Process

How we work, start to ship.

Four phases. The first one is free. The second is paid but small. By the time you sign off on the design, you know exactly what we're building, what it costs, and when you'll see it running.

04 — Process

How we work, start to ship.

  1. 01

    Assess

    Week 0 — free

    A 30-minute call. We send you a one-page brief: the agent's job, the data, the integrations, the success metric.

  2. 02

    Design

    Week 1

    We design the agent — prompts, tools, guardrails, fallbacks — and walk you through it. You sign off before we build.

  3. 03

    Build

    Weeks 2 – 4

    We ship in branches. You see it working at week 3, in your sandbox, on your data.

  4. 04

    Run

    Ongoing

    Once live, we monitor, retrain, and iterate. Monthly review, anytime adjustments. Or hand the keys over — your call.

Most clients are live in three weeks.

Average time from first call to first session in production · 2026

06 — FAQ

Things people ask before they book.

You do. The prompts, the schemas, the retrieval setup, the integration code. You can keep us running it, or take the keys and run it in-house. We document either way.
Zapier, Make and n8n are great for connecting steps. They can not reason about messy real-world inputs. The agents we build read intent, decide what to do, hold context across turns, and call those tools when they need to. The two stacks pair well — we usually keep the no-code tools where they earn their keep.
Yes. Some clients come for the agent and stay for the software around it. Some come for software outright — internal tools, dashboards, web apps, integrations. Same studio, same team, same engagement model. We are picky about scope, not picky about whether there is an LLM in the middle.
A working version in your sandbox at week three is the norm. Live with real users between week three and four for a first build. Bigger systems take longer; we tell you that in the assessment.
Your data stays in systems you control. We use providers that offer zero-data-retention modes (OpenAI, Anthropic) and host on infrastructure you approve — your AWS / GCP / Azure, or our managed setup. Nothing is used to train anything.
Every build comes with a written handover doc, a runbook, and access to all the underlying accounts. Two of us know every project. The codebase lives on your GitHub from week one.
The assess call is free. If we do not think the problem is a fit for an agent, we tell you and suggest what is.

07 — Next

Have a problem worth solving?