The ELV Designer Start a project
  1. Home
  2. Services
  3. AI-ready design

Layer 06 · AI

AI-ready ELV design

AI is not a separate system you buy at the end. It runs on the cameras, sensors, controllers and network of the other five layers, so it has to be planned with them. We design every layer so it can use AI now, or adopt it later without opening walls.

  • Video analytics
  • Occupancy and daylight
  • Fault prediction
  • Voice and meeting rooms
  • Privacy by design
Plan a project for AI
A glass-walled meeting room at dusk, with a conferencing camera bar under a wall display, a ceiling microphone above a long timber table and a linear pendant light.
Fig. 01 Meeting room, camera bar and ceiling microphoneIllustrative visualisation

01Planned, not added

AI depends on decisions made at the first drawing, not on what is bought at the end.

AI is moving into the systems we design: cameras that understand what they see, lighting and air conditioning that learn how a space is used, building systems that warn before equipment fails, and meeting rooms that frame and hear people on their own.

All of it depends on decisions made early. Where a camera sits decides what it can recognise. Sensor density decides what a room can learn. The network decides how much video and data can be processed, and open protocols decide whether systems can share what they know. We make those decisions with the rest of the design, so the space is ready whether AI is switched on at handover or years later.

02Where the layers converge

What each layer gives AI

AI works with what the other five layers capture. Choose what the space should do to see which sensors feed it, how the data travels and which layer acts on the result.

Sources · the five layers 05 · Connected 06 · AI · processed on site Actions · back to the layers Backbone 01 · Light Daylight sensor Occupancy sensor Dimmable drivers, energy data 02 · Media Ceiling microphone array Meeting room camera Voice control points 03 · Secure Cameras at the gate and drive Door and gate contacts Perimeter beams 04 · Control · BMS Pump vibration, current Chiller data Room temperature BMS Edge processing SeeVideo analytics LearnPatterns of use PredictTrends and faults ListenSpeech and sound Secure Alert to the owner: a person atthe gate after hours. A swayingtree in the same view is ignored. Light Control Lights dim as daylight rises, andcooling eases off in rooms thatnobody is using. Control Early warning: pump vibration isrising. The facility team servicesit before it fails. Media Control “Evening scene” by voice at home.In the meeting room the cameraframes whoever is speaking. Sources · the five layers 05 · Connected · backbone 01 · Light Daylight sensor Occupancy sensor Drivers, energy data 02 · Media Ceiling microphones Meeting camera Voice control points 03 · Secure Gate cameras Door contacts Perimeter beams 04 · Control · BMS Pump vibration Chiller data Room temperature 06 · AI Edge processing SeeVideo analytics LearnPatterns of use PredictTrends, faults ListenSpeech, sound Secure Alert: a person at the gate afterhours. A swaying tree is ignored. LightControl Lights dim as daylight rises;cooling eases off in empty rooms. Control Early warning: pump vibration isrising. Serviced before it fails. MediaControl A scene set by voice. The cameraframes whoever is speaking.
Schematic S-01 What each layer gives AI Functions See, learn, predict, listen Illustrative Not to scale

See. The gate cameras, door contacts and perimeter beams of the security layer send video and events over the network. Analytics tell a visitor from an intruder and a vehicle from a passer-by, and raise an alert only when it matters. It works because the cameras were placed for it.

See
Video analytics that tell a visitor from an intruder and a vehicle from a passer-by, with alerts only when it matters.
Learn
Occupancy and daylight patterns that drive lighting and climate, saving energy without anyone touching a switch.
Predict
Data from the building management system that flags pumps, chillers and other plant before they fail.
Listen
Voice control at home, and meeting rooms that pick up the right person and remove the noise.
Protect
Privacy planned in: what is captured, where it is processed and who can see it, decided before any camera goes up.
Choose freely
AI features are often locked to one brand. We compare them on what they do for your space, not whose logo is on the box.

03Decided at the first drawing

Where a camera sits decides what it can recognise

Most of what AI can do is settled long before anyone chooses an AI feature: by where things are mounted, how many there are, how they are connected and whether they can talk to each other. The same camera, drawn in two places:

Too high, too steep

Section · gate approach Steep Mounted high Camera view CAM 01 TOP OF A HEAD

From high on the wall the camera looks down on the top of a head. No analytics can recognise a face the camera never sees.

Planned height and angle

Section · gate approach Shallow Planned height Camera view FACE IN VIEW CAM 01

Mounted at a planned height and aimed along the approach, the same camera sees faces at a shallow angle. Its conduit and bracket were drawn into the wall before it was built.

Section C–C Gate approach, two camera positions Height and angle Decided with the facade Illustrative Not to scale

  • Camera position

    Where a camera sits decides what it can recognise. Height, angle and distance are fixed the day its conduit and bracket are drawn.

  • Sensor density

    Sensor density decides what a room can learn. One sensor can say a room is in use; several can say where, and for how long.

  • Network capacity

    The network decides how much video and data can be processed: bandwidth, power over the cable, storage and processing, with room to grow.

  • Open protocols

    Open protocols decide whether systems can share what they know, so a pattern found by one layer can act on another.

04Protect

Privacy, planned before any camera goes up

Three questions are answered on the drawings, not after installation: what is captured, where it is processed, and who can see it. Step through one camera, end to end.

  • Masked. Parts of a view that are never recorded, such as a neighbour’s window.
  • On site. Analysis and recording on equipment inside the building.
  • By role. Each person sees only what their role needs.
Privacy plan · one camera, end to end Site boundary Masked 1 · Captured Only what the camera is meant to see 2 · Processed on site Analysed and kept here Video stayson site Remote service(optional) Alerts and summaries only 3 · Seen by role Owner Live videoRecordings AlertsPlant data Security desk Live videoRecordings AlertsPlant data Facility team Live videoRecordings AlertsPlant data Privacy plan · one camera Site boundary Masked 1 · Captured Only what the camera is meant to see 2 · Processed on site Analysed andkept here Video stayson site Remote service(optional) Alerts andsummaries only 3 · Seen by role Owner Live videoRecordingsAlertsPlant data Security desk Live videoRecordingsAlertsPlant data Facility team Live videoRecordingsAlertsPlant data
Diagram P-01 Privacy plan, one camera Captured · processed · seen By design Illustrative Not to scale

Captured. Each camera is aimed and masked so it records the gate and the drive, not the neighbour’s window. What is not needed is never captured.

05What you receive

Readiness you can see on the drawings

AI readiness is documented on the same base drawings as the other five layers, so a camera position, a sensor or a spare cable is agreed once and built once.

  • Sheet AI-01

    Camera view study

    Each camera’s position, height and field of view, checked against what it must recognise.

  • Sheet AI-02

    Sensor layout

    Occupancy, daylight and plant sensors placed at the density each room needs to learn.

  • Sheet AI-03

    Network, storage and processing

    Bandwidth, recording and processing sized for today, with marked space for later.

  • Schedule AI-04

    Integration schedule

    Which systems share which data, and over which open protocol.

  • Plan AI-05

    Privacy and data plan

    Masked areas, where video is processed and kept, and who can see what.

06Across every layer

What AI asks of the other five layers

AI-ready design is not a sixth set of equipment. It is a set of decisions inside each of the other layers, made while they are being drawn.

  • Light

    Daylight and occupancy sensing, and circadian lighting that follows the time of day. The sensors, drivers and controls are chosen for it at the design stage.

    Lighting design

  • Media

    Meeting cameras that frame the speaker, voice pickup and noise suppression. Microphones and cameras are placed with the ceiling and the furniture.

    Audio and video

  • Secure

    Video analytics that recognise people and vehicles and cut false alarms. Camera heights, angles and night lighting are set for what each camera must recognise.

    Security

  • Control

    Systems that learn routines, save energy and flag faults before they fail. Plant is given its sensors, and the controls open protocols, from the start.

    Controls and BMS

  • Connected

    Bandwidth, storage and processing sized for the cameras and sensors AI depends on, with spare capacity marked for what is added later.

    Networking

07Where AI matters most

Different spaces, different intelligence

  • Homes

    Cameras that know family from strangers at the gate, lighting that learns the household’s routine, and alerts that stay quiet unless something is wrong.

  • Offices

    Cameras that frame whoever is speaking, noise suppression, space-use analytics and early alerts from the BMS.

  • Shops

    Footfall counts and heatmaps, queue alerts, and loss-prevention analytics on the cameras.

08Questions

What clients ask about AI-ready design

Do we have to buy AI features now?

No. We design every layer so it can use AI now, or adopt it later without opening walls. What matters at the design stage is the groundwork: where cameras and sensors sit, the cabling and power they need, the capacity of the network, and open protocols between systems.

Which AI products do you recommend?

None by name. AI features are often locked to one brand, so we describe what each feature must do for your space and compare the vendors’ proposals against that, on what they do rather than whose logo is on the box.

How is privacy handled?

It is planned with the rest of the design: what each camera captures and what it masks, whether video is processed on site or anywhere else, how long it is kept, and who can see what. Those decisions are made before any camera goes up.

Can an existing building be made AI-ready?

Often, yes. We start from what is already installed (camera positions, cabling, the network and how open the existing systems are) and tell you plainly what can be reused and what would need to change.

Is AI-ready design only for large buildings?

No. A home benefits too: a gate camera placed to see faces, sensors that let the lighting learn the household’s routine, and a network with room for more devices. Those decisions cost little at the design stage and a lot to change later.

Next step

The earlier we join, the more we can save.

Tell us about the space and where the design stands. We will come back to you to arrange a first conversation.

Start a project See how a project runs