The Head of MEP Job Nobody Needs to Fill Anymore

The Head of MEP Job Nobody Needs to Fill Anymore

A European industrial conglomerate — the kind with over 300,000 employees, its own corporate real estate arm, and a habit of calling its facilities "workplace environments" — is hiring a Head of Mechanical, Electrical & Plumbing. Part-time. The posting wants someone to own building services engineering across office and industrial sites, shepherd MEP concepts from design through handover, dream up "energy concepts," and hold the budget for technical building systems.

Read that again. Part-time. They are already telling you the job is not full-time-human-shaped anymore.

Here is how an AI system — a proper pipeline of models, solvers, and digital-twin agents — takes this seat.

What the role actually does

A Head of MEP is a translation layer. On one side: architects drawing pretty boxes, facility managers who want zero downtime, finance directors who want the capex under a ceiling, and sustainability officers chasing a certification. On the other side: duct sizes, chiller capacities, cable tray routes, load calculations, and contractor change orders. The human in this chair spends most of their time converting between those languages, checking other people's conversions, and signing off on compromises.

That is a review-and-reconciliation job with an engineering license on top. Which is exactly the shape of work that automates violently.

What automates first: the conceptual and design phases

The JD stresses "planning of MEP conceptual and design phases." This is where AI eats first.

Feed a building information model into a generative design loop. An LLM reads the architectural brief — square meters, occupancy profiles, climate zone, use class — and emits a structured MEP requirements document. That document is passed to a parametric solver (the kind already used in advanced engineering consultancies) that generates dozens of heating, ventilation, electrical, and plumbing layouts, each optimized against a weighted score: energy use intensity, first cost, maintenance access, flexibility for future reconfiguration.

The "Head" prompt then evaluates the top candidates. Not by looking at drawings, but by querying the model: Which option keeps electrical riser paths clear of structural transfers? Which one lets us swap a floor from open plan to lab use without re-running chilled water? What is the delta in 20-year operating cost between option 3 and option 7?

The human used to do this by sitting in a room with three engineers, a structural consultant, and a cost manager, arguing for six hours. The model does it in minutes and produces an audit trail explaining every trade-off. The "energy concepts" the JD asks for? That is a literature-aware LLM pulling current practices for heat recovery, geothermal integration, and demand-response control, then mapping them onto the specific building's load profile. It is not innovation. It is retrieval with geometry.

Budget and cost monitoring: the part that is already software

The JD wants someone who "establishes a budget for the cost categories of technical building systems and shares responsibility for cost monitoring." Construction cost tracking is a spreadsheet problem that grew teeth. An AI agent connected to the cost database and the BIM model can:

  • Pull quantity takeoffs directly from the model (duct length, cable runs, fixture counts).
  • Cross-reference against a live material price index.
  • Flag variance the moment a contractor submits a change order with a unit rate that drifts from the baseline.
  • Generate the budget burn-down report every Friday without anyone asking.

The human Head's "responsibility" here was to glance at a dashboard, frown, and call the project manager. The dashboard can now frown on its own and send the message.

The handover and operation phase

"Cost-effective operation of the buildings" is where it gets fun. The digital twin — the same model used for design — becomes the living as-built. Sensors feed it. The AI monitors HVAC setpoints against occupancy and weather, recommissions systems continuously, and predicts chiller failure before the maintenance contractor notices.

The Head of MEP used to walk a building at handover, checking that valves were labeled and the BMS was not lying. A computer-vision model on a drone does the walkthrough. An LLM reads the O&M manuals, ingests the BMS trend data, and writes the handover report. The human signature on the handover certificate becomes the last purely ceremonial act.

Where it breaks

Here is where the human clings on, and you deserve the honesty.

Code compliance and liability. European building codes are not clean machine-readable text. They are layered: national regulations, local amendments, fire marshal interpretations, and insurance requirements that contradict each other and get resolved through conversation. An LLM can draft a compliance narrative, but the authority having jurisdiction still wants a human with a stamp to look them in the eye. That stamp is a legal artifact. It will outlive the job it certifies.

Contractor management. MEP contractors lie. They lie about delivery dates, about what is in the wall, about why the chiller is three weeks late. An AI can detect the lie by comparing the BIM model to the progress photos and the procurement tracker. But resolving the lie — pressing a supplier, renegotiating, deciding whether to accept a substitute pump — requires a relationship and a threat. Models do relationships badly and threats worse.

The physical world is dirty. Models assume the duct goes where the drawing says. The duct goes where the installer could fit it, which is nowhere on the drawing. Someone has to catch that. Right now that someone is human. Computer vision on site photos is closing the gap fast, but "closing fast" and "closed" are different sentences in construction.

Politics. A Head of MEP at a large organization spends a surprising amount of time defending their budget against the IT director, who wants to spend it on racks, and the sustainability officer, who wants to spend it on solar. AI can model the trade-off. It cannot win the meeting.

So who stays?

One engineer. Not a Head — a custodian. Someone to carry the stamp, attend the jurisdictional meetings, shake down the contractor, and walk the site on the day the ceiling gets closed up. The rest — the design iteration, the energy concepts, the cost monitoring, the compliance drafting, the handover reporting — is a pipeline.

A job posting that asks for a part-time Head of MEP at a 300,000-person organization is a transition document. They are hiring the last human who will hold this chair while the system is built underneath them. The job description even admits it: they want "digital solutions" and "technologies." They are asking the human to build their own replacement and calling it innovation.

The real question is why a company this size thinks the transition only needs someone part-time.