We run the contracts-manager, estimator and document-controller work as a service — built from AI, owner-led. The MD approves the work; they don't author it. £2k–£6k a month, rolling monthly, no minimum term. Two new clients per quarter — that's the discipline.
Start where it hurts most: certs, tenders, or O&M manuals. One surface wired properly, with room to grow into the rest.
Two or three loops wired together. The middle-management layer as one service — contracts manager, estimator, document controller combined.
The full operations layer. Every place documents leave the building. Owner-led, with us holding the line monthly.
For context: £4k / month is roughly 5% of the cost of the four-role team (contracts manager, estimator, document controller, ops admin) that produces this work in most firms. One service, one operator, the AI doing the heavy lifting in the background.
What most firms call "too much admin" is three things stacked: rigid workflow software that doesn't fit how real jobs run; a patchwork of project-management tools (Monday, Airtable, Basecamp and similar) where no single one covers the process; and the stale-knowledge tax — even the tools that work get used every few weeks and need re-learning every time. AI removes all three. It absorbs the unstructured input (the late WhatsApp, the supplier substitution, the mid-job scope change), replaces the project-management patchwork, and carries the task context so there's nothing to re-learn. The documents that have to leave the building get drafted from the live mess, not a form.
The contracts manager exists to bridge that mess into structured outputs. AI is the bridge now.
A commercial install reaches handover. The client needs the O&M manual — eighty-plus pages, every datasheet current, every supplier substitution caught, every scope change reflected. Most firms have someone in the office assemble it by hand the week after the job's done.
Picture the brief landing the way these briefs actually land: a vague Teams call. A WhatsApp screenshot of a spec. The actual spec arriving two days late. A folder structure someone else started. Engineers who don't fill in forms. Supplier swaps and scope changes arriving across the project's natural arc.
The system absorbs all of it — email threads, call transcripts, folder paths, late specs. A task gets built with a plan. The AI holds the live mess as it lands, chasing missing data through short surveys, taking fragments as engineers and suppliers throw them in. The manual builds itself in parallel with the job, not after it.
The day the job finishes, the manual is finished too. Polished, current, accurate — every substitution and scope change in place. No post-job office-week of admin. The human work is a day or two of operator review, spread thinly across the months the job actually runs.
The work usually sits across four people — a contracts manager (~£55k), an estimator (~£45k), a document controller (~£30k), an operations admin (~£25k). With employer's NI, pension, holiday and sick cover stacked on top, the team comes in north of £200,000 a year.
That team running flat-out for a quarter produces roughly eighty deliverables. The AI produces several multiples of that. One person reviewing.
The offer is one thing. It shows up across different admin surfaces depending on which is hottest in your business. We start with the highest-pain loop, then expand.
BS 7671 EICs, PV commissioning certs, plot QA packs, MEIWCs. Engineer captures structured data once; the cert flows out. Faster cert means faster payment application.
AI reads the spec, pulls relevant past wins, drafts the pricing and the narrative in your voice. The MD signs off, doesn't write. Goes out at 4pm Thursday, not 2am Friday.
Generated from job data + product datasheets + the messy email trail, not assembled by hand. Eighty-plus pages drafted in an hour, every datasheet current.
Drafted from the live job data each month. Variations caught and paid — they don't slip through the net to next quarter. The application goes out before the cut-off.
AI chases, tracks, escalates in the firm's tone with the actual outstanding items. The pipeline holds without the MD remembering to ring round on a Friday.
Every file uploaded gets indexed by an LLM. "What was the model number on that boiler at the school in 2023?" — a 20-minute hunt becomes seconds.
The first applications wired depend on your hottest pain. The whole layer wires together over months two and three.
Sit inside the business. Map the actual work — not what's on paper. Identify the highest-pain admin loop.
Workspace built with your live jobs, templates, suppliers. First surface wired — usually certs or O&M. First documents go out under the new system.
Second and third surfaces wired. Your right hand learns to drive the daily review. The system starts holding more than the operator.
Monthly review, weekly check-in, urgent fires absorbed mid-cycle. The MD gets time back. The firm reads sharper to its clients than it ever has.
Honest exit, any month. Rolling monthly with 30 days' notice. If the work hasn't shifted what you do that day — in a way you can see and feel — you walk. We'd rather close a misfit cleanly than carry it.
Most M&E firms in our space already have a Quality Management System in place because the accreditor (NICEIC, NAPIT, ISO and so on) requires one. The QMS is designed for compliance audits, not for running the business day-to-day. The folders get updated when the inspection looms; for the rest of the year they go stale.
What we do isn't a parallel system — it's the engine that keeps the QMS structures populated as the work happens. The accreditor sees a QMS that's current. The firm gets the operations layer the QMS template implies but doesn't deliver on its own. The framework you already have starts paying back the cost of maintaining it.
Single monthly quote against the scope we agree in the first call — an indicative £100/hour of operator attention, in one of the three shapes above or somewhere between them. The AI does several multiples more in the background, so the monthly figure buys far more than the hours alone suggest. Rolling monthly, no minimum term — cancel any time with 30 days' notice. Switch shape between months without renegotiating. The figure is all-in: it covers operator time, your AI token budget (the AI's running costs) and your subscription to Forge, the platform we built ourselves and run our own back office on. No licences, no per-seat software, no usage bills on top.
No — they're example shapes at an indicative £100/hour to help you see roughly where the engagement is likely to sit. The discovery call lands on one number quoted to your specific scope. If the work needs to change shape later, the number changes with it. We send a one-page proposal with the final figure and the scope that justifies it.
Mostly remote — a few days of remote sessions with you and the team, with a site visit if it's the right call. We watch where time actually goes, identify the loop that's burning the hottest. Output: a prioritised hit list and a one-line agreed scope for month one.
Yes — that's what the £2k lower band is for. One focused admin loop. Rolling monthly: if it works, expand; if it doesn't, walk at any month-end with 30 days' notice. No minimum term to commit to.
No — not as the offer. If you need a £350 trade website we can quote that separately, but it's not what we run on retainer. For SEO or paid spend, we'll point you at people who do that properly.
Primarily UK M&E (mechanical and electrical) contractors and main contractors — £2m to £20m turnover, MD still close to the work. Adjacent trade contractors (solar, fit-out, building services) where the operational pain is shaped the same way.
Rolling monthly, 30 days' notice — walk at the end of any month. Honest exit. We'd rather you tell ten people we built something that didn't fit you than tell ten people we held you in a contract.
Two new clients per quarter, owner-led delivery. When the next slot is full, it's full — the wait list is real, not a sales tactic. The cap protects depth.
We ask about the business. You decide whether we're the people for the job.