Case study · MV Electrical

An M&E contractor's operations layer, rebuilt around AI.

MV Electrical is a UK mechanical and electrical contractor: solar PV, EV charging, commercial fit-out, BS 7671 compliance, across London and the Home Counties. Multi-million turnover, a lean team, owner still close to the work. Growth was throttled not by demand but by the admin tail behind every job. Here is what we rebuilt, and what changed.

Sector
M&E contractor
Turnover
Multi-million
Team
10 to 20 staff
Engagement
Consultant, then embedded
The starting point

The same volume of work, all of it leaning on one or two people.

Before the rebuild, the familiar trade-firm pattern. Project-management tools that each handle one workflow but never fit the shape of a real construction job, where multiple sites, multiple suppliers and multiple jobs all move at once. A patchwork between those tools and the supplier portals everyone has to use (MCS, DNO, warranty registrations, accreditation systems). And the quiet tax of stale knowledge: tools touched once a fortnight, so every return started with ten minutes of remembering how it worked.

The symptoms were the ones every owner-operator knows. Tenders going out at 2am. Saturdays in the office to keep things moving. O&M manuals taking a week of office time and still coming out nowhere near as polished as they should be.

What we rebuilt

Documents drafted from the live job, not from a form.

Over months, working first as a consultant and now embedded as Head of Digital Operations, the operations layer was rewired across the surfaces below. We do not replace the supplier portals. MCS, DNO and warranty registrations stay external; AI helps the firm navigate them faster. What we replaced was the project-management patchwork, and we bridged the messy reality of a live job into the documents that have to leave the building.

LayerBeforeAfter
Solar PV pipeline The same data re-entered across DNO submissions, MCS registrations, warranty registrations and handover packs. One missing document could hold up grid connection by weeks. Data lands once, at proposal stage and on the on-site capture. The downstream documents generate from those two sources. The admin tail of human work drops from weeks to days.
Project documentation Project data spread across boards, Drive folders, email threads and WhatsApp screenshots. No single place where the job actually lived. One workspace where the data, templates and AI tooling sit together. The same data flows into costing, customer records, payment applications and certificates without being re-typed.
O&M manuals A week of office time per manual. Eighty-plus pages assembled by hand, and still not as polished as they should be. The email thread, the spec, the call transcripts and the folder paths feed the system. The manual evolves across the job, every substitution and scope change reflected. Human work: a day or two of review, spread thin.
Quality-assurance packs Setting up an in-site form per project took days. Engineers struggled with it on the ground. Exports took days more. Output did not match the firm's brand. A form on site that is AI-authored for the specific job and pre-filled, so the engineer reviews and confirms rather than fills from scratch. Bulk corrections handled across plots. Output in the firm's brand.
Payment applications and variations Hand-drafted at month-end across multiple jobs. Variations slipping through unbilled. Applications going out at 11pm on the last Friday. Drafted from the live job data each month. Variations caught and evidenced, not lost to next quarter. Applications in on time, in full.
Document retrieval Hundreds of files scattered across folders. "What was the model number on that boiler at the school in 2023?" cost twenty minutes of hunting. Every file indexed with description, type and tags. A twenty-minute hunt becomes a sentence to the system and a couple of minutes to read the answer.
What changed

The MD stopped working Saturdays and started leading the business.

~50% → 0
The MD's week went from roughly half admin to none. He approves the work; he no longer authors it, not even an email.
Weeks → days
The admin tail behind solar handovers and O&M manuals, compressed. Documents finished the week the job finishes.
On time, in full
Payment applications in before the cut-off, variations evidenced. The firm reads as sharp to its main-contractor clients.
  • The same volume of work moves through fewer hands.
  • Engineers spend less time on admin and more on the tools, the work the firm actually charges for.
  • The office person does higher-value work, project coordination and supplier management, instead of typing certs and hunting for files.
  • The MD no longer authors anything. The system drafts; he approves. More throughput, better quality, and his week back for customers, tenders and growth.
The discipline behind the speed

Organise the chaos first. Then AI is exponential.

Truth oneAI on top of organised data is exponential.
Truth twoAI is also more than capable of taking chaotic data and making it organised.

The reconciliation is the sequence. The first thing we do is organise the chaos. After that, every downstream task gets exponentially easier. That is why this took months rather than weeks, and also why it did not take years.

The misconception we meet most often: "we need to clean our data before we can use AI." No. AI does the cleaning. What you need is an operator willing to sit inside the business while it happens. Drafts go to a single approver, and nothing leaves the building autonomously.

If you already have a QMS

The rebuild does not replace your quality system. It makes it actually used.

Most M&E firms of this size keep a Quality Management System because the accreditor (NICEIC, NAPIT, ISO) requires one. It is built for compliance audits, not for running the business day to day. The folders get updated when an inspection looms; the rest of the year they go stale.

What we do is not a parallel system. It is the engine that keeps the QMS structures populated as the work happens. The accreditor sees a system that is current; the firm gets the operations layer the QMS template implies but never delivers on its own. The framework you already pay to maintain starts paying you back.

Next step

Want this for your firm? Tell me what's in the way.

Thirty minutes, direct. We ask about the business; you decide if we're the people for the job. No deck.

Draft for sign-off. Off-nav until the named firm has approved publication.