Build Your Compounding People Network

Most AI agents start every task from zero because they have no shared memory of your relationships. This Blueprint builds that memory: one research pass per contact becomes the intelligence layer your whole AI stack reads from, so meeting prep, intros, follow-ups, and birthday notes all start with full context.

Built for tech leaders and go-to-market professionals who want to build their professional network the way top networkers actually do. This is the team version: every member of your team runs it as themselves from one shared setup, so the whole team draws on one consistent contact memory instead of a dozen private ones, at any company.

The problem: relationships cool and reset to zero because the volume of contacts is too high and CRM systems get out of date. Manual approaches (Salesforce notes, Google contacts, human memory, chiefs of staff/EAs) don't scale to keep full and current context on your key business relationships.

The outcome: relationships compound instead of cooling, because every interaction is anchored in genuine human context: hometown, family, last conversation, the thing they care about that nobody else asks. And your AI assistants share that full context as they help you communicate and follow up with your most important business relationships.

Invoke the skill with a casual natural-language request like "add Sarah Chen at Anteriad to my contacts" or "refresh David Leinweber." The skill loads your configuration, then runs a 12-step pipeline behind the scenes:

Aggregation (parallel). The skill scrapes ten sources simultaneously: Gmail (5-year lookback, extending to 10 years for long relationships), Calendar (past 6 months plus upcoming events), Granola (meeting transcripts), Slack 1:1 DMs (last 50 messages or 90 days), Slack channel mentions, ZoomInfo (verified work history and employment trajectory on a 90-day cache), G2 Marketplace (product reviews and vendor-alias awareness; optional, on only if your company sells software on G2), LinkedIn via Claude in Chrome (live profile check plus connection status, drafted personalized connection requests), Salesforce (dedup by email + name, Account parent-company retry), and Google Contacts (existing record and notes for dedup and refresh).

Synthesis. The skill builds a canonical contact record with structured sections: LAST TOUCHPOINT, LAST UPDATED, TYPE, Education, CAREER, G2 FOOTPRINT, MEETING HISTORY, RELATIONSHIP ARC, PERSONAL / FAMILY, OPEN LOOPS, and RELATIONSHIP.

Write to three audience-tiered destinations. Google Contacts Notes (private, personal context only), JSON sidecar at $COWORK_WORKSPACE/memory/contacts/<email>.json (local cache for downstream skills), and Salesforce Contact fields (FirstName, LastName, Email, AccountId, Title, Description, LeadSource, Lead_Source_Detail__c, Phone, MobilePhone, LinkedIn_Profile__c, visible to anyone at your org with Salesforce access), logged under your own sourcing values. Personal anchors stay in Google Contacts; business context goes to Salesforce; and colleagues on your own company domain are treated as internal and never written to Salesforce.

Output. You get an updated Google Contact with a full structured Notes block, a Salesforce record with verified phone and LinkedIn URL, a cached JSON sidecar for downstream skills (intro drafts, meeting prep, birthday reminders), and a personalized LinkedIn connection-request draft ready to paste if you're not already connected.

Key features:

• Shared agent memory: one contact file that intro drafts, meeting prep, one-on-one prep, and birthday reminders all read automatically.
• Ten-tool research pass: Gmail, Google Calendar, Granola, Slack, ZoomInfo, G2 Marketplace, LinkedIn (via Claude in Chrome), Salesforce, and Google Contacts, orchestrated by the add-contact-team skill.
• Team-ready configuration: each teammate sets their first name, work email, company domain, and CRM sourcing values once; the company domain marks who is internal, so your own colleagues are never logged as leads.
• Audience-tiered writes: personal anchors stay private in Google Contacts; business context syncs to Salesforce for your whole org.
• Anti-hallucination guard: captures only what is literally written in a source; anything less than 95% certain is left out.
• LinkedIn connection drafts: a personalized request ready to paste when you're not yet connected.
• Dedup-safe CRM writes: email plus name matching with Account parent-company retry in Salesforce.

After running the skill, everyone on your team has full structured context on their professional contacts (personal anchors, career arc, recent touchpoints, relationship history) ready for the next call or meeting. Your AI assistant ecosystem inherits that context automatically (intro drafts, meeting prep, birthday reminders, one-on-one prep all read from the same contact file), so both you and your agents stay aligned on how to best engage and collaborate with everyone in your professional network. Relationships compound instead of cooling.

Runs best with the Claude Desktop Cowork interface today to invoke this add-contact-team skill. You need a paid Claude plan plus licenses to the systems you will call via MCP for full relationship context: Salesforce, Google Workspace, ZoomInfo, Slack, and Granola. Nothing about a specific person or company is hardcoded: each teammate sets a small configuration once (first name, work email, company domain, CRM sourcing values), and the company domain is what determines who counts as an internal colleague. Every field has a working default, so setup is fast; the attachment includes a copy-and-paste configuration block and setup steps. The CRM write path is Salesforce-specific, so a team on a different CRM would adapt those steps.

Be conscious that you may be capturing some private information from your communications with contacts. Only store that data in secure systems that meet the user's data privacy rights for their country of residence, your company's data privacy policies, and any industry-specific privacy regulations that govern your business and user relationships. A shallow mode skips the deeper personal-context mining for contacts you barely know or any sensitive internal situation.