14 features scanned from 2025–2026 real products (Boardy, Intro.co, Lunchclub,
LinkedIn AI, Dex/Clay, PeopleGrove/Hivebrite/Graduway, WhatsApp AI summaries, privacy-preserving
matching) and mapped to Kalsiqi's pillars — Jobs, B2B vendors, Investments + pitch, Emergencies,
Collective wisdom. Full detail + sources: docs/research/2026-10-04-ai-social-features.md
AI proposes a specific intro with a one-line reason; both sides confirm before contact info is shared.
Example: Boardy (double-opt-in brokering), Intro.co marketplace
Fits: Jobs · Investments · B2B vendors — extends Meena's existing intro brokering
Risk: low — matches existing DPDP opt-out posture.
A question gets routed to the 2–3 best-placed alumni instead of broadcast to everyone.
Example: AI event-matchmaking routing pattern (EventHex)
Fits: Emergencies · Collective wisdom — upgrades existing vendor-answers/blood-registry routing
Risk: must not leak why someone was selected if that exposes sensitive profile data.
Free-text query ("fintech founders in Dubai who did ops") returns ranked people, not filters.
Example: LinkedIn's 2026 conversational search (Sales Navigator → global rollout)
Fits: Jobs · B2B vendors · Investments — upgrades the directory search UI
Risk: must respect is_suppressed()/opt-out gating on every result.
Auto-summarize unread group activity into a short "what you missed" + action items.
Example: WhatsApp's own AI unread-chat summaries; ChatBrief, GistGem
Fits: Collective wisdom · Events — Meena already parses group messages
Risk: must honor EXCLUDED_GROUPS and suppressed members.
AI nudges both sides to actually meet after an intro, tracks outcome, asks for an update.
Example: Boardy Pro (June 2026) — "now I make deals happen"
Fits: Jobs — same as the planned Job/CV lifecycle + referral-outcome tracking
Risk: low — status tracking of an already-consented interaction.
AI proactively calls/voice-messages members to refresh what they need and queue intros.
Example: Boardy AI — 237k people called, ~1,500 intros/day
Fits: Jobs · Investments · Collective wisdom
Risk: voice-call fatigue for an India/NRI WhatsApp-first base — needs opt-in windows.
AI pairs two members weekly for a short call based on stated goals, not static overlap.
Example: Lunchclub — weekly curated 1:1 video matches
Fits: Collective wisdom · Jobs — recurring karma-earning loop
Risk: match-quality complaints are Lunchclub's #1 gripe — needs "skip this week".
AI summarizes "what you know about this person" before you message/meet them.
Example: Dex (follow-ups) and Clay/Mesh (auto-context), ~$10–12/mo
Fits: Collective wisdom · Living-profiles mission
Risk: needs an edit/dispute path once members see inferred facts about them.
AI drafts the "ask ABC about X" message in the member's voice; member reviews before sending.
Example: NetworkAI/Wonsulting-style LinkedIn outreach drafting
Fits: Jobs · B2B vendors · gamified "do you know ABC" mechanic
Risk: must stay human-in-the-loop, never auto-send.
Conversational AI gives first-pass career, CV, or pitch-deck feedback, not just routing.
Example: LinkedIn's AI Career Coach in LinkedIn Learning
Fits: Jobs · Investments (monthly pitch pillar)
Risk: must disclose it's AI feedback, not a verified alumnus's opinion.
AI auto-tags low-quality/duplicate posts and routes to the right pillar before a human sees it.
Example: AI-moderated-community triage pattern used across large Discord/Slack communities
Fits: All pillars — protects signal as 34k-alumni volume grows
Risk: false positives silencing real asks — needs an appeal path.
Private "open to a new role / looking for a co-founder" flags stay hidden until both sides match.
Example: Privacy-preserving multidimensional-intersection matching research (2026)
Fits: Jobs (passive seekers) · Investments
Risk: low if the non-match side is never logged.
Incumbent alumni platforms bolting AI matching onto mentorship/career modules.
Example: PeopleGrove Career Access Platform, Graduway/Gravyty, Hivebrite AI personalization
Fits: Validates the "replace IIMBAA app" direction as a baseline, not a build item
Not a build item — track as baseline.
A mini-agent per member answers basic "what does X do / open to" queries when they're away.
Example: Microsoft's internal "DigitalMe" digital-twin-at-work concept
Fits: Collective wisdom · B2B vendors (busy/traveling members)
Risk: highest on the list — impersonation risk; must only use member-approved facts with a visible AI label.