It was nearly midnight in Adyar when Kavya pasted the client's brief into a chatbot. She was the design head at a twelve-person firm, the client wanted a quote for a 4BHK on OMR by morning, and she had been at it since 8.
The chatbot was fast. In under a minute it produced a tidy BOQ: rooms, items, quantities and, to her surprise, rates. Wardrobes at a neat figure per square foot. False ceiling, painting, a modular kitchen, all priced, all totalled. It looked like something a senior estimator would have made.
She almost sent it. Then she looked at the wardrobe rate and realised it was nothing like what her firm paid its factory, let alone what it charged. The kitchen was priced as if the hardware were basic, when the client had asked for premium fittings. The total was confident, clean and wrong for her firm in a dozen small ways.
Kavya had just met the most important rule of using AI in an interior firm. AI is a brilliant drafter. It is a terrible decider.
What AI is good at, and what it cannot know
Modern AI is very good at reading, sorting, summarising and drafting. Give it a brief and it can list the rooms and items. Give it a floor plan and it can read room sizes. Give it a rambling voice note and it can pull out the three things that need doing.
What it cannot know is your firm. It does not know what your factory charges this quarter, which vendor gave you a better rate last month, what margin you need on a small project, or what you promised this particular client on a call. It does not know whether a ceiling is really finished, because it is not standing in the room.
So the useful question is not “can AI do this?”. It is “who is responsible if this is wrong?”. Where the answer is your firm's money or reputation, a person decides.
Where AI saves real hours
Used well, AI takes away the slow, repetitive parts of the work and leaves the judgement to your team. These are the places it earns its keep in an interior firm.
- Drafting BOQ lines from a brief. The rooms, items and quantities, laid out in your structure, ready for someone to check. Not the rates.
- Reading floor plans. Room sizes pulled from the plan instead of measured with a scale and re-typed. Checked by a person before they are used.
- Making changes in plain words. “Change all wardrobes to PU” applied across a forty-line quote, instead of editing each line by hand.
- Turning voice notes into tasks. A site engineer's voice note, in whatever mix of languages he speaks, becomes a task, an issue or a material request with an owner.
- Answering questions from drawings. “What is the tile size in guest toilet 01?” answered from the drawing set, with the page shown so the engineer can see it for himself.
- Preparing a salesperson. A short summary of a lead before a call, so nobody walks in cold.
Notice the pattern. In every case, the AI produces a draft or an answer, and a person can see where it came from and confirm it. The drawing answer comes with the page. The BOQ draft comes with the lines laid out for checking. The voice note becomes a task that someone owns.
Take the voice notes alone. A site engineer might send fifteen a day, each one a minute of half-sentences over the sound of a grinder. Someone in the office used to listen to all of them and type out what mattered. Now the AI can draft the tasks and issues, and the engineer confirms with a tap. The listening still happens. The typing does not.
| Reading the plan and noting room sizes, by hand | 2 hours |
|---|---|
| Typing out every line | 4 hours |
| Looking up and entering rates | 1.5 hours |
| Checking the whole thing | 1 hour |
| Total by hand | 8.5 hours |
| With an AI draft: reviewing the drafted lines and sizes | 1 hour |
| Filling rates the library did not have | 0.5 hours |
| Checking the whole thing (this step does not shrink) | 1 hour |
| Total with an AI draft | 2.5 hours |
| Time saved on one BOQ | 6 hours |
All figures are illustrative. The point is where the time goes: the drafting shrinks, the checking does not.
Where a person must decide
There are three decisions in an interior firm that should always belong to a named person. Not because AI is weak, but because each one is a commitment that someone must stand behind.
- Prices. Every rate should come from your own rate library, built from your real costs and margins. When a line has no rate, it should be flagged for a person to fill, not guessed.
- Payments. Money leaves your account only when a person approves it, after seeing proof that the work is done. Ideally, the person who raises a payment is not the person who approves it.
- Completion. Whether a stage is finished is a judgement about a physical room. It belongs to the engineer who confirms it and the person who checks it, with photos and measurements on record.
Why AI should never write a price
Of the three, prices are where firms are most tempted, and where the damage is most quiet.
A price is a promise. When you send a quote, you are telling the client that you will deliver this item, to this specification, for this amount. That number is built from your factory's cost, your hardware supplier, your crew's output, the overheads of your studio and the margin you need to stay in business. None of that lives in a general AI model.
What AI produces instead is a plausible number. It looks like a real rate because it was shaped by many real rates. And that is exactly the problem.
Plausible is the most dangerous kind of wrong. It reads fine, so nobody checks it.
A wildly wrong number gets caught. A number that is ten percent too low on wardrobes, across a whole quote, sails through. You find out three months later, when the factory invoice arrives and the margin on the project has quietly gone.
The safe rule is simple: AI may lay out the line, but the rate always comes from your library. If there is no rate, the line is flagged, and a person decides.
A short checklist before you trust an AI tool
Whether you are trying a general chatbot or a tool built for interiors, ask these questions first.
- Does it show where each answer came from? An answer without a source is a guess you cannot check.
- Can it write a price on its own? If yes, be very careful how you use it.
- Who confirms? Every draft should land with a named person before it reaches a client, a vendor or a bank.
- Is there a record? You should be able to see what was drafted, who changed it and who approved it.
- Where does your client data go? Think twice before pasting a client's floor plan and budget into any tool you have not checked.
Kavya still uses AI every day. It drafts her BOQ lines, reads her floor plans and saves her most evenings. But the rates come from her firm's library, and she reads every line before it goes. The client gets her quote by morning, and it is her quote.
How Pakka handles it
Pakka uses AI where it drafts and keeps decisions with people. On the Sales Desk, Talk to Quote drafts quote lines from the client's brief, reads room sizes from the floor plan and takes changes in plain words. The AI never writes a price: rates come from your own library, and lines without a rate are flagged for you.
On site, voice notes become tasks, issues and approvals, and engineers who ask the drawings a question get the answer with the drawing page attached. Nothing is marked done without the engineer's tap, QC is checked with photo evidence, and payments become payable when QC passes, approved by a named person.