A 4-inch tile becomes 5 or 6 inches in the rendering. The room still looks beautiful. A client could easily miss the discrepancy, or catch it and begin wondering which other details are wrong.
Kristi Carré Freeland, founder of rendering studio Carré Designs, used that example in a July 2026 interview with Business of Home. She has embraced AI for conceptual design and remains careful about what reaches a client. Her reason is simple: a polished image can carry small inaccuracies, and the client may read the whole thing as a promise.
This is the complication inside faster concept work. AI can give a designer more visual directions to test before a presentation. It can also produce a first draft with the finish of a final one. The creative work becomes easier to begin, while the conversation about what has actually been decided becomes harder.
Realism gives an image authority
Clients rarely read a rendering as a designer does. A designer can look past an invented chair leg or an uncertain junction and focus on proportion, atmosphere, or palette. The client sees her future room.
Photorealism strengthens that response. Tile has grout, lamps throw light, and fabric folds over the sofa. The image feels resolved because it contains the visual cues of a finished photograph, even when several parts came from an exploratory prompt.
The more convincing the image, the more deliberate the designer has to be about its role. A rough study can support a broad reaction: warmer or cooler, quieter or more layered, open or enclosed. When the room looks realistic, the client asks about the exact sofa, the precise wall finish, and whether the light will fall that way in the afternoon.
Those are different client conversations. Sending the second image while hoping for the first response asks the client to ignore the information the image appears to provide.
A client reads the finish of a rendering as a sign of how settled the room is.
Accuracy depends on the information supplied
Carré Freeland asks designers to brief her team as if they were briefing the builder. A floor plan helps establish the room. Elevations add more information. Finish schedules, furniture details, dimensions, and custom specifications build the next layers.
That process explains why a plausible image can still be wrong. A style direction gets the rendering started; accuracy comes from the decisions that govern scale, transitions, clearances, material behavior, and construction.
AI concepting often begins earlier, when many of those decisions remain open. That is part of its appeal. A designer can test an atmosphere before documenting the room. Trouble begins when the image moves into a client presentation without a change in the accuracy standard.
The designer then has to inspect every visible claim the client might rely on. Does the room geometry match the survey? Are openings in the right place? Does a repeated material use the specified scale? Is the furniture drawn to dimensions that fit? Which products exist, and which are visual stand-ins?
The image may still be valuable if some answers remain open. Its purpose has to match that uncertainty.
Keep exploratory abundance backstage
Fast generation can encourage a designer to show more. If six directions took less time to produce than one traditional rendering, putting all six into the deck can feel generous.
It also transfers the designer’s unresolved work to the client. The client now has to compare rooms with different architecture, furniture, lighting, and finish details while guessing which differences matter. A conversation about one design decision becomes a referendum on six complete-looking interiors.
Backstage, the abundance is useful. The designer can test whether the palette survives a darker room, whether the composition needs more weight, or whether a material direction has enough range. In front of the client, show the image that advances the decision she is ready to make.
That may mean cropping to the relevant area, reducing the finish of the image, or presenting a material study beside it. It may mean holding the rendering until the room has enough documented information to support close reading. Choose a format that matches the stage of work.
Name what the client is approving
“Concept only” is easy to overlook under a persuasive image. A useful presentation states the decision in the language of the project.
The designer might be asking the client to approve the room’s overall direction, the relationship between two materials, or the placement of major pieces. State that before the image appears. Then identify any visible element that is serving as a stand-in and is likely to attract attention.
Good client communication depends on judgment about where the image looks more certain than the design is. If the generated window arrangement differs from the survey, the image needs correction before the client sees it. If a side table is present only to test visual balance, the designer can say that the exact piece will follow.
Carré Freeland’s team uses renderings as part of a revision process. Designers see that lounge chairs are too large or that a fabric fails in the room, then change the design. An AI rendering can support the same review when the designer remains in charge of what the client is invited to conclude.
Add an accuracy gate before the presentation
Use the speed at the beginning of the process. Client-facing review still needs time.
Compare the rendering with the source material the studio has already approved. Check architecture first because it governs the room. Then inspect scale, specified materials, major furnishings, and details that carry cost or client attention. Remove visual inventions that could become accidental commitments.
The final check is conversational. Ask what the client is likely to point at first and what answer the designer will give. If the truthful answer is “ignore that part,” the image probably needs another edit or a different place in the presentation.
That gate can feel disproportionately expensive after an image appeared so quickly. It is also the step that makes an attractive image safe to present. Carré Freeland’s warning about tile scale matters because the error is small enough to seem harmless and visible enough to damage the client’s faith in the rest.
Let the first draft stay a first draft
AI is useful in the loose part of design, when a studio is trying directions and looking for a stronger question. A client presentation needs to make the next decision easier while staying honest about the details still under review.
The designer can keep those jobs separate. Generate freely, edit hard, and choose the image according to the conversation it needs to support. A concept can remain provisional even when it looks beautiful. The designer should explain that status herself.
