A technician photographs the signed work order in the driveway, and the office opens a job row with the customer, the address, the job number, and the date already filled in. That is the working outcome, live in your workspace, built from one prompt.
The app embedded on this page is the Application Tracker Board, a real Taskade Genesis build you can work right now. It is a stage-by-stage tracking board rather than a work order desk, so treat it as the closest starting point: click "Use this app" to clone it in about ten seconds, rename the lanes to Received, Reviewed, and Invoiced, then add the upload form, the reading step, and the extraction described below.
Here is what the build gives you concretely. A Work Orders Project holds one row per job. A published form carries a technician field, a customer field, and a File field labeled "Photo or scan of the signed work order". An automation fires on submit. The Convert File to Text step reads the page first, and this is the part that matters most: a HEIC photo straight off an iPhone camera roll, an image-only PDF from the office scanner, and a typed PDF from dispatch all pass through that one step. The automation gets a single path instead of a sort by file type. An Ask Agent step then returns customer name, service address, job number, date of service, technician, parts used, and total as separate named fields. Those values land in columns, the uploaded photo attaches to the same row, and status moves to Reviewed only when every required column has a value. When one is missing the row stays at Received, so nobody has to guess which jobs still need a human.
The same job data renders in every layout the office needs:
- Table view for the full job register with customer, job number, tech, and total side by side
- Board view to drag jobs through Received, Reviewed, and Invoiced
- Calendar view to see service dates across the whole crew
- Gantt view for multi-day installs and their dependencies
- Plus List, Mind Map, and Org Chart across the 7 project views
Automations carry the job past intake. Wire it up with Taskade automations and a submitted work order pings dispatch in Slack, emails the customer a confirmation, and drops the totals into your accounting sheet without anyone opening the app. Across 100+ bidirectional integrations, Gmail and web forms pull paperwork in while Slack, Google Sheets, and calendar invites push the finished job out. The AI agents inside the app carry 34 built-in tools including persistent memory, file analysis, and web search, so the agent that reads today's sheet remembers how last month's sheets from the same customer were coded.
Ownership is simple. Clone the app, invite your crew and your office staff, and the jobs, the photos, and the extraction rules are yours to keep and reshape.
Start from the full walkthrough in Build a Document Intake App, check the upload behavior in File Uploads in Your App, and see which formats are read off the page in Read Scanned Paperwork and Photos. Browse live builds in the Community Gallery or shape the destination columns with Databases in Projects. Pair this with Send Supplier Invoices Straight to an Approval Board for the money half of the loop, or Log Safety Inspection Sheets as You Collect Them for the compliance half.
Frequently Asked Questions
Will a phone photo taken in bad light actually read?
Usually yes, and the honest answer is that light and focus matter more than resolution. The reading step handles HEIC, JPG, PNG, and WebP, so a camera-roll photo needs no conversion. Keep the sheet flat and fill the frame. When a value comes back empty the row stays at Received rather than saving a wrong number, so a bad photo becomes a visible task rather than a silent error.
What size limit applies to a photographed work order?
Keep scans and photos under 20 MB. That ceiling covers reading words off an image, which is the path every phone photo takes. A typed PDF sent from the office is read as text instead, so it is bounded by your plan upload size rather than that number. Most phone photos land far under both.
Does reading the page cost anything?
Reading words off a scan or a photo uses AI credits, the same as any other AI step in an automation. A file that already has selectable text is read straight through and does not use them. See AI Credits and Models for how credits work across a workspace.
Can technicians see only their own jobs?
Yes. Turn on app sign-in through App Users and every upload arrives attached to a named person instead of an anonymous submission. You then publish two surfaces from the same data: technicians see the jobs they submitted, office staff see the full dashboard.
What happens to the original photo after the values are pulled?
It stays. The file is stored with the submission, attaches to the job row, and lands in your workspace Media tab. When a customer disputes a signature or a part number you open the actual page rather than trusting a transcription.
How do I check the automation is reading correctly before I trust it?
Submit one test work order, then open the automation History tab and look at what the reading step returned before wiring anything after it. Build in small prompts with a look in between. That habit turns a debugging session into a thirty-second glance.
Can I run this on the Free plan?
You can build it and test the path. Free includes 6,000 one-time AI credits, 3 Taskade Genesis apps, and 10 flow runs in total, which is enough to prove the intake works end to end. Running a real crew on it needs a paid plan, and Pro is $10 per month billed annually.
