How a Building Services Company Cut Quoting Time by 80% with AI

The Estimator Bottleneck

A 60-person building services contractor in Ireland was losing reactive work because quotes took too long to produce. Twenty-five field technicians servicing commercial HVAC, mechanical, and electrical systems. Somewhere between 30 and 50 quotes every week across planned maintenance contracts, emergency callouts, and system replacements. This was a classic automated quoting problem: the data existed, but the process was entirely manual.

Complex quotes took one to two hours each. Planned HVAC overhauls, chiller replacements, multi-trade scheduled maintenance. The estimator had to dig through the ERP, find similar past jobs, check current parts costs against their schedule of rates, calculate labour hours and margin, and factor in contract-specific pricing. A chiller replacement meant pulling previous chiller jobs, checking current compressor and refrigerant costs, calculating crane hire, disposal, and commissioning labour. Most of a morning, gone on one quote.

Simple reactive quotes took twenty to thirty minutes. A failed AHU, a burst pipe, a boiler lockout. In reactive maintenance, the first company to quote credibly usually wins the job.

The real risk ran deeper. Estimating knowledge was concentrated in two people, one of whom was approaching retirement. Five years of pricing instinct, supplier relationships, and margin judgments locked in their heads, not in any system. When they were on leave or busy, quotes stalled. Pricing consistency was poor, with different people quoting the same job type producing 15–20% variance. And thousands of completed jobs with full cost breakdowns, bill of materials, and labour hours sat in the ERP, never analysed or reused.

In commercial building services, speed wins reactive work and accuracy wins planned tenders.

How It Started: The €1,250 Discovery

The company started with an Enterprise Ireland Digital Discovery engagement. Brian Egan — an Enterprise Ireland approved consultant, former Marie Curie Research Fellow, and Founder of Deep Purple AI Consulting — led the assessment. Total cost: €5,000, with Enterprise Ireland funding 80%. The client's net investment: €1,250.

Over eight weeks, we spent time on-site with the operations team, interviewed the estimators and the MD, mapped the full quoting workflow from callout to invoice, and analysed data quality in their ERP across five years of job history. We identified four AI opportunities, ranked by projected ROI.

The ERP data was messier than expected. Inconsistent job categorisation across five years of entries. Missing fields. Pricing that had been manually overridden hundreds of times. But we had seen this before. Messy data is the norm, not the exception. The Discovery included a data quality assessment that identified what was usable, what needed cleaning, and what the AI could work around.

The quoting bottleneck was the number one opportunity. The data was there but nobody was using it systematically.

The company decided to proceed to a full build. We handled the technical input for the Enterprise Ireland grant application for the build phase. The client's team provided roughly four hours of input per fortnight during the build, leaving them free to run the business while we did the heavy lifting.

What We Built

An AI quoting system analyses historical job data to generate accurate draft quotes automatically. It connects to your existing ERP, finds similar completed jobs, and produces itemised estimates including parts, labour, and margin, in minutes instead of hours. Your team reviews and approves every quote before it goes to a client.

Here is how the system works, step by step:

To make it concrete: a chiller replacement quote that used to take the estimating team most of a morning now takes fifteen to twenty minutes to generate a review-ready draft with compressor costs, crane hire, disposal, and commissioning labour all pre-populated from historical patterns.

The system also pre-populates method statements and preliminary safety plans from templates, removing more admin from the estimating team's day.

The Design Decisions That Mattered

Not every quote can be automated. Bespoke one-off jobs with no historical precedent still need human judgment from start to finish. But for the 70–80% of quotes that follow known patterns, the system handles the research and the first draft.

The Results

How we measured this: We tracked baseline quoting time across three categories (reactive callouts, planned maintenance, and system replacements) over a four-week period before deployment. Post-deployment, we measured the same categories over the first six weeks of live use. The 15–20 hours per week figure is based on approximately 35 quotes per week averaging 25–30 minutes of manual research time each, reduced to 3–5 minutes of AI-assisted review time, spread across two estimators and one operations admin.

The estimating team went from being the bottleneck to being a review-and-approve function. Management gained visibility into pricing patterns for the first time: which job types are most profitable, where margins are thin, where material cost changes are quietly eroding margin on repeat contracts.

The company can now quote more jobs without adding headcount. The system handles the research; humans handle the judgment.

And the retiring estimator's knowledge? It is captured in the system now. The business risk of losing that person did not disappear, but it got a lot smaller.

"The big thing for us was that Declan's pricing knowledge, all the stuff that was only in his head, is now in the system. The rest of the lads can actually produce accurate quotes without waiting for him."

Results vary by business. Figures shown are measured and estimated during delivery.

What It Actually Cost

Placeholder introduction to the cost breakdown.

What was included: Read-only ERP integration, AI quoting engine, confidence scoring, estimator review dashboard, audit trail, method statement templates, four-week calibration period, and retained support.

What was not included: Replacing the ERP, implementing a new job management system, or rewriting the company's commercial terms or pricing strategy.

Project costs vary depending on scope, data complexity, number of integrations, and what grant funding is available. This project was a relatively focused build: one ERP integration, one core AI function, and a review dashboard. Projects with multiple system integrations, more complex data structures, or additional functionality will cost more. We scope and price every project individually during the Discovery phase, so you will know the full investment before committing to a build.

The Discovery phase is available to any Enterprise Ireland client company with 10 or more employees. LEO offers a similar programme for smaller businesses. The build phase was funded through Enterprise Ireland's Digital Process Innovation programme, covering up to 50% of implementation costs. Eligibility depends on Enterprise Ireland client status and project fit. We confirm the best route during Discovery.

The Discovery stands alone. There is no obligation to proceed to a build. At the end of it you receive a full AI readiness report, a prioritised roadmap of opportunities with estimated ROI, and a clear recommendation on whether to proceed and, if so, exactly what to build first.

We provide the technical inputs and project description required for the grant applications and keep the business case in plain English.

What Happened Next

The system is live and processing quotes daily. We provide retained support and monitoring. When parts pricing shifts or the company takes on a new type of contract, we adjust the system.

They are now exploring Phase 2: reporting on job profitability trends and predictive scheduling for planned maintenance renewals. As new jobs are completed and costed, the system updates its baseline patterns from the company's own data only.