passes
Run these in order before a proposal leaves the building. The first four take a couple of minutes each with a find tool; the fifth uses a second AI chat.
- Number trace. Every number needs a source tag from the prompt. Delete any number without one, then check each tagged number against the sheet, the datasheet or the quote.
- Outcome verbs. Search for will, ensures, guarantees, eliminates, stops, prevents, reduces, lowers, cuts, slashes, keeps, saves and pays. Each hit must describe what you do, not what the customer gets.
- Comparatives. Search for best, premium, superior, most efficient, cheapest, perfect and ideal. Replace each with a property from the datasheet, or delete it.
- Safety words. Search for fire, safe, non-combustible, healthy, breathable and eco. Use the datasheet's own wording or nothing.
- Fresh reviewer. Paste the finished text into a new chat that has not seen your instructions and ask it to list, not fix, every sentence that states or implies a saving, a payback, a comfort or health outcome, an absolute, or a comparison.
On the option D first draft, a reviewer chat returned findings like these.
| Flagged phrase | Why it was flagged | What replaced it |
|---|---|---|
| "gives a perfect seal" | Absolute claim | "We spray 100 mm to the underside of the roof" |
| "highest rating available" | Unchecked comparison | The datasheet figure with its unit and thickness |
| "eliminating draughts" | Outcome promise | Deleted |
| "protecting your roof timbers" | Outcome promise, possibly untrue | The trade-off: the foam covers the timbers |
Budget about 20 minutes of checking per proposal at first, falling as your trade-off library and word lists mature. Compare that with a hand-written four-option proposal at 50 minutes: filling the sheet is already done at the survey, the AI draft takes 5 minutes, the five passes take 20, and you add 10 for your own edits and the price lines. That is 35 minutes, and the work you keep is the checking, which is the part you could not hand over anyway. The marketing version of the same rule, with a fact sheet and a trigger-word scan, is in insulation contractor marketing: autumn campaigns and local pages with AI.
When the customer asks how much they will save
You will be asked, so have the answer ready in a form you can say on the doorstep and print in the proposal. Here is wording that stays truthful without sounding evasive.
I can't give you a saving figure I would stand behind, because it depends on how your house is built, how you heat it and how you live in it. What I can do is tell you exactly what goes in, how thick, what the manufacturer rates it at and what each option involves, so you can compare it with any other quote. If you want a number for your own home, an independent energy assessment is the way to get one.
Add one standing sentence to every proposal as well: "We do not estimate savings, because they depend on the home, the heating and how it is used. Ratings in this proposal come from the manufacturers' datasheets." If a customer arrives with a figure from a chatbot or a rival's leaflet, thank them, say you cannot confirm it for their house, and steer back to what you can show: the options, the thicknesses and the datasheet ratings. For a comparable approach to another trade, heat pump quote template: explain options with AI, no savings promises covers the same discipline for heat pumps.
Insulation-specific rules that may apply to you
Some markets go further than general advertising law and regulate insulation sales directly. In one well-documented example, the installer must give the customer a dated, signed receipt or contract showing, for each part of the house, the coverage area, the thickness and the rating taken from the manufacturer's data, without adding up areas or ratings across different parts. For loose-fill the receipt must also show the initial installed thickness, the minimum settled thickness and the number of bags used. The same rule says that if an advertisement or sales material implies lower fuel bills, the seller needs a reasonable basis for it and must include a fixed disclosure that savings vary, although television, radio and online ads too small to carry the wording are excused that disclosure.
Check whether a rule like that applies where you trade, using your trade body or a qualified adviser, because its exact wording beats anything an AI drafts. Even where none applies, two of its habits are worth copying: a settled-thickness line in every loose-fill option, and a dated record of area, thickness and rating for each part of the house. One more habit goes beyond what that rule asks. Leaving the product labels from each package with the customer at handover shows what was used and how much of it, and gives the customer something to compare against the proposal.
Three slips from drafts like these
The unit mix-up. A draft quoted "R-8" next to a metric datasheet figure of 5.5 m²K/W, because the AI tried to be helpful and convert. The numbers were not even the same layer. The prompt now bans conversion, and the datasheet figure is pasted with its unit.
The borrowed fire claim. For a rigid board option the AI added "non-combustible" because it had seen the phrase attached to mineral wool. The board's datasheet said nothing of the kind. Safety descriptors now come from the datasheet or stay out, and "non-combustible" is on the search list.
The boarding advice. A cheerful line told the customer they could "board the whole loft for storage afterwards". Boarding laid straight onto the rolls compresses them, which is exactly why option C includes a raised platform. The trade-off library now carries that rule so the AI repeats it rather than contradicting it.
The proposal explains what you will do and why it costs what it does. The quote that sits beside it carries the exact lines, areas and exclusions, and insulation quote template: material, thickness and area lines with AI shows how to build that part. If the survey also turns up damp, the notes in damp survey report: write findings and quote lines with AI and home energy audit report: turn findings into a client report with AI cover the neighbouring paperwork.
Questions contractors ask about savings wording
Can I quote the manufacturer's savings figures in a proposal?
Treat them as the manufacturer's claim, not yours. Savings depend on the customer's house, climate, heating and habits, so a general figure does not apply to a specific proposal. Safer wording states the datasheet's thermal rating with its unit, says what you will install, and points the customer to an independent assessment if they want a savings estimate. If your market has an insulation advertising rule, follow its wording.
Should the proposal show R-values at all?
Yes, if you copy them from the datasheet with their unit and say which thickness they apply to. They describe the material, not the customer's result. Add one plain sentence explaining that a higher number means more resistance to heat flow and that the figure does not predict bills. Never let the AI supply or convert a rating for you.
Can AI compare my options in a table for the customer?
Yes, provided every cell comes from your survey sheet or datasheets. Compare what goes where, thickness, rating with unit, what each option needs from the customer, trade-offs and price. Leave out payback and savings columns. Check the table cell by cell, because an AI will sometimes fill a gap with a plausible figure.
What if a customer brings an AI-generated savings estimate?
Thank them for it and be clear you cannot confirm it for their home, because the estimate is general and the real figure depends on the climate, the house, the heating system and how the household lives. Offer to explain the options and the datasheet figures, and suggest an independent assessment if they want a number.
Further reads
- Insulation Leads: What They Cost and How AI Qualifies Them — What insulation leads cost and how AI qualifies them.
- Insulation Takeoff From Plans: Can AI Do It, and What to Verify — Measuring areas from plans, and what to verify.
- Loft Conversion Quote Template: Stages and Provisional Sums With AI — A quote template that handles provisional sums clearly.
- Heat Pump Installer Software: Surveys, Heat Loss and Quotes, AI Compared — Survey-to-quote software for a neighbouring trade.
- AI Solar Design Software Compared: Cost per Proposal for Small Installers — Cost per proposal for small installers using design software.
- How to Back Up Every Claim AI Writes in Your Marketing — A general routine for proving what AI writes.
- 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.
- Solar Proposal Software: Adding AI-Written Cover Letters and FAQs — A worked proposal showing how to draft cover letters and FAQs with AI from your own modelled figures, without savings or green-energy promises.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: a public home-insulation fact sheet (R-value, compression, thermal bridging, ventilation and the factors behind energy savings) and the text of an insulation labelling and advertising rule, read in November 2026.