Getting Started

Welcome to the BYU AI Foundry. You are joining a delivery team with a paying client. This guide gets you productive in the first two weeks.

Before the Project Starts

Sign the client NDA or the Foundry standard mutual NDA
Accept your GitHub invitation to the ai-foundry-byu organization
Confirm your team, your role, and your QC lead
Set up access to the project repository and approved tools
Complete the AI Foundry baseline survey, the six Builder axes
Read the client’s proposal and any background material provided
Review the Build Standards and the Program Overview

First Two Weeks

Attend the kick-off with the client and confirm the point of contact
Write a one-page restatement of the problem in your own words and send it to the client
Agree with the client on what “done” looks like and how success will be measured
Establish the standing weekly client call and the internal team cadence
Inventory the data you have, the data you need, and the data you are not going to get
Identify the three assumptions that, if wrong, break the whole project

Tip

The most common failure in a Foundry project is building a correct answer to a question the client did not need answered. Spend the first two weeks making sure the question is right. Clients expect you to question the brief in that time.

Team Roles

Every team has the same three roles. They rotate by milestone unless the team agrees otherwise.

Role	Owns
Client lead	The relationship. Runs the weekly call, sends the update, owns scope conversations.
Builders	The work. Analysis, modeling, code, documents.
QC lead	The standard. Nothing goes to the client without QC sign-off.

The QC lead’s sign-off is required. Client-facing work that contains an unreconciled number or an unexplainable model output is a QC failure, and it is graded as one.

Where You Start on a Team

Roles say what you do. Tier says how much of it you own. Your starting tier comes from what you have already built and shipped. Your baseline survey ratings do not set it.

Tier	What it means
Shadow	You are on the project, in the meetings, reading the repo. You do not own a workstream yet
Contribute	You own a piece of the work, with an experienced teammate checking your work
Own	You own a workstream end to end, and the client sees your name on it

Client lead is an Own-tier role. Students at Shadow or Contribute do not run the client relationship in their first engagement.

Tiers move during the semester, and they are meant to. Ask your QC lead what would have to be true for you to move up, then go do that. Nobody stays at Shadow because of a survey answer.

Teams are staffed so that every Builder axis is covered: someone on the team is strong on each axis, though no one is strong on all six. If you are the only person on your team strong on an axis, you were staffed for that axis.

Your Sprints

Every student runs the same two-week sprints, staffed or not. Your engagement’s own dates (kick-off, phase reviews, roll-off) are set with the client and do not change the sprint calendar.

Sprint	Starts	Plan due	Wrap-up due
1	Fri, Oct 23	Mon, Oct 26, 11:59 PM	Fri, Nov 6, 8:00 AM
2	Fri, Nov 6	Mon, Nov 9, 11:59 PM	Fri, Nov 20, 8:00 AM
3	Fri, Nov 20	Mon, Nov 23, 11:59 PM	Fri, Dec 4, 8:00 AM

Set up your Foundry repo

We do this together in studio on October 23.

Create your repo from the builder template, named <your NetID>_foundry, private, under your own GitHub account.
Add Professor Murff (sdmurff) as a collaborator.
Open the repo in VS Code, start Claude Code, and run /sprint-plan. The first time, it asks which course the repo is for: answer AI Foundry.
Commit and push your Sprint 1 plan, and paste its GitHub link into Sprint 1 Plan on Canvas.

If you also take MSB 341, keep the two repos separate.

Your Responsibilities

Confirm expectations and ask clarifying questions rather than guessing
Deliver on schedule. Tell the client about a likely slip as soon as you see it
Come prepared to client calls with an agenda, progress, and specific asks
Raise blockers within 48 hours. Not at the weekly call, and not at the end of the sprint
Protect client data per the confidentiality terms in the Syllabus
Run every sprint: plan by Monday, review and Loom demo by the Friday it ends. This is individual and separate from the team’s client update
Carry your share. Your teammates grade you on this

What You’ll Be Working On

Foundry teams ship software. Depending on the engagement, that includes:

Full-stack applications. Net-new web and mobile products, B2C, B2B, or internal tools, built end to end and deployed to production
AI integration and orchestration. Agents and copilots over the client’s own data, automations across Slack, Notion, and their existing stack, document processing, extraction, routing, and summarization
AI enablement. Setting the client’s own team up on current agent tools such as Claude Code, working alongside their staff, then handing off

Discovery work comes first: where AI helps this business, what is worth building, what is worth automating. The purpose of discovery is to decide what to build. The engagement ends with software the client is running.