Claude Code and agents · Coming soon
Answer your team’s repeat questions from your own docs, with sources
You answer the same questions every week, and the answers already sit in documents nobody finds. You build a Claude Project or custom GPT from those documents, with instructions to cite its source, a test set of your team’s real questions to check it against, and a routine for keeping stale files out.
The course: Claude Projects and custom GPTs as a team knowledge base
$15 $29 early-bird price
This course is being built now. Join the early-access list to reserve the early-bird price and shape what it covers.
Joining the list is free. You only pay if you buy at launch, at the early-bird price.
If this sounds like your week
- The same questions arrive in chat every week, and you answer each one by hand.
- New starters ask you first because the wiki is hard to search and partly out of date.
- You tried a custom GPT and it answered confidently from something that was not in your files.
- Nobody owns the documents, so whatever you load goes stale.
What you do now, and where it stops working
Today the answers live in your head, a wiki and a few pinned messages, and you are the search engine. A Project or GPT with every file uploaded looks like the fix, until it gives one wrong answer with confidence and people stop asking it.
What changes after the course
- A Claude Project or custom GPT holding only the documents that answer your team’s real questions.
- Answers that name the document they came from, so anyone can check them.
- A test set built from questions your team asks, rerun after every change.
- An update routine with a named owner, so the files stay current.
How it works: Questions first, documents second
- 1Collect the questions people ask you, from chat and email.
- 2Choose and clean only the documents that answer them, using the preparation checklist.
- 3Load them with instruction templates that ask for a cited source, and an “I do not know” when the files are silent.
- 4Run your question test set and fix what fails before you share it.
- 5Set the update routine: who refreshes which files, and when.
What you get
- Document preparation checklist
- Instruction templates for Claude Projects and custom GPTs
- Question test set template
- Video walkthrough of a team knowledge base built from scratch
This is for you if
- You lead a team or run operations and field the same questions weekly.
- Your team uses Claude or ChatGPT already, or is allowed to.
- You have documents that answer most of those questions, even if they are messy.
Skip it if
- You need company-wide search connected to live systems. This course builds a focused assistant from a curated set of files.
- Your documents hold information your company does not allow in AI tools. Check your policy first.
- You want it to run with nobody maintaining it. It needs an owner.
Before you join
- What if it gives confident wrong answers?
- That is the main risk, and most of the course. The instructions ask for a source on every answer, you test against real questions before anyone else uses it, and you rerun the test set after each change, so drift shows up in your test rather than in a colleague’s question.
- Claude Project or custom GPT: which should we use?
- Use the one your team already pays for. The course covers both, with instruction templates for each, and the method is the same: curated documents, cited answers, a test set.
- Do I need a paid or team plan?
- Sharing a Project or GPT with colleagues depends on your plan, and those rules change. The course lists what to check on each vendor’s plan page before you commit.
- Our documents are a mess. Is that a blocker?
- No, but they come first. The preparation checklist covers choosing, trimming and structuring files before anything is loaded, starting with the documents behind your most frequent questions.
- Who keeps it up to date?
- Someone has to, and the course makes that explicit: the update routine names an owner and a schedule, and the test set shows when an answer has gone stale.
Rather have it built for you? Describe the project and the same work can be done for your team.