AI for interior design: more than a render

If you use an image generator to show a client an idea, you've already solved the easy part. The hard part begins when that person points at the sofa in the image and asks how much it costs, who sells it and how long it takes to arrive.

That leap —from an approved render to a proposal someone can quote and buy— is today the real bottleneck of AI for interior design. It's what decides whether the technology saves you time or just moves it somewhere else. It's also the problem I founded Kouzee to solve. Here I go through what the data says about where AI is working in architecture and interior design studios, where it still fails, and what a proposal needs in order to go from image to executable project.

Where AI is actually helping

It's helping mostly to produce ideas faster, and much less to improve design decisions. That distinction is the most useful one for deciding where to invest your time.

The RIBA AI Report 2026, from the Royal Institute of British Architects, shows adoption is no longer marginal: UK practices using AI went from 59% in 2025 to 74% in 2026. 75% report productivity gains. But only 17% —fewer than one in six— agree their designs are better thanks to AI, and 77% hold that it can never replace human creativity.

That study measures the British market, not the Latin American one. Still, the pattern is recognizable in every studio I know: more production speed, same quality of judgment. In the US, the Houzz report on AI in construction and design tells a similar story from another angle: firms that use it estimate a productivity gain of about US$74,000 a year, almost all of it in administrative and production tasks, not design.

A second source helps locate where that saving happens. Chaos and Architizer surveyed close to 800 professionals —70% architecture, 25% interior design— for their AI in Architecture Report 2026. Among those already using AI, 86% report some time saving and 59% save at least five hours a week. Broken down by stage, according to the analysis they published afterwards, 48% perceive relevant savings in concept design and ideation, while in material selection the figure drops to around 25%.

Almost twice the savings in imagining as in specifying. That's the problem.

Why an approved render is still not a project

Because a render describes an aesthetic intention, not a set of purchasing decisions. And the client approves the intention believing they approved the decisions.

I lived this scene many times as an architect and I've seen it in every studio I've worked with: three images are presented, the client picks one, and only then does the work nobody charges for begin. You have to identify exactly what that pendant lamp is, find an equivalent sold in the country, ask for a price, confirm stock, check delivery times, adjust the budget and —if something doesn't fit— present again. That cycle can repeat several times before the first purchase order.

The cost isn't just hours. It's credibility. When the visual proposal promises a level the real budget can't sustain, the conversation with the client turns into a downward negotiation, and the designer ends up defending cuts instead of defending judgment.

That's why I find it useful to separate two distinct objects that usually get confused:

Neither replaces the other. The mistake is presenting the first and charging as if it were the second. And there's a technical reason the objects in a render have no origin: an image generator invents every pixel anew. I explain it in how AI image generation works.

What a proposal needs to be executable

A proposal is executable when the client can approve it and buy it without you having to rebuild it from scratch. In practice, that demands five conditions:

1. Product traceability. Every relevant piece must be identifiable: brand, model or reference and supplier. A “similar to” object is useless for quoting.
2. Dated price. A value without a date ages in weeks. Stating when it was checked protects the studio when the market moves.
3. Real availability. A discontinued product or one with a twelve-week import changes the project, not just the spreadsheet.
4. Equivalent alternatives. At least one replacement option per critical item keeps a stockout from halting the job.
5. A deliverable the client understands. The proposal must come out in a format that can be reviewed, commented and archived: a presentation, a spreadsheet or a document. Not a link that only works inside a tool.

How to bring AI into the workflow without redoing it all

My recommendation is not to change the whole process, but to intervene at the two points where the most time is lost today: initial exploration and assembling the proposal. An order that has worked for me:

1. Define the brief before generating anything. Program, budget cap, deadlines and constraints of the space. Without that, AI produces pretty images and impossible decisions.
2. Use image generation only to explore direction. It's the stage where the data shows the biggest savings. Treat them as sketches: they serve to agree on a course, not to commit to a result.
3. Ground every image in real products as early as possible. The later this step happens, the more expensive it is to correct. Ideally, before the first formal presentation.
4. Present image and list together. When the client sees the atmosphere and its cost in the same sitting, the hard conversations happen early and cost less.
5. Keep a record of what was approved. Version, date and scope. It's what lets you charge for a scope change without it turning into a conflict.

Why I built Kouzee

This is the problem that gave rise to Kouzee. I founded it with Samuel Larraín, business engineer, and Alejandro Sandoval, software engineer and CTO; I come from architecture, and this was exactly the part of the problem that hurt me most. In June 2026 the entrepreneurship ecosystem of Universidad del Desarrollo published a piece about the project, and Samuel summed up the focus better than I could:

Kouzee lets architects, interior designers and designers create proposals quickly using real products, with real prices and stock.

Samuel Larraín, Kouzee co-founder, in Emprendimiento UDD (translated)

The same piece covers the work with the Chilean Association of Interior Designers (AdDI), our time in Acelera UDD, mentoring by David Basulto —founder of ArchDaily— and Corfo's Semilla Expande award. In concrete terms, what I'm after is for the design board and the shopping list to be the same object, and for that object to leave the platform in the formats the client already uses. How we got there, and what was hardest, I tell in what I learned building Kou.

Questions I always get

Does AI replace the interior designer?

The data doesn't point that way, but it doesn't rule out effects on employment either. In the RIBA AI Report 2026, 77% hold that AI can never replace human creativity; at the same time, 59% believe its adoption will lead to staff reductions in the profession. Both coexist: what gets automated first is production, not judgment.

Are AI renders good for presenting to a client?

They're good for agreeing on an aesthetic direction, as long as they're presented as such. The risk appears when the client reads them as a promise of outcome. A simple practice: accompany every generated image with the list of products behind it, or state explicitly that it's still exploratory.

What happens to prices and stock if the project drags on?

They change, and that's exactly why you date every value. In projects that stretch over several months, revalidate prices and availability before each purchasing milestone, and put in writing that the references correspond to the proposal's date.

Where to start this week

Take the last project you presented and count how many items in that proposal you could quote today without opening the browser to search again. That number is your real starting point: it measures how much of your creative work arrives converted into decisions and how much stays as an image.

AI will keep improving the speed at which you produce ideas. The competitive edge, though, is moving toward those who get those ideas to the client with price, origin and lead time.