D-Tools tracked the quote, the equipment, the labor and the schedule, but not the drawings, photos and handover docs that explain them. I designed Files from a blank page: a shared structure that follows the deal, a filing rule that keeps it clean, and a status system teams can read at a glance.
Project files lived outside the system of record, so finding one depended on who you asked.
Structure first, storage second. Every file gets a home before it's uploaded.
End to end: problem framing, IA, interaction, visual design and specs.
Lifecycle folders, file-before-upload, visible permissions and status tags.
A single AV project collects proposals, engineering drawings, site photos, programming files and closeout packets across months and several teams. With no home inside D-Tools, those files scattered, and the context around them scattered too.
The cost isn't storage. It's the install tech working from an old drawing, the PM rebuilding a closeout packet by hand, and the salesperson who can't tell which brief the client approved.
Files isn't used by one persona. It's a relay: sales hands to engineering, engineering to programming and install, install to handover and service. Mapping who touches which documents, and when, is what turned a storage feature into an information architecture.
The folders aren't categories. They're hand-offs. If the tree mirrors the order work happens in, anyone on the project can find anything without being told where it is.
Designing from nothing means every default is a decision. These are the ones that mattered most, with the option I turned down and the trade-off I chose to accept.
Sales documents start on the quote. When the opportunity is won they carry over, and six project folders are added automatically. A shared convention means a PM can open any project and already know where the drawings are.
The picker's subtitle says it plainly: files aren't uploaded until a folder is selected. “Sort it later” is how files ended up scattered in the first place, so the product never creates a pile to sort.
Sales briefs and admin files are sensitive, but hiding them makes the structure look different for every person, and a missing folder reads as missing data. A visible lock explains itself and keeps the map identical for everyone.
States overlap. A floorplan can be Needs Review and Client-facing at once, which a single dropdown can't express. Six starter tags cover urgency, review state and client visibility, and teams can add their own.
The first picker put the whole new-tag form inline. It worked, but it made the dialog taller, pushed Save File down, and put a secondary task in the middle of the main one. The fix was subtraction: a dashed + New tag chip that opens a compact popover.
Tag creation competes with filing for attention, and the dialog grows by a third.
Filing stays the focus. Creating a tag is one click away, in the place you'd look for it.
A filing rule is only good if it fits the moment. On the quote, deals move fast and documents often arrive before anyone knows where they belong. So the quote view is lighter, and it's where I tested how far the rule should bend.
New uploads require a folder
Same rule as the project. Cleanest data, but adds friction at the fastest-moving stage.
New uploads land here as unfiled files
Matches how early sales work happens, with drag-to-file and a visible “not in a folder” area as the nudge.
I'd let the data decide: if unfiled quote files still exist when deals convert, the forgiving variant is creating debt, and the strict rule should win.
A file manager earns trust in small moments: whether a lock explains itself, whether download is one click, whether a column of files can be scanned before it's read.
Badges are colored by file type, so a folder of drawings, sheets and photos is scannable before anyone reads a filename.
Every folder shows its file count in the tree and the picker, so active folders stand out from empty ones.
It's the most common action, so it ends every row and every card instead of living in an overflow menu.
Global search covers the whole project. Each folder has its own search and filter for when you already know where to look.
Every folder ends in a drop zone, and the quote view accepts a pasted URL. Uploading never needs its own screen.
Every tag pairs a color with a dot and a label, so status still reads for color-blind users and in grayscale prints.
A filing system succeeds when people stop thinking about it. These are the signals I'd hold it to, each tied to one of the four questions, followed by the bets the structure makes possible.
With no legacy to lean on, the four questions became the brief, the review checklist and the success metrics. They turned taste debates into yes-or-no checks.
Folder-first filing, locked folders and a fixed palette are all restrictions. Each one works because the interface shows the rule and the reason at the moment it applies.
Folder names went through rounds of their own, and you can see the drift between the picker and the tree. Next time I'd lock vocabulary with integrators before any layout work.
A net-new feature designed end to end: a structure built on hand-offs, a filing rule that prevents the mess, access you can see, and status you can read.