Projects
Status, phase, room, what's approved and what's still sourcing.
"Summarize the Ashford Residence project."
Platform · Walter AI
It reads your studio.
It reads your studio.
And it knows the trade.
Walter answers from your projects, invoices, contacts and timelogs — the records your team already keeps — and it's been trained on the things a designer is expected to know: codes, clearances, standard dimensions, how the trade works. It will also tell you plainly what it can't do.
Hi, I'm Walter.
Your project insights assistantI read across your studio — projects, invoices, contacts, and timelogs — to help you find, summarize, and update things fast.
Have a support question? Walter can't help with billing, bugs, or account issues. For those, email the success team or use the chat icon in the bottom-left.
What invoices are unpaid?
Summarize the Ashford Residence project.
How many hours did we log this week?
Add a note to my newest project saying client approved the sofa.
Ask about your projects — "Summarize my Ashford project", "What invoices are unpaid?"
Walter is AI. It is right far more often than it is wrong, and it can still be confidently wrong — check anything load-bearing against the source.
Trusted by the studios doing the work
4 records
projects, invoices, contacts and timelogs — the ones you already keep
0 tabs
opened to answer a question you could have just asked
Ask or tell
it answers questions, and it writes notes back to the record
Says no
billing, bugs and account questions go to a person, not to Walter
Grounded
An answer is only as good as what it read
Most AI in this category is a general model with your logo on it. Walter is pointed at four kinds of record and nothing else — which is why its answers are specific enough to act on, and why it can be wrong in ways you can check.
Invoices
What's been sent, what's been paid, what's sitting unbilled.
"What invoices are unpaid?"Contacts
Clients, vendors and reps, with the trade discount and the history attached.
"Who's the rep at Stark?"Timelogs
Hours by person, project and phase, billable and not.
"How many hours did we log this week?"
And what it was taught
It also knows the things a designer is expected to know
Walter carries trade knowledge as well as your records — codes and clearances, standard furniture dimensions, how the sequence of a job normally runs. So "will a 96-inch sofa work in that room" is a question it can take a view on, not just a question about your data.
That knowledge comes from training, not from your project. Treat it the way you'd treat a well-read colleague's opinion — useful, usually right, and worth confirming before it reaches a client or an inspector.
Ask it, or tell it
The fourth suggestion isn't a question
Three of the prompts Walter offers are things to find out. The fourth — "add a note to my newest project saying client approved the sofa" — is a thing to do. That's the difference between a search box and an assistant: you can hand it the sentence you'd have typed into the project yourself.
- It knows what "my newest project" means, so you don't have to name it.
- The note lands on the record, where the next person to open the project will find it.
- Chats are kept, so "what did I ask about the Ashford invoices" is itself answerable.
- Check its work. An answer that moves money or changes a record deserves the same glance you'd give a junior's.
G
Which projects are over their estimated hours?
W
Based on connected studio data · just now
Robin Residence is 54.1 hours over estimate. Most of the overage is in design development. Zapata Residence is 7.4 hours over in procurement.
Estimated20.0 hrs
Logged74.1 hrs
Over+270%
G
Add a note to Robin Residence saying the client approved the sofa.
W
Added to Robin Residence.
Project note · todayClient approved the sofa.
The honest part
It tells you what it can't do, before you ask
Walter's own opening screen says it: billing, bugs and account issues aren't its job, and it points you at a person instead. We'd rather it declined clearly than guessed confidently — a wrong answer about an invoice is worse than no answer, because you'd act on it.
- Support goes to the success team, by email or the chat icon in the app.
- Check its sources. Walter is trained on trade knowledge as well as your records, and like any AI it can be confidently wrong. Verify anything you'd be embarrassed to get wrong in front of a client or a building inspector.
- Your data stays yours. It's never used to train models for anyone else, and never crosses between accounts.
Walter can't help with
BillingInvoices from us, plan changes, payment methods
BugsSomething in Workroom behaving incorrectly
Account issuesAccess, permissions, seats, log-in trouble
Those go to the success team, who are people, and who can actually change something.
The connected record
Why it can answer at all
An assistant is only useful in a system where the records already talk to each other. Walter isn't doing the connecting — the platform did that. Walter is what it looks like when you can finally ask.
Hours meet money
"Which projects are over their estimated hours" is only answerable because timelogs and estimates sit on the same project.
Invoices know their project
Unpaid isn't a list of documents — it's a list of clients and jobs, because that's how the invoice was built.
Contacts carry their history
A vendor's discount, their POs and their reps are one record, so a question about them has somewhere to land.
Notes go back in
Asking is half of it. Walter writing a note onto the project is what stops the answer dying in a chat window.
Recent chats persist
The thread you started on Tuesday is still there on Friday, under its own name in the sidebar.
Scoped to your account
Walter reads your studio's records and no one else's, and your data never trains a model for another firm.
Set it once, per firm
The reason AI writing gets deleted is that it does not sound like you
Every studio has a voice, and clients can tell when it slips. A draft that says “reach out” when you always say “connect” is not wrong, it is just somebody else — and rewriting it takes longer than writing it would have.
Describe the voice once, in a sentence. Add the word swaps you actually care about. From then on every draft Walter writes — replies, rewrites, email summaries — honours them, and confirmation stays on by default so nothing is committed without you seeing it first.
- One plain-English description of your voice
- Word swaps Walter always honours
- Confirmation on by default — nothing commits silently
- Walter only sees data inside your own workspace
Brand voiceTell Walter how your studio sounds. This shapes every draft Walter writes — replies, inline rewrites, and email drafts and summaries.
Voice & tone
Warm and personal, never stuffy. Short sentences. We speak to clients like a trusted friend who happens to be an expert. British spelling.
Say this, not that
Word swaps Walter always honours.
“reach out” → “connect”“utilise” → “use”“spend” → “investment”
Autonomy
Let Walter make changes without confirmingWhen off, Walter describes what it would write and asks you to reply “yes” before committing.
Walter can only see data inside your own workspace.
Questions we hear
Walter, specifically answered
What can Walter actually see?
Two things. Your studio's own records — projects, invoices, contacts and timelogs — and general design knowledge it was trained on, such as codes, clearances and standard furniture dimensions. Answers about your firm come from your data; answers about the trade come from training, and both are worth verifying before you act on them.
How accurate is it?
Treat it as a well-read assistant, not an authority. Walter is right far more often than it's wrong, and it is still AI — it can state something incorrect with complete confidence. Anything load-bearing — a code requirement, a clearance, a number going to a client — should be checked against the source before you rely on it.
Can it change things, or only answer?
Both. Adding a note to a project is one of the four suggestions it offers on the opening screen. Treat anything that changes a record the way you'd treat a junior's work — useful, fast, and worth a glance before you rely on it.
Why won't it help with billing or a bug?
Because it would be guessing, and a confident wrong answer about your account is worse than no answer. Walter says so on its own opening screen and points you at the success team, who can actually change something.
Is our data used to train models?
No. Walter analyzes your firm's own records to serve your account only. Your data is never used to train models for other customers and never crosses between accounts. More on security and trust.
Who on the team can use it?
Anyone with access to the records in question. Walter respects the same permissions the rest of Workroom does — it can't surface a margin to someone who isn't allowed to see margins.
See the connected workflow
Bring a real project. Ask it something hard.
The demo works best when you ask Walter a question you actually can't answer today — about your hours, your unbilled work, or a client you haven't looked at in a month.