How to extract a sign schedule from tender drawings with AI

You can upload architectural plans or a signage schedule straight into an AI tool like ChatGPT or Claude and have it extract every sign into a table, type, quantity, location, and specification, in minutes instead of an afternoon. Done properly, it's the single biggest time-saver AI currently offers a sign estimator. Done carelessly, it's a fast way to price a job off a hallucinated schedule. Here's how to get the first outcome and not the second.

Why this works now

Modern AI tools read uploaded PDFs, including drawings, directly. You don't retype anything or copy-paste text out of a plan set. You upload the file, ask for what you want, and the AI reads the document the way it reads anything else. For tender packs, that means the slow, careful first pass, finding every sign reference scattered across a drawing set, can be done by the machine, with you checking its work instead of doing the hunt yourself.

The basic extraction

Upload the signage schedule or the relevant drawings, then ask for a structured pull. A prompt that works:

"I've uploaded the architectural plans and signage schedule for a project. Read them and pull out every sign into a table: sign type, quantity, location or drawing reference, and any specification given (size, material, illumination). Flag anything referenced but not fully specified so I know what to query."

The drawing reference column matters. Every row in the AI's table should point back to where it found the sign, so you can verify each one against the source in seconds.

Then hunt the gaps

The second pass is where the real value is. Ask:

"From the documents I've uploaded, list the information an estimator would need to price this signage package that isn't specified, things like substrate, fixing details, access, power supply, or consent requirements. Give me a clear list to raise with the contractor before pricing."

That list becomes your RFI. Tender documents are reliably silent on the things that cost sign companies money, the substrate behind the feature wall, whether power's been allowed for the illuminated sign, who's responsible for access equipment. Getting those questions in early, before the price, is the difference between a quote with allowances and a quote with holes.

The verification rule

Never price off the AI's table without checking it against the drawings. This is the non-negotiable part. AI reads documents well, but it can misread a quantity, miss a sign that only appears on one elevation, or confidently invent a specification that isn't there. The table is a fast first pass, not a final schedule.

The efficient workflow: AI extracts, you verify each row against its drawing reference, you correct what it got wrong. That's still dramatically faster than building the schedule from scratch, and the verification pass often catches things you'd both have missed.

What this doesn't replace

The extraction tells you what's in the documents. It can't tell you what's on site. A tender schedule says "6 illuminated fascia signs", it doesn't say the substrate is failing render, the power's forty metres away, or the access needs traffic management. Those answers still come from a site visit, and they're the difference between the tender price and the real price.

Use AI to master the paperwork fast, then spend the time you saved where it actually pays: on site, getting the information the documents don't have.

---

SignPad is a site survey app for sign companies. Once the tender's won, it captures the site data the documents never had, photos, substrates, measurements, access and power against every sign location. See a sample report.