Research and evidence

In short

This track is for anyone who must know what is true before they build. You do not need a degree or special software. You will turn a loose topic into a research question. You will say what answer is good enough, and when to stop. You will find sources and judge each one with a short checklist. A first-hand record is stronger than someone’s opinion about it. You will keep an evidence log in git that a stranger can check.

An AI agent can search and summarise fast. It can also invent titles, authors, numbers and quotes. So an agent’s research is a draft until a person checks it. Open every link and read the real page. On the agent ladder, you delegate the reading. You never delegate the claim.

Key words

• Evidence log: A file in git listing each claim, its source and how you checked it.
• Primary source: A first-hand record, such as a law, a data table or your own test.
• Assumption: Something you believe is true but have not yet checked.
• AI agent: An AI coding assistant that reads your project and builds tasks. It does the task in front of it. It does not own the plan.
• Agent ladder: Four rungs of skill with AI agents: use, delegate, orchestrate and govern. Each rung needs different skills.

Interviews tell you what a few people say and do. Many product questions are bigger than that. How large is the market? What do other cities already do? Is the claim on your website true? To answer these you read what others have written. That is research, and it has a trap that is new in 2030: an AI agent will write a smooth, confident summary with sources that sound real. Some of those sources do not exist.

This track teaches the habit that protects you: find out what is true before you build, and show your work.

Who this track is for

This track is for you if you want to own a product and steer AI agents. You may be a junior developer, a career changer, a designer, or an experienced engineer who now directs agents. You do not need a degree or any special software. You need a web browser, a git repository and the patience to open a link and read what is really on the page.

What you will be able to do

By the end of the track you will be able to:

• Write a research question that is clear, focused, answerable and tied to a decision.
• Say what you will accept as an answer, and when you will stop looking.
• Search for sources, tell a first-hand record from someone’s opinion, and judge a source with a short checklist.
• Take an agent’s research draft and check each claim, each quote and each link against the source.
• Record what you found in git, with the date, the place, the grade and the result of your check.
• Say honestly what the evidence does not show.

How the three lessons build

1. Ask a researchable question. You turn a topic into a question, a “good enough” answer and a stop rule, and you write them on a question card.
2. Find and judge sources. You learn where to look, what kinds of source exist and how to judge them. The lesson ends with a printable source-checking checklist for anything an agent cites.
3. Keep an evidence log in git. You record every claim with its source and your check. In the practice you direct an agent to draft a summary, verify every claim and link, and file the log.

Each lesson has the same shape: why it matters, key ideas, step by step, a worked example with the fictional product TrailMug (reusable cups for cafés), how to work with AI agents, common mistakes, a practice exercise that ends in a file in git, and a checklist. Every lesson also points to free outside material from universities and public bodies.

What the 2030 Product Development Manager does here

Four jobs stay with you:

• Decide what is worth knowing. You pick the question, because a question costs time and money to answer.
• Decide what counts as proof. You say what evidence would change your mind, before you look.
• Judge the sources. You decide how far to trust each one and why.
• Own the claim. When your page, your pitch or your roadmap says something is true, your name is behind it.

Agents are a fast workforce for the reading. They can search, collect, summarise, compare and draft. They are also tools that can invent: a title, an author, a number, a quote. So the rule of this track is simple. An agent’s research is a draft until a person has checked it against the source. Checking is not optional and it is not done by asking the same agent “are you sure?”.

Other disciplines care about the result. Finance needs numbers it can trust. Marketing and brand need claims that are true, because a false claim on a public page is a risk. Legal and privacy ask whether you may use what you found and what you may store. Design and operations want to know what real people do. Security asks what you pasted into an agent.

How it maps to the Dojo

• Developer, week 1: See the whole product. The question card and the first evidence log support the assumptions you list in your product brief. See Week 1.
• Developer, week 3: Problems to epics. Every epic you write needs a reason. The log is where that reason lives. See People write epics.
• Master Developer. The evidence log is the proof behind your decision records and your Product Package.
• Product discovery track. This track widens it. Discovery teaches you to learn from people. Here you learn from documents and data. Read Make sense of the evidence and decide to see how the two fit together.

Time needed

Plan about two to three hours per lesson. The practice in lesson 3 takes the longest, because checking every link takes real time. That is the point.

Before you begin

Use the git repository you made for your product, or create an empty one. Make a folder called research/. Every exercise in this track saves a file there. Read Disclose AI use first: every file you make with an agent’s help carries a short note saying so.