Last month, we shipped dayBrain Volt — an AI-powered quoting engine built specifically for trade sector operators: HVAC businesses, electrical contractors, construction firms. The kind of companies where quoting is both essential and expensive.
During development, our initial cost to generate a single professional quote using AI was approximately £1.00. After a focused round of prompt re-engineering and a deliberate model switch — from Claude Sonnet 4 to Claude Haiku 4.5 — that cost dropped to £0.14 per quote.
That is an 86% reduction. And it changes the economics of quoting fundamentally.
This post is not a product announcement. It is an honest account of what we did, why it worked, and what the numbers actually mean for a business that sends dozens or hundreds of quotes every month. If you run operations, own a trade business, or make technology decisions for one, the margin implications here are worth understanding in detail.
The Real Cost of Quoting (That Most Businesses Do Not Measure)
Before we get into the AI cost reduction, we need to establish what quoting actually costs — because most operators dramatically underestimate it.
The visible costs are obvious: quoting software licences, maybe a dedicated estimator's salary. But the invisible costs are where the real money goes. We have worked with enough trade businesses at Daybrain Consult to know that this is one of the most consistently undercosted functions in the sector.
The Hidden Cost Stack
Consider a mid-sized HVAC contractor sending 60 quotes per month. A realistic cost breakdown looks something like this:
- Estimator time: 45–90 minutes per quote at fully-loaded cost of £50–£80/hour = £37–£120 per quote
- Admin overhead: Chasing information, formatting documents, following up = another 20–30 minutes
- Opportunity cost: Senior time pulled from billable work or business development
- Error correction: Requoting after scope misunderstandings, pricing errors, or omitted line items
- Delay cost: Quotes that take 48–72 hours to produce, sent to prospects who have already signed with a competitor
Run those numbers honestly and a single quote costs between £60 and £150 when you account for all inputs. At 60 quotes per month, that is £3,600–£9,000 in monthly quoting overhead — before you factor in the quotes that never convert.
Now consider your win rate. Industry averages for trade contractors sit between 25% and 40% on competitive tenders. That means for every job you win, you are paying for two or three quotes that generate zero revenue. The true cost per won job, from a quoting perspective alone, is closer to £150–£450.
This is the problem dayBrain Volt was designed to solve. But the engineering decisions we made along the way revealed something broader — a methodology for building AI-powered tools that are not just functional, but genuinely cost-effective at scale.
What We Built and Why Cost Matters More Than You Think
dayBrain Volt takes job inputs — scope description, materials, location, client type, job complexity flags — and produces a structured, professional quote document. Not a rough estimate. A properly formatted, line-itemised proposal with cover language, terms, and pricing that reflects real-world trade rates.
The AI component handles the language generation, scope interpretation, and document structuring. A human still reviews and approves before anything goes to a client. That is by design — the tool is built to eliminate the labour-intensive drafting work, not the professional judgement.
Why Per-Quote Cost Is the Right Metric
When we talk about AI costs in most enterprise software contexts, the numbers are abstract. Token counts, API pricing tiers, context window limits — none of it means anything to a business owner deciding whether to invest in new technology.
Per-quote cost is concrete. It is directly comparable to your existing cost base. It scales with your volume. And it determines whether an AI quoting system actually improves your margins or simply shifts where the cost sits.
At £1.00 per quote, the AI cost is manageable but not transformative. You are still paying around £60–£150 in human labour per quote, so adding £1.00 in AI cost gives you modest savings only if the tool meaningfully reduces that labour time. The business case is thin.
At £0.14 per quote, the calculus changes entirely. The AI cost becomes genuinely negligible relative to the value created. A business sending 200 quotes per month pays £28 in AI costs. If the tool reduces average quoting time from 60 minutes to 12 minutes — a reasonable outcome based on our testing — the labour saving alone is worth several thousand dollars monthly.
That is a real return. And it compounds as volume grows.
How We Achieved the 86% Reduction: The Technical Story
The reduction came from two distinct interventions, and it is worth being precise about each one. Vague claims about 'optimisation' are not useful to anyone making technology investment decisions. Here is exactly what we did.
Intervention 1 — Prompt Re-Engineering
The first version of the dayBrain Volt prompt was, frankly, what you get when engineers build something that works before they build something that is efficient. It was comprehensive. It was also bloated.
The initial prompt architecture included extensive context-setting preamble, repeated instructions that duplicated logic already embedded in the system prompt, verbose output formatting specifications, and example outputs that added significant token overhead without proportional quality gains.
Prompt re-engineering is not about cutting corners. It is about surgical precision — identifying which tokens are doing real work and which are noise. We went through four iteration cycles:
- Audit: Every instruction in the prompt was categorised as essential, conditional, or redundant.
- Consolidation: Redundant instructions were removed. Conditional instructions were restructured to activate only when relevant job parameters were present.
- Output specification: Instead of describing the desired output at length, we moved to a structured output schema — telling the model the exact format to return rather than describing it narratively.
- Quality validation: Each iteration was tested against a dataset of 50 representative job types to ensure output quality did not degrade.
The result was a prompt that used roughly 40% fewer tokens on average while producing outputs that scored marginally higher on our internal quality rubric. Tighter instructions, it turns out, produce more consistent outputs.
Intervention 2 — The Model Switch
The second intervention was the model switch from Claude Sonnet 4 to Claude Haiku 4.5. This is where most of the cost reduction came from — and it is the decision that requires the most explanation, because it is not obvious and it is frequently misunderstood.
The instinct in AI product development is to use the most capable model available. More capable model equals better output equals better product. This logic is understandable. It is also often wrong.
The relevant question is not 'what is the most capable model?' but 'what is the minimum capability required for this task to meet our quality bar?' Those are very different questions.
Generating a professional trade quote is a structured, well-defined task. The inputs are consistent. The output format is fixed. The domain knowledge required is specific but not deep. It does not require the broad reasoning capability of a frontier model. It requires reliable instruction-following, good language generation, and accurate formatting.
Haiku 4.5 handles all of that comfortably. And it costs a fraction of Sonnet 4.
The key to making this switch work was the prompt re-engineering that preceded it. A sloppy prompt fed to a smaller model produces poor outputs. A tight, well-structured prompt fed to a smaller model produces excellent outputs — because the model is not being asked to compensate for ambiguous instructions with reasoning effort.
Sequence matters: engineer the prompt first, then evaluate which model can execute it reliably. Most teams do this in the wrong order and conclude, incorrectly, that they need a more expensive model.
The Combined Effect
Prompt re-engineering alone reduced our costs by approximately 35–40%. The model switch, enabled by the cleaner prompt, drove the remaining reduction. Together: £1.00 to £0.14. The quality of the output — assessed blind by trade professionals who were not told which version they were reviewing — was rated equivalent or better in 47 of 50 test cases.
The Margin Math: What 86% Actually Means at Scale
Abstract percentages are less useful than concrete numbers. So let us work through the margin impact for a realistic trade business at different quoting volumes.
Worked Example: A Mid-Sized Electrical Contractor
Profile: 12-person electrical contracting business. Sends 80 quotes per month. Average job value: £8,500. Win rate: 32%. Current quoting process: senior estimator spends average 55 minutes per quote at a fully-loaded cost of £70/hour.
Current state (no AI quoting):
- Monthly quotes: 80
- Labour cost per quote: ~£64 (55 min at £70/hr)
- Total monthly quoting labour cost: £5,120
- Jobs won per month: ~26
- Quoting cost per won job: ~£197
With dayBrain Volt at £1.00/quote (pre-optimisation):
- AI cost per quote: £1.00
- Estimator time reduced to 15 minutes (review and approve): ~£17.50
- Total cost per quote: £18.50
- Monthly quoting cost: £1,480
- Saving vs current: £3,640/month
With dayBrain Volt at £0.14/quote (post-optimisation):
- AI cost per quote: £0.14
- Estimator time: £17.50 (unchanged)
- Total cost per quote: £17.64
- Monthly quoting cost: £1,411
- Saving vs current: £3,709/month
- Saving vs £1.00/quote version: £69/month
The direct monthly saving from the 86% cost reduction — comparing £1.00 to £0.14 — is £69 at this volume. That is not nothing, but it is not the main event. The main event is the £3,640 monthly saving that AI quoting delivers regardless of which model version you are running.
So why does the 86% reduction matter? Three reasons.
Reason 1 — Volume Ceiling Removal
At £1.00 per quote, a business generating 500 quotes per month faces £500 in monthly AI costs. That is real money that creates psychological and budgetary friction around scaling volume. At £0.14, the same 500 quotes cost £70. The volume ceiling effectively disappears.
This matters because one of the strategic advantages of fast, cheap quoting is the ability to quote more aggressively — to bid on jobs you might previously have passed on because the cost of preparing a losing quote was too high. Lower per-quote cost enables higher bidding volume, which directly impacts win count even without changing win rate.
Reason 2 — Product Economics and Commercial Viability
For any business considering building or commissioning a quoting tool — rather than buying an off-the-shelf solution — the per-unit cost determines whether the product is commercially viable to offer at a price customers will pay.
At £1.00 per quote, a SaaS product priced at £99/month for 100 quotes is barely viable — AI costs alone consume the entire gross margin. At £0.14, the same product has substantial gross margin to invest in infrastructure, support, and growth.
This is the difference between a product that exists and a product that can be sustained and scaled.
Reason 3 — Competitive Positioning
Your competitors are almost certainly not thinking about this. The businesses that understand the unit economics of AI-powered tools — and optimise for them deliberately — will build structural cost advantages that compound over time. The ones that adopt AI tools at face-value pricing, without optimisation, will find the economics erode as volume grows.
The Framework: How to Evaluate AI Cost in Any Business Tool
The dayBrain Volt story is specific, but the methodology generalises. If your business is evaluating any AI-powered tool — whether you are building it, commissioning it, or buying it — here is a framework for assessing whether the economics actually work.
The AI Unit Economics Checklist
Use this before committing to any AI tool that charges per-use or where cost scales with volume:
1. Define your unit
What is the fundamental transaction this tool performs? A quote, a report, a customer response, a document. Cost per unit is the only metric that matters for margin analysis.
2. Establish your current unit cost
What does this transaction cost today, fully loaded? Include labour time at real cost, overhead allocation, error correction, and delay cost. Most businesses undercount this by 40–60%.
3. Get the vendor to commit to a per-unit cost
If a vendor cannot tell you what their tool costs per transaction, that is either a competence problem or a transparency problem. Either is a red flag. Push for a number. Hold them to it.
4. Model three volumes: current, 2x, 5x
Costs that look fine at current volume can become significant at scale. A tool costing £0.50 per transaction is fine at 100/month (£50). It is a material cost at 2,000/month (£1,000). Run all three scenarios before deciding.
5. Ask what has been optimised
If a vendor is using AI and has not explicitly optimised their prompt architecture and model selection, they are almost certainly paying more than necessary — and passing that cost to you. Ask directly: which model are you running? Have you optimised the prompt? When did you last benchmark against alternative models?
6. Calculate payback period at conservative win assumptions
Do not use best-case productivity gains. Use 60% of the vendor's claimed efficiency improvement. If the payback period is still under six months at 60% of claimed benefit, the investment is robust. If it only works at 100% of claimed benefit, be cautious.
7. Identify the volume inflection point
At what monthly volume does the tool pay for itself from cost savings alone, excluding any revenue upside? This is your minimum viable adoption target. If it is unrealistic given your current operation, the tool is not right for you yet.
Applying the Framework to dayBrain Volt
Using the electrical contractor example above:
- Unit: One completed quote
- Current unit cost: £64 labour
- dayBrain Volt unit cost: £17.64 (AI + review labour)
- Saving per unit: £46.36
- At current volume (80/month): £3,709/month saving
- At 2x volume (160/month): £7,418/month saving
- At 5x volume (400/month): £18,544/month saving
- Payback at 60% efficiency: Positive from month one
The framework forces you to be specific. Specific is uncomfortable, but it is the only thing that lets you make a defensible decision.
Sales Velocity: The Advantage Nobody Talks About Enough
We have focused on cost so far because the 86% reduction is a cost story. But the more significant commercial advantage of AI quoting is speed — and speed affects revenue in ways that cost reduction does not.
In the trade sector, quoting speed is a competitive weapon. Contractors who respond to enquiries within hours rather than days win more work, often without competing on price. Clients who receive a professional quote before they have finished shopping around frequently do not finish shopping around.
The research on this is consistent: response time is one of the strongest predictors of B2B sales success. A study from Harvard Business Review found that companies responding to leads within an hour were seven times more likely to qualify the opportunity than those responding an hour later. The trade sector is no different.
What Faster Quoting Actually Enables
Same-day quoting: A job enquiry received at 9am can have a professional quote in the client's inbox by noon. Before AI quoting, this was possible only for simple jobs or businesses with excess estimating capacity. dayBrain Volt makes it the default.
Quote while on site: A technician completing a service call identifies additional work. Rather than promising a quote 'in a day or two', they can generate a draft on a tablet, review it, and send it before leaving the property. The client's interest is highest in that moment.
Bid on more jobs: When quoting costs £64 in labour, you make judgements about which jobs are worth quoting. At £17.64, those judgements shift. You quote more aggressively, because the downside of a lost quote is materially smaller. Higher bidding volume, even with an unchanged win rate, means more won jobs.
Quantifying the Revenue Upside
Returning to the electrical contractor example: if faster quoting improves win rate from 32% to 36% — a conservative assumption given the evidence on response time — the revenue impact on 80 monthly quotes at £8,500 average job value is:
- Additional wins per month: 80 × 0.04 = 3.2 jobs
- Additional monthly revenue: 3.2 × £8,500 = £27,200
- Additional annual revenue: £326,400
A 4-percentage-point improvement in win rate, driven by faster and more consistent quoting, is worth over £300,000 per year to a business of this size. The cost reduction story — £3,709/month — is real and meaningful. The revenue story is an order of magnitude larger.
What This Means for Businesses Considering AI Tools
The dayBrain Volt development process surfaced a pattern we see repeatedly when working with businesses on technology transformation. The gap between 'AI tool that works' and 'AI tool that creates genuine competitive advantage' is almost always about unit economics and operational fit — not about the technology itself.
Most businesses that adopt AI tools do so at default settings, with default prompts, on whatever model the vendor chose when they built the integration. They get results. They do not get optimised results. And they pay more than they need to for the results they do get.
The 86% cost reduction we achieved was not accidental. It came from deliberate engineering decisions — decisions that required both technical knowledge and a clear-eyed view of what the business problem actually was. The technical capability to build an AI quoting tool is now widely available. The operational discipline to build one that is genuinely cost-effective at scale is much rarer.
Three Questions to Ask Before Investing in Any AI Business Tool
1. Has the vendor done unit economics analysis on their AI costs, and will they share it?
If they have not, they are operating on instinct rather than data. If they will not share it, they are obscuring something. Either way, you need to push harder or look elsewhere.
2. Is the tool designed for your specific workflow, or is it a general AI tool with a trade-sector wrapper?
General tools adapted to vertical markets rarely perform as well as purpose-built tools on the specific task that matters. The quality of the output — and the reliability of the output — depends heavily on how well the AI has been configured for the actual job.
3. What happens to your cost as your volume doubles?
Linear scaling is the minimum acceptable answer. If costs scale superlinearly with volume, the tool becomes a liability as you grow. If costs scale sublinearly — because fixed infrastructure costs get spread across higher volume — that is a genuine advantage.
The Broader Consulting Lesson
We built dayBrain Volt as a product, but the work that made it commercially viable was consulting work: diagnosing the real cost problem, designing the solution architecture, and then iterating the implementation until the economics worked. That sequence matters.
The businesses we work with through Daybrain Consult consistently face a version of the same challenge. They know they need to modernise. They have often already tried one or two technology solutions that did not deliver the promised return. The problem is rarely the technology — it is the absence of a clear diagnostic process before implementation begins.
Quoting is a useful case study because the cost structure is so transparent. But the same methodology applies to any business process where AI or automation is being considered: customer service, scheduling, reporting, compliance documentation, procurement. The questions are identical — what does it actually cost today, what is the minimum capability required, and what does the unit economics model look like at scale?
Businesses that answer those questions before they build or buy make better decisions. They spend less on implementations that do not deliver. They get to value faster. And they end up with tools that create structural advantages rather than tools that simply exist.
The 86% cost reduction in dayBrain Volt is a good number. But the real result is a quoting engine that costs less per quote than a cup of coffee, produces professional output in under two minutes, and gives trade contractors a genuine competitive edge in markets where most of their peers are still quoting the same way they did in 2015.
That is what optimised AI implementation looks like. Not the flashiest deployment. The most disciplined one.
Where to Go From Here
If you run a trade business and you are still quoting manually — or using a quoting tool that was not built for your sector — the numbers in this post are worth sitting with. The cost advantage is real and quantifiable. The speed advantage is real and probably larger than the cost advantage in revenue terms. The competitive edge is real and, for now, accessible to businesses willing to move before their peers do.
If you are a business owner or operations director in any sector thinking about AI or automation investment, the framework in this post applies. Define your unit. Establish your current cost. Demand specificity from vendors. Model the economics at realistic volumes before you commit.
And if you want help doing that analysis properly — working through the diagnostic before the investment, rather than discovering the problems after — that is exactly what the team at Daybrain Consult does. We work with businesses that are serious about modernising and want to do it in a way that actually improves their margins, not just their technology stack.
The calendar link below is a direct booking for a consultation. No sales pitch, no discovery call that turns into a proposal call that turns into a second proposal call. A single conversation where we look at your actual situation and give you an honest assessment of where technology investment will and will not move the needle for your business.
Book a consultation with Daybrain Consult →
dayBrain Volt is a Daybrain Digital product. This post was written by the Daybrain Consult team, drawing on the engineering and commercial decisions made during the dayBrain Volt build. The cost figures cited are real production numbers from the dayBrain Volt development process.