How to Build an AI CRM That Connects People to Ideas
For sales and business development professionals · Based on Matt Giaro AI Second Brain Build
// TL;DR
Sales and business development professionals meet dozens of people at conferences and lose the context that makes follow-ups meaningful. This framework builds an AI-powered CRM in Obsidian and Codex where the CRM pillar is the primary value driver. After each event, say 'Add to CRM: [Name] — met at [Conference], discussed [topic]' and the system creates a contact record cross-linked to wiki pages on that topic. Months later, ask 'What did I discuss with [Name] about [topic]?' and get an answer drawing from both the contact record and your knowledge base. Use it to follow up with genuine relevance, not generic outreach.
Why do great conference conversations lead to weak follow-ups?
You have a fantastic 20-minute conversation at a conference, exchange cards, and fully intend to follow up. Three months later you remember the name but not the topic, the context, or why the connection mattered. Your CRM has a name and an email — nothing that makes a follow-up feel personal. This is the Dumping Ground Problem applied to relationships.
The AI Second Brain Build fixes it by making the CRM an active, cross-linked layer. Contacts aren't just records — they're connected to the ideas you discussed, so follow-ups carry real substance.
How do you set up a second brain where the CRM leads?
The Wiki/Knowledge Base remains the gravitational core, but for sales professionals the CRM pillar is the primary value driver. Your three pillars: Wiki (your store of industry knowledge and topics), CRM (contacts and events), and Journal (reflections on deals and conversations).
Setup:
1. Install Obsidian, create a vault, and note the folder path.
2. Add the Obsidian Web Clipper to clip industry articles, thought-leadership pieces, and conference talk recordings into RAW — these become the wiki pages your contacts link to.
3. Open the vault in Codex, build the Karpathy wiki architecture, and configure agents.md with your CRM rules.
Seed the wiki with content on the topics your market cares about so there's something for contacts to cross-link to from day one.
How do you capture a contact after an event?
Open Codex and say: 'Add to CRM: Priya Nair — met at SaaStr, discussed usage-based pricing, VP of Revenue at a mid-market SaaS, follow up about their renewal cycle.'
The system creates or updates a file in /CRM named after Priya, capturing contact details, how and where you met, relationship context, and notes. It maintains /CRM/index.md alphabetically with a short bio, logs the update in log.md, and — crucially — cross-links Priya to any wiki pages on usage-based pricing.
How do you prepare for a follow-up months later?
This is where it pays off. Prompt: 'What did I discuss with Priya about usage-based pricing?' The Grounded Response draws from both her CRM record and the relevant wiki content, giving you: the exact context of your conversation, her role and renewal timing, and the latest thinking you've saved on the topic.
You walk into the follow-up able to say 'When we spoke at SaaStr about usage-based pricing and your renewal cycle, you mentioned...' — the kind of specificity that turns a cold follow-up into a warm one.
How do you keep the system current without manual admin?
Set up the hourly automation in Codex so new articles you clip get processed automatically, keeping your knowledge base fresh on the topics your contacts care about. Connect a private GitHub repo for hourly backups — your relationship intelligence is too valuable to lose.
Use the Obsidian graph view to see your network form: contacts clustering around the topics they care about reveals patterns in your pipeline and which themes connect the most people.
Next step: Create an Obsidian vault, clip three articles on your market's hottest topic, then log your three most recent conference contacts into the CRM and test a follow-up query in Codex.
// FREQUENTLY ASKED QUESTIONS
How is this different from Salesforce or HubSpot?
Traditional CRMs store structured contact fields but don't connect people to the substance of your conversations or your evolving knowledge. This framework cross-links each contact to wiki pages on the topics you discussed, so a follow-up query returns both the relationship context and the latest relevant thinking you've saved. It's local-first, fully customisable via agents.md, and grounded in your own content.
Can I bulk-add contacts after a busy conference?
Yes. You can add contacts one at a time with the 'Add to CRM' trigger, or prompt Codex to process a list. Each entry creates a cross-linked contact file and updates the alphabetical index. For best results, include how you met, the topic discussed, their role, and a follow-up note so the system can cross-link to the right wiki pages.
What if I want to track deals, not just people?
Customise agents.md to add a deals or opportunities layer alongside the CRM. Since the system is entirely prompt-driven, you can define new folder categories and entity types without rebuilding anything. Link deals to the contacts and topics already in your wiki so you get a connected view of people, conversations, and pipeline all grounded in your saved knowledge.