# AI for interior designers: what it's good for and where it falls short

By Arnaud Doucet. Published August 27, 2026.

## Summary

- AI is good at reading and rewriting. That covers drafting emails, summarising meetings, pulling details off a product page and sorting a library.
- It's unreliable at facts it hasn't read. Asked for a product, price or lead time from memory, it produces plausible details that may not exist.
- Give it the source, such as the product page, the quote or the email, and check what it extracts before the result reaches a client or an order.
- A chat assistant only knows what you paste in. One connected to your tools can read and update your real records, which is more useful and needs more care.

AI tools for designers mostly do one of two things: make images, or read and write text. Both can save time, and both can put something wrong in front of a client if nobody checks. The difference between a useful tool and a risky one is usually whether it's working from a real source or from memory.

## What can AI do for an interior designer today?

AI is reliable at reading, summarising and rewriting text you give it, and useful for quick concept images. It's unreliable at facts it hasn't been shown: specific products, prices, dimensions, availability and regulations.

In a design practice, that splits the work roughly like this:

| Task | Good fit? | What to check |
| --- | --- | --- |
| Drafting client emails and proposals | Yes | Tone, and every fact, date and price |
| Summarising meeting notes into decisions and actions | Yes | That nothing agreed was left out |
| Reading a product page into name, size, finish and price | Yes, from the page itself | Price, dimensions and the selected option |
| Sorting or tagging a product library | Yes | Items with brief or ambiguous descriptions |
| Reading supplier emails for dates and references | Yes, with review | Dates, quantities and which project it's for |
| Concept images and quick renders | For direction | That nobody mistakes them for specified pieces |
| Suggesting specific products from memory | No | Usually invents products, sizes or links |
| Building codes, fire ratings and compliance | No | Use the actual standard or a qualified person |
| Current prices and lead times | Only from the source | The supplier's page, price list or quote |

## Why does AI make up product details?

Language models write the most plausible next words, not the true ones. Asked for "a 220 cm oak dining table from a Danish brand", a model with no source in front of it produces a description that sounds right, sometimes with a model name, a price and a link that don't exist.

The fix is to give it the source. A model reading the actual product page or the supplier's price list is summarising real information, which it does well. Keep the source link on every product, so anyone can check a line against the original in one click.

## How should I check AI output before a client sees it?

Check anything that affects money, measurements or a decision, and compare it with the source rather than reading it for plausibility.

1. **Prices:** confirm against the supplier's page or quote, and check whether it picked up a retail, sale or member price.
2. **Dimensions and options:** make sure they match the size, finish and fabric you're specifying, not the default shown on the page.
3. **Names and links:** open the link. If it doesn't resolve, the product details may not be real.
4. **Dates and quantities** read from emails: compare with the original message.
5. **Anything sent to a client:** read it as if you wrote it, because to the client, you did.

## What's the difference between a chat assistant and one connected to my tools?

A chat assistant knows only what you paste into the conversation, and anything it produces has to be copied back into wherever your project lives. An assistant connected to your tools can read your actual records and make changes for you, with your permissions.

Connected assistants, such as Claude with a connector added, are more useful because they start from your real data. "Which products on the Smith Residence schedule still have no supplier?" gets an answer from the schedule itself, not a guess. They also need more care, because they can change and delete things.

When you connect one:

- **Check what it can access.** It should see and change only what your own account can.
- **Ask it to describe a change before making it,** especially anything that deletes or overwrites.
- **Review its work** the way you'd review a new assistant's, until you trust the kind of task you've given it.

Inside Outside connects to Claude, so you can ask about and update your projects, products, schedules and orders in plain language. Claude works with your permissions, and you can disconnect it at any time. See [Connect Inside Outside to Claude](https://docs.insideoutside.com/getting-started/connect-claude/).

## Where does AI already save time in day-to-day work?

The biggest savings come from turning information you already have into structured records, rather than from generating anything new. Copying a product page into a schedule, reading a supplier email for a delivery date, or sorting two hundred products into categories are all transcription. AI is good at transcription.

In Inside Outside, AI reads a product page when you add a product from its URL, suggests categories for library products, and reads forwarded supplier emails into suggested updates. In each case you review the result before it's saved. See [Add a product from a URL](https://docs.insideoutside.com/schedules/add-from-url/) and [Categorize products automatically](https://docs.insideoutside.com/library/auto-categorize/).

## Will AI change what clients expect?

It already has for some clients, who arrive with AI-generated room images and expect quick visual options. Those images are a useful starting point for a conversation about direction, but the pieces in them usually don't exist.

Treat a client's AI image like a page torn from a magazine: a reference for mood, scale and palette. Your work is turning it into real products with real prices and lead times, and explaining where the real versions differ.

## How do I start using AI in my studio?

Pick one repetitive task, try a tool on it for a few weeks, and write down what you check each time.

The [Studio Field Guide](/field-guide) has free AI skills you can try for tasks such as sourcing furniture, building a room schedule and preparing a client proposal.

1. **Choose a task that's mostly transcription,** such as meeting summaries, product data entry or first drafts of routine emails.
2. **Give the tool the source** every time, rather than asking it to recall facts.
3. **Write a short checklist** of what you verify, and keep using it.
4. **Check the tool's data terms** before using client information.
5. **Expand only once the first task is routine,** and keep anything client-facing under your own review.

## Frequently asked questions

### Can ChatGPT or Claude make an FF&E schedule?

They can lay out a schedule's structure, and fill it accurately from product pages or a supplier list you provide. Asked to suggest products from memory, they tend to invent names, sizes, prices and links, so every line needs checking against the real source.

### Should I tell clients when I use AI?

There's no single rule. Be clear whenever it changes what the client is looking at, for example when a concept image shows furniture that doesn't exist, and follow anything your contract says about how you work.

### Are AI renders useful in client presentations?

They're useful for showing a direction early on. Label them as concept images, because the pieces in them usually can't be bought, priced or specified.

### Is it safe to put client details into an AI tool?

Check the tool's data terms first, including whether it stores your inputs and whether they're used for training. Share only what the task needs, and keep sensitive personal details out unless the terms suit your obligations to the client.
