Job Intake Cleaner
See how a customer message becomes an editable job card, then open the separate product for a fictional preview or connect your own OpenAI API key.
Product preview · bring your own OpenAI keyStart with what the customer actually said.
The next scene uses a prepared extraction of this exact fictional message.
No provider call occurs in this portfolio.
Casey Morgan asks to replace two ceiling fans and examine an outdoor outlet. A phone number is supplied, but the service address is absent and Friday has no reference date.
Reduced motion is on. Use the scene controls to explore the complete story at your own pace.
Read the complete step-by-step story
- A messy request arrives
Casey Morgan asks to replace two ceiling fans and examine an outdoor outlet. A phone number is supplied, but the service address is absent and Friday has no reference date.
- Details become a job card
The prepared illustration separates the contact, phone, requested work, timing, and access notes. Exact source quotes support extracted details; they do not certify accuracy.
- Show what needs attention
Three of four required checks are satisfied. The address is missing. Friday stays in the customer's words; no calendar date is invented. A usable phone is enough—email is not also required.
- Correct the service address
The fictional operator supplies 24 Example Lane, Exampleton, NY 10001. This is marked as a user correction, not a source extraction. Try editing the address: the next draft and export use your current value. Human review remains outstanding.
- Prepare the relevant follow-up
With the service address entered, the follow-up asks which calendar date Friday means. Clearing the address adds an address question again. The stated phone preference is retained; nothing is sent or booked.
- Export the current job card
The fictional JSON download contains the current corrected address, original timing, an unresolved exact date, and draft review status. The actual application also implements copy, CSV, and printing. This walkthrough makes no provider calls.
From the problem
to the system.
What needed to work better.
Customer requests arrive as messy messages and phone notes. Contact details, service locations, and requested work can be incomplete or contradictory, making the next conversation harder to prepare.
Design the connection.
CPL designed and built the intake experience, shared validation rules, editable job card, follow-up generation, export tools, and protected server-side extraction adapter.
A deliberate flow.
The separate product provides a fictional preview and a Settings path for a visitor-supplied OpenAI API key. Its workflow includes source evidence, editable details, required-information checks, follow-up drafts, and current-state exports. Live extraction requires a valid key and configured server protection. The portfolio separately illustrates the workflow with a fictional electrical request.
- The portfolio walkthrough shows a missing service address and ambiguous Friday timing without inventing a date.
- An operator correction updates the fictional follow-up and JSON export; human review remains visible.
- The application implements copy, CSV, JSON, and print outputs for a current edited record.
The walkthrough is scripted, uses fictional data, and makes no AI calls. The extraction adapter and editing/export workflow are implemented. Bring-your-own-key live processing has not been independently verified; illustrated playback and a working preview do not establish live readiness.
Explore the architecture & technologies
- Paste a customer message or phone notes into the separate application
- Extract a structured draft through the protected server-side adapter when configured
- Review source evidence, required details, and conflicts
- Correct the job card and prepare a relevant follow-up
- Review and export the current record
Explore the actual build.
Application screenshots with fictional data. Read each caption for its context; open an image to inspect the details.
Inspect workflow ↗