Behind the Build
How We Scope a Custom AI Agent in One Week
Before we write a single line of code, we spend a week figuring out exactly what we're building and why. Here's what that looks like.
One of the most common things we hear from prospective clients is some version of: "We tried to build this internally and it took six months and we're still not sure it works."
The reason is almost never technical. It's scoping.
Building an AI agent that actually gets used requires answering a set of questions that most teams skip because they're eager to start building. At Revival Group, we don't write code until those questions are answered. Here's the process we run in the first week of every engagement.
Day one is about the human workflow, not the AI. We spend the first conversations understanding what a person is doing today — step by step, tool by tool. We want to see the actual inputs: the email thread, the spreadsheet, the CRM record. We want to understand what "good output" looks like and how the person using it will know it's good. Most importantly, we want to understand where things go wrong in the current process and what the cost of those failures is.
Days two and three are about defining the agent's job description. An AI agent, like a human employee, works best when it has a narrow, well-defined role. We draft a one-page brief that describes: what the agent receives as input, what decision or action it takes, what it produces as output, and what it should do when it's uncertain. This brief becomes the foundation for everything that follows.
Day four is about integration. We map out every system the agent needs to touch — where it reads data, where it writes, what APIs are involved, what authentication looks like. Integration is usually where timeline estimates go wrong, so we surface all of this before committing to a build plan.
Day five is a scope confirmation and kickoff. We present a proposed build plan with a clear definition of done, a delivery timeline, and an honest list of assumptions. The client either confirms or we iterate. We don't start building until both sides are aligned on what exactly is being built.
This process sounds deliberate because it is. The clients who see results fast are almost always the ones who invested time upfront getting the scope right.
If you're trying to figure out what to build and where to start, that first conversation is free.