The challenge
The studio designs residential interiors in presentation decks, then re-enters everything in those decks into a project management tool to produce line items and purchase orders. They estimated the re-entry at sixteen hours per project and wanted to know whether that could be automated.
Rather than take the estimate on faith, we asked for one real project file and took it apart with a script before scoping anything. The file was one gigabyte, 2,204 individual files inside a single document, 68 slides, roughly 25 room sections. It contained 468 unique product hyperlinks across 724 link instances, pointing at 114 distinct vendor domains. It carried 821 free-text specification notes, about twelve per slide.
Sixteen hours across 468 items is 2.05 minutes per item. That is exactly what the clip-and-paste workflow costs: paste the link, wait, highlight the fields, add the size and the finish, save, next. Their estimate was not a rough guess. It was arithmetic, and it was correct.
The finding that reframed the whole engagement was this: of those 821 specification notes, six contained a dimension a computer could read. Not six percent. Six.
The notes themselves are complete. "Artistic tile ocean blue vein cut, wainscot halls in entire room, Schluter greige, Nordic Timber 48 tall, 8 x 48 tiles, run one piece vertically, Tec sterling power grout." Everything a person needs to place that order is there. It is sitting in a text box, beside an image, near a hyperlink, with nothing connecting the three to each other.
So the sixteen hours is not a data entry problem. It is a data recovery problem. Every fact needed to build those line items was known and typed by a human, once, during design, and then flattened into a slide where the image, the link and the note stopped being related. The team was going back into their own deck to reassemble information they already had.
Two smaller findings came out of the same teardown. The file was a gigabyte because two individual images were 73 MB each: raw, uncropped photos straight off a phone, with nothing resizing anything on the way in. And the ordering was far more concentrated than the 114 vendors suggested, with a single vendor accounting for 153 of the 468 links and the top ten covering roughly half the project.
The solution
Three phases, each delivering something usable alone, with a decision point after each.
The first extracts and specifies. It reads the deck, pulling every hyperlink, text note, image and the room and slide structure, which was already proven working on their real file. It visits every product page and retrieves the actual data: name, vendor, SKU, list price, dimensions, available finishes and sizes, current availability. It matches notes to products using the position of the text relative to the image and link on the slide, which is the step that recovers the connection the deck destroyed.
Then it flags the mismatches, which is the part that saves the most pain rather than the most time. When the link resolves to the 84-inch and the note says 96, or the link is Fog and the note says Oatmeal, it says so before anybody places an order. It flags dead and repriced links too. Across 468 products in a project running April to July, some are discontinued or out of stock, and today the studio discovers that at ordering time rather than at review time.
The output is a review screen grouped by room where the team corrects anything wrong, then an export ready for import. Sixteen hours of retyping becomes roughly an hour of reviewing a list somebody else assembled, and it applies to decks already built rather than only future ones.
The second phase bridges into their project management tool, and it opens with three tests run in their own account and reported honestly: whether the tool's CSV catalog import actually creates project line items or only populates a library, whether its newer paste-a-URL feature already solves part of this, and whether an existing twelve-dollar app covers a meaningful chunk. The commitment, in writing, is to tell them if the answer is that they should use software they already pay for and stop.
The third phase is the design canvas that replaces the deck entirely, capturing product data at the moment of design so the first two phases stop being necessary. It was deliberately placed last, after the extraction, validation and integration paths are proven, rather than leading with the biggest rebuild.
The results
Sixteen hours per project collapses to roughly one hour of reviewing a list somebody else assembled. That is about a 94% reduction, and it is a more trustworthy number than most automation estimates because the denominator was measured rather than reported: 468 items counted in a real file, at a per-item cost the studio's own estimate independently confirmed.
It works backwards as well as forwards. The links are still sitting in decks the studio has already built and already billed for, so the first run can be pointed at finished projects rather than waiting for the next one to start.
Vendor concentration is what makes the build tractable. Extraction gets written vendor by vendor, and a raw count of 114 sounds unbounded. One vendor covering a third of the project and ten covering half means the first pass captures most of the value, with the long tail handled by the review screen rather than by more engineering.
The mismatch detection is worth separating from the time savings entirely, because it is not an efficiency gain. When the link resolves to the 84-inch and the note says 96, or the link is Fog and the note says Oatmeal, catching it at review instead of at delivery avoids a wrong shipment, a restocking fee and a schedule slip on a project where the client is watching. Across 468 products on a job running April to July, dead and repriced links are a certainty rather than a risk.
Three weeks to the first working phase, and the phases are ordered so the studio can stop after any one of them. The design canvas that would replace the deck entirely sits last, after extraction, validation and the integration path are all proven, rather than leading with the largest rebuild and hoping it pays for itself.