Background
Priya and her co-founders had deep data science expertise but limited full-stack bandwidth. They'd budgeted 6 weeks for the beta: two weeks for infrastructure, two weeks for core product, and two weeks for integration and bug fixes. The two-week infrastructure estimate was optimistic — they'd gone over on every previous project. When their seed investors moved the demo date forward, they needed a way to compress the timeline without cutting product scope.
The challenge
The team needed production-quality auth (with team/workspace support for B2B), usage tracking per account, an admin panel that would let their non-technical ops person manage users and subscriptions, and a clean enough architecture that two engineers could work in parallel without stepping on each other. They also needed the whole thing to be maintainable — a quick hack that fell apart in three months wasn't acceptable.
How they built it
Workspace auth wired in half a day
Clio needed per-workspace auth for their B2B model. The Better Auth foundation in ShipAI already handled multi-user session management and protected routes. Priya extended the default auth patterns to add workspace scoping — the extension points were clear and the session model was documented in the code. What she'd estimated as three days took about four hours.
Admin panel used by ops on week one
ShipAI ships with eight admin views covering user management, subscription controls, usage analytics, trace viewer, and log explorer. Priya handed the admin URL to their operations person on day three. No customization required — the views connected to real application data immediately. The ops team was using it to answer user questions before the engineers had finished the analytics core.
Usage metering for per-workspace billing
Clio charges per data query. The built-in usage metering tracked per-user counts and the plan enforcement middleware enforced limits at the API route level. This was integrated into the analytics query endpoints in an afternoon — the pattern was already established in the codebase, so the engineer doing this work could follow existing examples rather than design from scratch.
Two engineers, zero stepping on each other
The clear service boundaries — UI components, API routes, data model, AI handlers — meant both engineers could work in parallel with minimal merge conflicts. One owned the analytics core; the other owned the workspace and billing layer. The folder structure made the boundaries obvious enough that they didn't need explicit coordination beyond a morning standup.
Outcomes
Beta shipped in 18 days vs. 6-week estimate
Infrastructure that was estimated at 2 weeks took 3 days, freeing the team to focus their remaining time entirely on the analytics product.
Admin panel in use by day 3
The operations team member started using the admin panel on day three without any customization or training documentation.
Zero backend specialist hired
The team shipped a production B2B product with team auth, usage metering, subscription billing, and an admin console without adding backend headcount.
Parallel development with no coordination overhead
Two engineers worked simultaneously across different product layers with minimal merge conflicts across the 18-day sprint.
In their own words
The admin panel is the thing I didn't know we needed until we had it. Our ops person was using it to manage users within days of launch. We'd have built something simpler and more brittle. The full version being there from day one set a standard that made the whole product feel more serious.
“We were two engineers and a designer. We needed to move fast without hiring a backend specialist. ShipAI gave us auth, usage metering, and an admin panel that our ops team actually uses. We shipped the public beta in 18 days instead of the 6 weeks we'd budgeted.”
— Priya Ramachandran
Frequently asked questions
How did the team handle B2B workspace scoping?
Priya extended the Better Auth session model to include a workspace ID. The protected route patterns already existed — she applied workspace validation in the same middleware layer, following the existing pattern.
Did the team use the AI infrastructure features?
Not in the initial beta. They used auth, billing, usage metering, admin panel, and the landing UI system. The AI handler infrastructure was available for the second phase — Clio planned to add AI-assisted analytics summaries in the following sprint.
What would have taken longest to build from scratch?
Priya says the admin panel and usage metering together would have taken the most time — probably a full week each. Having them pre-built and connected to real data from day one was the single biggest time saving.