VelvetWorked examples
Five things that happen while nobody is watching.
Each one runs left to right. Scroll sideways, use the arrows, or the arrow keys. Pick a scenario to start.
Built from the questions these businesses actually get. Figures are examples, not a real price list. Of the two LOOM scenarios, the first is the part that works today and the second describes the reporting half, which is still being built. The SILK scenario describes a product that has not been started.
01 / 00
23:41 · Thursday
The shop is closed. The customer is not.
She is putting together client handover gifts, after her own working day, which is when most of this happens.
Without the bot: a contact form, an answer on Sunday, and by then she has ordered somewhere else.
23:41 · First question
It answers from your price list.
The bot has your actual document: packages, prices, what is in each one. It quotes that.
Note the format: a list question gets a list, with prices. Not a paragraph she has to read twice.
- Corporate Thank You Basket · AED 350
- Luxury Chocolate Basket · AED 450
- Classic Comfort Basket · AED 220
23:42 · Logistics
Delivery zones and lead times, without you.
The two questions that decide whether an order is even possible, answered from your own zone list.
From your areas, not from a map. Small delivery businesses price by zone, not by the kilometre, so the bot looks the area up and gives the real fee.
23:44 · It asks for the order
Then it collects what you would have asked for.
Name, number, and the request in her own words, and it reads the details back so nothing is misheard.
Per market: gifting asks occasion, budget, date and area. Another industry asks its own four things.
23:44 · Your phone
The whole thing arrives as one message.
Not an email you will read on Sunday. A WhatsApp notification with everything she said, on the number you nominate.
Your own WhatsApp is never connected. The sending number is ours; customers never see it and never message it.
23:44
- Name
- Reem A.
- Phone
- +971 5X XXX XXXX
- Wants
- Eight corporate gift baskets, delivery to Business Bay, budget AED 2,500, needed Thursday.
Everything she told it, in the order she said it.
08:10 · Next morning
You reply once, already knowing everything.
No discovery call, no back and forth to establish what she wants. The quote goes out before the shop opens.
What changed: the enquiry did not wait, and you did not stay up for it.
01 / 00
09:02 · Monday
Ramadan pricing goes up on the website.
She edits the hamper page herself, the way she always has. One place, five minutes, done.
The thing she does not do: log into anything else. There is no second copy of her prices to keep in step.
10:15 · The honest part
For a couple of hours, the bot is behind.
It last read the site at six this morning. Until the next pass, it is still quoting Monday's old figure.
Worth knowing rather than hiding: this window is exactly why LOOM has a re-scan button. Automatic covers the days you forget; manual covers the morning it matters.
12:00 · Scheduled
It will re-read the site on its own.
Coming soon: four times a day, every six hours, WEFT will fetch the pages again and rebuild what it knows. Nobody will need to trigger it.
Why a schedule and not every message: re-reading on every question would be slow and wasteful. Every six hours is fresh enough for prices and cheap enough to run forever.
12:14 · The next customer
Same question. Correct answer.
New price, and the new lead time along with it, because the bot re-read the page and not just the number.
The website stays the single source of truth. She maintains one thing, and everything downstream follows it.
End of season
Nobody had to remember anything.
Prices changed four times across the season. The widget was correct within hours of each one, and she never opened a dashboard.
The failure this prevents: a bot quoting a price you no longer honour is worse than no bot. It commits you, in writing, to a number you have moved on from.
01 / 00
Tuesday · 09:40
The answer exists. It is in five places.
An events company, four people, one shared inbox. A client rings about a quote from March. The figure is in an email, the revised menu is a PDF on somebody’s drive, and the person who sent it is on leave until Sunday.
This is not a filing problem. Everything was saved properly. It is a retrieval problem, and folders have never solved that one.
Setup · once
It reads what you already have, where it already is.
Mail, drives, the document library, and anything somebody uploads by hand. LOOM indexes them in place. Nothing is moved, renamed or reorganised.
Nobody has to change how they work. The great tidy-up that every document system asks for, and never gets, is not part of this.
09:41
One question, in the words you would have said out loud.
Not a filter, not a saved search, not a folder to guess at. The question is the whole interface, and it can cross the inbox and the drive in one go.
The client is still on the phone. That is the bar this has to clear, and twenty minutes of searching does not clear it.
The part that makes it usable
Every line says where it came from.
The answer arrives with its sources attached: which email, which file, which date. You can open any of them and read the original.
This is the difference between an assistant and a guess. An answer you cannot check is one you have to verify by hand anyway, which puts you back where you started.
Where this sits
This is the half of LOOM that works today.
Asking your own mail and documents a question is built and running. The reporting half, the metrics and the keywords and the gap report, needs a history of enquiries to read before it can say anything true.
Which is why they arrive in this order. Answering works on the day you connect it. Reporting gets better every month it runs.
01 / 00
The 1st · In development
A month of conversations, and no time to read them.
1,284 people asked this business something. 213 became enquiries. The other 1,071 went unexamined, as they do everywhere.
This scenario describes LOOM, which is being built. It is what we are making, not something you can buy today.
The question
Asked in plain language, not built in a report.
No filters, no pivot table, no dashboard to configure first. The enquiries are indexed, so you ask them things.
Every answer is traceable. It returns the conversations it drew from, so you can disagree with it by reading them.
- Flower arrangements · 119 conversations
- Abu Dhabi delivery · 84
- Balloon & décor setup · 52
- Cakes · 31
The finding
One hundred and nineteen people asked for flowers.
Each of those was one polite refusal, months apart, forgotten by the next morning. Counted together they are a business case.
This is the number no analytics tool has. Page views tell you who arrived. Only the conversation knows what they wanted and did not get.
The decision
A partnership, not a guess.
She does not become a florist. She arranges supply with one, adds a line to the site, and the next request is an order instead of an apology.
And the loop closes: the site changes, WEFT re-reads it, and the bot starts selling the new line without anyone configuring it.
Why this is the bigger-company product
Under a few hundred enquiries, reading them still works.
LOOM earns its place when the volume is past what any one person can hold in their head: an agency, a group of branches, a business with a marketing team.
Which is the honest split: WEFT is for the owner who answers their own phone. LOOM is for the company that stopped being able to.
01 / 00
The decision · Not built yet
You changed provider. Nine years of mail did not come with it.
Six mailboxes, a folder structure people actually navigate by, and every attachment a client ever sent. None of it moves because the domain did.
This scenario describes SILK, which has not been started. It is on the list, not in the product. The two scenarios above it are real code.
The quote
A consultant, a weekend, or losing some of it.
The usual routes are an export file per mailbox, or dragging folders between two open clients and hoping the connection holds.
The failure mode is quiet. A drag that stops at 70% does not announce itself. It is found months later, by someone looking for one specific email.
Setup · 4 minutes
Sign in to the old one. Sign in to the new one.
Both ends authorise through their own provider. No password is ever handed over, and nothing is typed into a spreadsheet.
Read-only on the source. The mailbox you are leaving is never written to, so the old account stays exactly as it was until you close it yourself.
Running · overnight
It copies. It does not cut.
Folders map across, dates stay the dates the mail arrived, read and unread survive, attachments come with their messages.
Interruptions are expected, not fatal. If the link drops at 60% it continues from 60%, because a migration that has to restart is a migration nobody finishes.
Morning
The part nobody else does: proving it arrived.
Every folder is counted on both sides and the two numbers are put next to each other. Any that disagree are named.
A migration you cannot check is one you have to trust. This is the difference between believing it worked and knowing.
Where SILK sits
The one you need once, and badly.
Nobody migrates mail every month. They do it when they rebrand, merge, leave a provider, or inherit a company that never left one.
Which is the honest split: WEFT answers your customers, LOOM reads what they asked, SILK is what you use on the day everything moves.
Get in touch
Want to see one of these with your own prices in it?
Send us the questions you answer most and we will build the same walkthrough for your business.
velvetworksco@gmail.com