Personal CRM

Most CRM tools are designed for sales teams, not individuals. They require rigid data entry, structured fields, and disciplined logging. Moltbot takes the opposite approach: just tell it what happened in natural language, and it remembers everything.

You: Note — met with Alice today. Her company is building AI customer
service using RAG + Claude. She's interested in our project.
Scheduled a follow-up for next Wednesday.

That is the entire "data entry" step. No forms, no fields, no app switching.

Prerequisites

• Moltbot running and connected to Telegram
• Memory system enabled — This recipe depends entirely on Moltbot's ability to store and retrieve information. See Memory System

How It Works

Moltbot's memory system uses vector embeddings to store information semantically. When you log a note about a person, the content is embedded and stored. When you later ask a question, Moltbot performs a semantic search across all stored memories.

This means:

• You do not need to use exact keywords. Asking "Who works on AI customer service?" will find Alice even if you never tagged her with those exact terms.
• Context accumulates over time. Each new note about Alice adds to her profile. After several interactions, Moltbot has a rich picture of the relationship.
• Queries can be broad or specific: "What does Alice work on?" or "List everyone I met in January who is interested in our project."

Setup

Step 1: Define CRM Behavior in SOUL.md

Add guidelines for how Moltbot should handle contact-related information:

## Personal CRM

When the user shares information about a person they met or interacted with:
1. Extract the person's name, company, role (if mentioned), and key discussion points
2. Note any follow-up actions or scheduled meetings
3. Save to memory with the person's name as a key identifier
4. Acknowledge what was saved with a brief confirmation

When the user asks about a person or their contacts:
1. Search memory for all entries related to that person
2. Compile a summary including: last interaction date, company, role, discussion history, and pending follow-ups
3. If asked for a list (e.g., "all contacts in AI"), search broadly and return matching entries

Step 2: Start Logging Interactions

No special syntax is required. Just talk naturally:

You: Note — met with Alice today. Her company is building AI customer
service using RAG + Claude. She's interested in our project.
Scheduled a follow-up for next Wednesday.

You: Had coffee with Bob from DataCorp. He's their VP of Engineering.
They're migrating from AWS to self-hosted infrastructure.
Might be a good fit for our enterprise tier.

You: Quick call with Carol — she's the one Alice introduced me to.
Works on developer relations at Anthropic. Shared some feedback
on our documentation. Very helpful.

Step 3: Query Your Contacts

Retrieve information naturally:

You: What does Alice work on? What did we discuss last time?

You: What meetings do I have next week?

You: List all my contacts working on AI

You: When did I last talk to Bob? What was it about?

You: Who did Alice introduce me to?

Vector search handles the matching. You do not need to remember exact phrasing from your original notes.

Step 4: Set Up Follow-Up Reminders (Optional)

Combine with Moltbot's reminder capabilities to never miss a follow-up:

You: Remind me to follow up with Alice next Wednesday at 10am.
Include a recap of what we discussed.

When the reminder fires, Moltbot does not just send a generic ping — it retrieves your conversation history with Alice from memory and includes a summary. See Context-Aware Reminders for more details.

Step 5: Weekly Relationship Review (Optional)

Set up a cron job to review your networking activity:

cron:
  - name: weekly-crm-review
    schedule: "0 9 * * 1"
    channel: telegram
    prompt: |
      Review my contact interactions from the past 7 days:
      1. List everyone I logged notes about this week
      2. Highlight any pending follow-ups that are overdue
      3. Suggest anyone I haven't interacted with in over 30 days who might be worth reaching out to
      If no contact activity this week, do NOT send a message.

Example Queries and Use Cases

Before a meeting:

You: I have a meeting with Alice tomorrow. Give me a full briefing:
everything we've discussed, her company's situation, and any
open action items.

Networking event prep:

You: I'm going to the AI Engineering Summit next week.
Which of my contacts might be attending? Who works in
the AI/ML space that I could reconnect with?

Finding introductions:

You: I need to talk to someone who knows about Kubernetes
operations at scale. Do any of my contacts fit, or
know someone who might?

Tracking shared commitments:

You: What have I promised to send or do for people this month?
List all pending action items from my contact notes.

Edge Cases and Troubleshooting

• Name ambiguity: If you know multiple people named "Alice," add distinguishing details: "Alice from DataCorp" vs. "Alice from the conference." Moltbot uses semantic context to differentiate, but explicit identifiers help.
• Outdated information: People change jobs and roles. When you learn new information, log it: "Update on Bob — he left DataCorp and joined NewStartup as CTO." Moltbot stores the new entry alongside the old ones, and when you query, it will present the most recent information.
• Privacy considerations: All contact information is stored in Moltbot's memory system on your server. No data is sent to third-party CRM services. However, be mindful of what you log — Moltbot remembers everything you tell it.
• Bulk import: Moltbot is designed for incremental logging, not bulk data import. If you want to import an existing contact database, you would need to send entries one at a time or write a script to feed them through the API.
• Group conversations: If you use Moltbot in a group chat, be aware that contact notes logged there may be visible to other group members (depending on your configuration).

Pro Tips

• Log immediately after meetings. The best time to record notes is right after a conversation, while details are fresh. A quick voice-to-text message on Telegram works great for this.
• Include emotional context. Notes like "Alice seemed excited about the partnership" or "Bob was frustrated with their current vendor" add valuable context that pure CRM tools miss.
• Use Moltbot to draft follow-up emails. After logging a meeting note, ask: "Draft a follow-up email to Alice thanking her for the meeting and summarizing our next steps." Moltbot will use the context from your note.
• Cross-reference with other recipes. If Alice mentions an article, save it with the Read-It-Later recipe and note: "Alice recommended this article about RAG architectures." Now the article and the contact are linked in memory.
• Periodic cleanup. Every few months, ask: "List all contacts I haven't interacted with in 6 months." Decide whether to reach out or let those connections go dormant.

Related Pages

• Memory System — The foundation that makes the personal CRM possible
• Scheduled Tasks — Automate follow-up reminders and weekly reviews
• Context-Aware Reminders — Reminders that include relationship context
• Weekly Report — Include networking activity in your weekly summary
• Creative Use Cases — More automation ideas