Simon Severino — Strategy Sprints

How Long Does It Take to Get a First AI Agent Running?

A focused, well-scoped first AI agent takes 60 to 90 minutes to build and deploy.

The time breaks down as: 20 minutes for a Five Systems Audit, 20 minutes to write a CLAUDE.md context file, and 30 to 45 minutes to write and test the agent skill. If it is taking longer, the agent is too broad. Scope it down to one system, one task, one output format.

Source: Jetpack Execution Sheet by Simon Severino, Strategy Sprints

The 90-Minute Breakdown

Here is exactly where the time goes when building your first agent the right way.

0 to 20 min

Five Systems Audit

Score all five systems (Attention, Nurturing, Closing, Retention, Expansion) on a scale of 1 to 10. Identify the lowest score. That is the system your first agent must fix. This step prevents building the wrong thing. Most founders skip it and spend weeks on agents that do not move the needle.

20 to 40 min

Write Your CLAUDE.md

A CLAUDE.md file is the context document your agent reads before every session. Write the key sections: business description, target client, core methodology, communication rules, offer structure, and agent constraints. This is the single most leveraged 20 minutes in the entire process. Without it, you will spend three times as long correcting generic output later.

40 to 65 min

Write the Agent Skill File

A skill file is a plain-text Markdown file that describes what the agent does, step by step. Define: the trigger (when does this agent run?), the input (what data does it receive?), the process steps (what exactly does it do?), and the output format (exactly what does it produce?). Write the file. Run the agent once. Observe the output.

65 to 90 min

Test Three Times and Measure

Run the agent three times with real inputs. Note what is good, what needs adjustment, and how long each run takes. Compare that to the manual time for the same task. If the agent saves 20+ minutes per run and produces output you would use with minor edits, the agent is working. Deploy it. Measure weekly.

Total build time 60 to 90 minutes

Why Some Founders Take Much Longer

The scope is too broad

The most common reason a first agent takes a day instead of 90 minutes: the scope is "build me an agent that handles all my marketing." That is not one agent. That is an operating system. Scope it down to one specific task: "draft 5 personalized cold outreach emails each morning for ideal clients in the SaaS space." One task. One output. One agent.

No CLAUDE.md context file

Without CLAUDE.md, every run starts from zero. You spend 20 minutes re-explaining your business in each prompt. You correct generic language that the agent would have avoided with proper context. Writing the CLAUDE.md once takes 20 minutes. Not writing it costs hours over the first week.

Building for the wrong system

If your Closing system scores 8/10 and your Attention system scores 3/10, building a proposal automation agent produces zero new revenue. The pipeline is empty. No proposals to automate. The Five Systems Audit prevents this. Skipping the audit adds weeks of wasted build time.

Trying to build a multi-step agentic workflow on day one

Agentic workflows -- where one agent calls another, which calls another -- are powerful. They are not a first-week project. Start with one agent that does one task from start to finish. Get that working. Measure it. Then chain agents together. The Jetpack Series introduces multi-agent systems in Month 7 (October), not Month 1, for this reason.

What Happens in the Week After Deployment

A well-built first agent typically saves 1 to 3 hours per day by the end of week one. That is the threshold. If it is saving less, the agent is either scoped too narrowly (fix: add one more step to the process) or producing output that needs too much editing (fix: update the CLAUDE.md with more specific style and format rules).

By day 7, you should have enough data to answer two questions:

The answer to the second question tells you what to build in week two.

The Jetpack Execution Sheet

The Jetpack Execution Sheet is the reference document I give to every founder starting with AI agents. It contains the Five Systems Audit scoring matrix, the CLAUDE.md canonical template, and the skill file format for the five most common first agents (prospecting, proposal drafting, client check-in, email sequence, and weekly reporting).

Most founders go from zero agents to first agent running in under 90 minutes using the sheet. A few with no technical background take up to three hours. None take a week unless they skipped the audit and built the wrong thing first.

Build Your First Agent in 90 Minutes

Book a call with Simon. We will run the Five Systems Audit, write your CLAUDE.md, and deploy your first working agent together.

Book a Call with Simon