Interior Space Visualisation Agent

Interior Space Visualisation Agent

Business Challenge

Business Challenge

A sample flat is one interpretation of a space, executed once, in one design language, for one imagined buyer. Every prospect who does not see themselves in it is being asked to look past it.

Sales teams handle this verbally. The buyer asks whether the living room would work in a darker palette, or with a study rather than a formal dining area, and the answer is a description. The conversation stays abstract at exactly the point it should become concrete.

What We Built

What We Built

A visualisation agent that generates furnished interiors of the same space across design languages, palettes, material choices and usage scenarios, so a sales conversation can respond visually to what the buyer actually says they want.

How It Works

How It Works

Model and surface
Claude for scene specification, paired with image generation. Delivered on a tablet-usable interface for the sales team to run inside a live conversation, and to the marketing team for campaign asset production.

Connectors
The approved floor plans and, where available, the spatial specifications already produced by the render agent, so the two systems share a dimensional basis. The specification schedule for what is actually delivered. The approved material palette.

Skills
The design language library is encoded: each style the client offers or references, with its defining material, palette, furniture and lighting characteristics, so a style request produces a coherent result rather than a generic one. Spatial rules are encoded separately, furniture scale against real room dimensions, circulation clearances, and the fixed constraints of the shell that no amount of styling may move.

A separate skill enforces the delivery boundary: which depicted elements are part of the handover specification and which are buyer-fitted, and the disclosure that must accompany the latter.

Data flow
The salesperson states the space and the buyer's stated preference. The agent generates the interior against the spatial specification and the requested design language, holding the shell fixed and varying only what is genuinely variable.

Human review
Standard variants are pre-approved. Anything generated live in a customer conversation carries the fitted-versus-delivered disclosure, and the sales team is trained on that boundary rather than left to manage it by instinct.

Governance
The shell never changes. Buyer-fitted elements are disclosed as such, every time, because a visualisation shown in a sales conversation carries representational weight regardless of what a brochure disclaimer says.

One Cycle End To End

One Cycle End To End

A buyer in a site office says the living space feels formal and asks what it would look like as something warmer, with the second bedroom used as a study.

Within the conversation, the salesperson generates both: the same room in a warmer material language, and the second bedroom laid out as a study. The buyer is now looking at their own stated preference in their own prospective home rather than being asked to imagine it.

Outcome

Outcome

1,296 hours saved per year

Measured on a live deployment: 1,296 hours saved per year on interior visualisation production, from 27 hours per week. The measurable saving is production time. The material effect is on the sales conversation, which moves from a fixed image and a verbal description to a visual response to what the buyer actually asked for.

The Cutting Edge School

Enquire for Custom AI Trainings

We design and deliver practical, enterprise-ready AI training that turns everyday teams into high-output, AI-powered operators.

The Cutting Edge School

Enquire for Custom AI Trainings

We design and deliver practical, enterprise-ready AI training that turns everyday teams into high-output, AI-powered operators.

The Cutting Edge School

Enquire for Custom AI Trainings

We design and deliver practical, enterprise-ready AI training that turns everyday teams into high-output, AI-powered operators.