Infrastructure for Referral Network Intelligence

Referral network intelligence requires documented definitions of what constitutes a referral relationship, how referral sources are categorized (client, alumni, partner), and what tracking fields must be populated in CRM. These process definitions exist at L2. However, the nuanced judgment about which client relationships are warm enough to ask for an introduction — the relationship depth assessment that drives referral likelihood — is entirely tacit partner knowledge that resists formalization.

Referral network intelligence requires systematic capture of relationship interactions that signal referral potential: meeting frequency, project satisfaction scores, tenure of relationship, and actual referral events (opportunity source = 'referral' + referrer contact). Template-driven CRM workflows with required fields for referral source attribution and relationship activity logging provide the foundation. This systematic capture enables the AI to identify referral patterns by relationship type, industry, and engagement level.

Identifying referral paths requires structured relationship data: Contact → Account → Relationship Type → Interaction History → Referral Events. The CRM's Account→Contact schema provides the foundation. Standard industry codes and relationship classification fields enable the AI to compute network distance between a referral source and a target account. Without consistent schema, graph traversal to find warm intro paths — the core output of referral network intelligence — cannot be computed reliably.

Referral network intelligence must query CRM relationship and opportunity records, access alumni network data where maintained, and pull target account information. Modern CRM APIs enable this programmatic access. The system can traverse the relationship graph stored in structured CRM data. However, LinkedIn connections and personal network intelligence — often where the warmest referral paths live — are inaccessible, constraining the network graph to formally recorded relationships.

Referral network quality degrades rapidly as contacts change jobs, relationships cool, and alumni networks evolve. The baseline shows account data decaying without ownership. For referral intelligence, a stale contact record — recommending a warm intro to someone who left the target company 8 months ago — produces actively harmful recommendations that damage partner credibility. Maintenance at L2 (periodic cleanup) means the network graph will contain significant relationship decay between refresh cycles.

Referral network intelligence primarily draws on CRM relationship data and outputs recommendations back to sales teams through CRM or email. Point-to-point integration between CRM and the intelligence engine covers the core workflow. Alumni network data, where maintained, typically lives in a separate HR or marketing system with limited integration. LinkedIn and personal network data isn't systemically accessible. The capability functions with existing point integrations despite this constraint.