Case Volume Without Headcount

Case study - back-office intake for an insurance-policy research firm

Red Folder Research does insurance-policy research for personal-injury law firms. I built them a case-intake API over the Jira factory they already had. It went live inside a month. Well over 200 new cases a working day, with the same operations team. Their Jira automation bill disappeared. They avoided an account upgrade costing thousands.

The state I walked into

Cases arrived by email, or through a Jira service-desk form. Automations crawled over live work - routing, copying, hoping this week's fields still meant what last month's fields meant. Volume was going up. The Atlassian bill was going to follow it. Headcount was the other lever, and they didn't want to pull it.

They asked for an API. The first spec they handed over was a textbook. OAuth 2.0, two grant types, a role for every lawyer at every firm, Swagger, auto-scaling. Well done homework. Not the actual pain.

What was breaking was simpler. Their clients should not fill Jira forms. Their researchers should not wait on a pile of automations to finish inventing a case. And they should not pay Atlassian for that pile getting bigger.

Build to the use, not the bid

The bid wanted a multi-tenant law-firm product with a login screen. The use was B2B case intake: a machine submits a case, the factory keeps running, nobody learns Jira who doesn't already live there.

So I cut the user-account machinery: service-to-service tokens, firm-level access, not a permission matrix for every associate. I didn't replace Jira. The factory can stay; clients just don't work in it.

A case is a packet: plaintiff, date of loss, the other party, maybe a VIN, a police report, a file that arrives later. Submit it. We write it down first, including the ones that fail validation. The Jira form would redline you and make it hard to continue. Here the submission exists, and you can see it and fix it. Jira finds out in the background. If Jira is having a day, the request is still there.

Status is something you poll. Files attach to the case that already exists, not to a form that bounced.

Then the work that used to be automations crawling live Jira issues: routing, transforming, filling the next screen. We moved that into the service, nearly all of it, in code - versioned, tested, and released like ordinary software. Good luck doing any of that in Atlassian's automation UI. They own the automations the way you own a repository. This was before agentic coding was any good. It is a lot better now, and they are still collecting dividends.

The cases live in their database. They can query them, keep them, and report on them without waiting for Atlassian to be up, or to allow an export. A dump of Jira is not a record you control.

Along the way the processing picked up augmentations - some AI, some plain old heuristics. Each one takes a slice of researcher time off the case. A researcher still works it. They start further along. That's where the mandays went. The Atlassian bill was the obvious line item. The hours were the actual ones.

I stayed after that first month. Support, extensions, the rest of the pillar. One of their largest clients hits that API now. Almost 40% of traffic, on Kevin's numbers.

The same API, later, grew adapters. Litify, SmartAdvocate: pull from the case system the firm already runs, so a large client doesn't fill a form at all. The adapters are how you stop asking a law firm to call your API.

What I didn't nail

Billing and QuickBooks. That needed a two-hour room and a whiteboard. It got a telephone game instead - three or four people relaying what "the invoice" meant. We played whackamole against a moving explanation. I should have insisted on the room. Intake didn't get that treatment: we were looking at actual cases, not a spec that wanted to be a platform.

Kevin, afterward:

"Tin built the service that let Red Folder handle growing case volume without adding operations headcount or paying to scale Jira automation. It was live within a month and still handles hundreds of cases a day. He later built an API that's now live on one of our largest clients and handles almost 40% of our traffic."

— Kevin Crotteau, Technical Owner at Red Folder Research

If the next hundred cases would mean another hire, or another vendor tier - let's talk →