Process · Blue Jay Build Co

Fig. 08a

Build

— Weeks 1–2

Faciendum est

We create the agent in a staging environment that mirrors your systems. You see the work-in-progress as it goes. By the end you have a real, working version, not a slideshow about one.

— Milestone 01

Plug-and-play agent workflow
Live demo on a real input
Technical scoping & feasibility report

Fig. 08b

Integrate

— Weeks 3–5

Coniungendum est

We connect the agent to your real systems and operational tools. Authentication, error handling, edge cases — the parts that matter when something goes wrong at 2 AM and there's nobody at a desk.

— Milestone 02

Production integration with your stack
Robust testing on real volume
One month of tuning support, free

Fig. 08c

Tune

— Ongoing, optional

Curandum est

If you want us around for the long tail — model upgrades, new edge cases, the inevitable third-party API change — we're around. Either as part of a Studio-time retainer or a la carte.

— What's included

Quarterly tune-up sessions
Model & SDK migrations
Monthly check-in calls

❦

Fig. 09 — How payment works

Low-risk by design. Refundable by contract.

AI agents in 2026 are still new. Expectations are evolving. Every part of the payment shape exists because we don't want you to feel locked in if the build isn't earning its keep.

Fig. 09a — At kickoff

50%

Of Phase 01 fee. We start work the same week.

Fig. 09b — At Milestone 01

50%

Of Phase 01. Reviewed together. Approve to proceed.

Fig. 09c — Phase 02 start

50%

Of Phase 02. Only after you're happy with Phase 01.

Fig. 09d — At Milestone 02

50%

Of Phase 02. Production-ready, tested, integrated.

If a build proves technically infeasible, you get a full refund.

Not "credit toward another engagement." Not "we'll keep working until you give up." A wire transfer back. We've issued one refund in our short history. We'd rather issue a hundred more than keep money for work that didn't pay back.

If a build is feasible but you decide mid-way it isn't worth finishing, you keep what's been built and we part ways at the milestone. No prorated fight.

❦

Fig. 10 — In practice

Multiple agents can run in parallel.

Most engagements are 2–3 agents, not one. They sit at different phases at the same time — one in Build, one in Integrate, one in Tune.

— Sample 12-week engagement (3 agents)

Triage AgentTickets

Recap AgentQBR drafts

Pulse AgentHealth scoring

Week

1–2

3–4

5–6

7–8

9–10

11–12