From inbox to signed affidavit: the operations platform behind a process-serving firm
Before any AI, we automated a process server's paper trail: email intake, field sheets, barcoded print runs, and affidavits assembled from marked-up paper.
MithraAuto · MithraProcess serving runs on paper. A law firm emails a request with court documents attached. Someone keys the case into a system, prints a field sheet, and hands it to a server. The server attempts delivery, records what happened, and comes back with a sheet full of handwriting and checkmarks. Then someone turns that sheet into a sworn affidavit, with the right case number, the right address and the right wording, and files it on time. Every handoff is a chance to retype something wrong, and an error in an affidavit is not a typo. It's a problem in court.
At a glance
| Client | Mithra, a process-serving and investigations firm |
| Industry | Legal support |
| What we built | An internal operations platform: email intake, job tracking, generated field sheets, barcoded print runs, a client portal, and affidavits assembled from scanned field sheets |
| Where AI fits | It doesn't. This is deliberate, deterministic automation, and the groundwork that later AI document work was built on. |
| Stack | Python web apps with a MySQL database, PDF generation and image processing, plus a small mobile wrapper app ("PaperPush") |
| Status | An operations platform for a process-serving firm, from email intake to signed affidavit |
The old way
- Read each emailed request and type the case, parties and addresses into a tracking system.
- Fill in a field sheet by hand or from a template for every serve attempt.
- Print and sort paperwork for the day's routes, flagging rush jobs by word of mouth.
- Collect the completed field sheets from servers.
- Retype what the server marked into an affidavit, check it against the original request, and send it to the client.
The same facts were typed three times: at intake, on the field sheet and on the affidavit.
What we built
MithraAuto turns that chain into one record that flows from start to finish.
- Intake. A scheduled job reads the firm's intake mailbox, saves each request as a job with its own job number, and keeps the attachments with it.
- Field sheets. Each job generates a printable field sheet, pre-filled from the database, with a barcode that ties the paper back to the job.
- Print queue and rush alerts. The day's paperwork goes into a print queue automatically, and rush jobs trigger a notification instead of relying on someone noticing.
- Reading the paper back. When a completed sheet is scanned, the system straightens the page and measures how dark each answer bubble is against the page's blank baseline, to read which boxes the server marked.
- Affidavits. The marked answers, the intake data and the firm's standard wording are combined into a filled affidavit, with alternate addresses chosen from what the server recorded.
- Client portal. Law firms submit jobs and follow progress without emailing for updates.
How the workflow changes
| Step | Before | After |
|---|---|---|
| Intake | Read the email, type the job in | Requests become jobs automatically, numbered and filed with their documents |
| Field paperwork | Fill out a sheet per attempt | Pre-filled, barcoded field sheets generated from the job record |
| Daily print run | Gather and sort by hand | A print queue assembles the day's paperwork; rush jobs raise an alert |
| Recording the serve | Handwriting to be retyped later | The server marks bubbles on paper they already carry; a scan reads them back |
| Affidavit | Retype everything, then proofread against the request | Assembled from the same record that started at intake, so the case number is typed once |
| Client updates | Phone calls and emails | A client portal for submitting and tracking jobs |
Inside the build
One record from the first email to the signed affidavit.
Under the hood
- Paper is a first-class input. Servers keep working on paper, which is reliable in a stairwell with no signal. The software meets them there instead of forcing a phone app on every attempt.
- Reading marks without guesswork. Bubble detection compares each answer's darkness with a blank region of the same page, so a faint scan or a different printer doesn't flip the result.
- Deskewing first. Pages scanned at an angle are straightened before any answer is read, because a few degrees of tilt is enough to read the wrong box.
- One source of truth. Job numbers come from the database, and the affidavit pulls from the same tables as the field sheet, which removes the retyping step where most errors entered.
By the numbers
- 283 commits, built as the firm's internal operations system between May 2024 and May 2025. We wrote most of it, and contributor omarmajz added 58 commits.
- About 22,500 lines of Python across the operations app and the client portal.
- One record from intake to affidavit: job numbers come from the database, and every document pulls from the same tables.
Its real value to Second Voyage is what came next. Knowing the process-serving workflow in this detail is what made the later AI work possible: our Hussar Investigations platform reads uploaded court documents with a language model and pre-fills the job, which is the intake step here, done with AI.
Example scenario
Example scenario (illustrative): A small process-serving firm handles 40 new jobs a week from a dozen law firms. Today, an office manager spends each morning keying requests and each afternoon typing affidavits. With an intake-to-affidavit pipeline, the morning becomes a review of auto-created jobs and a printed stack of barcoded field sheets. The afternoon becomes scanning returned sheets and proofreading affidavits that are already filled in, against the same record they came from.
What this means for your business
The lesson is that automation starts with the workflow, not the model. Before adding AI, get every fact entered once and flowing through to the final document. Then AI can take over the hardest remaining step. This fits:
- Legal-support firms: process servers, court reporters, investigators
- Field-service companies that still run on paper job sheets
- Insurance, title and compliance back offices that turn forms into sworn or signed documents
- Any team retyping the same facts into three systems
Second Voyage builds AI systems that are held to a standard you can check. Talk to us about yours.