Individual AI accelerates the person with the prompt. Institutional AI accelerates the venue. Thynk is the institutional layer underneath — Salesforce-native, five-portal, audit-traceable. Any AI that runs on it inherits coordination, governance, and venue-domain depth that generic tools cannot replicate.
Want the longer version? Read the manifesto →
In March 2026, the venture investor George Sivulka — a16z partner, founder of Hebbia — published the now-canonical essay distinguishing Institutional AI from Individual AI. It has become the most widely-circulated AI framework of the cycle.
Seven dimensions separate the two. We build for every one of them.
Sivulka's framework on the left. Thynk's implementation on the right. Each underlying capability is real, not a roadmap commitment.
Three reps, three AIs, three contexts. Sales, ops, exhibitions accelerate independently. The venue gets nothing.
An inbound RFP triggers a coordinated chain across sales, operations, exhibitions, suppliers and finance — on shared Salesforce data, on the same permission model, with one audit trail.
Underlying: the Salesforce hospitality data model. 212 objects. 4,173+ fields. Five portals reading and writing the same records.Generates more output — more proposals, more emails, more reports, more BEO drafts. Volume goes up. So does slop.
Filters every inbound — Cvent inquiries, Outlook emails, ESC complaints, ops incident logs, occupancy data — and surfaces what the team should be looking at before they ask. Deterministic. Auditable.
Underlying: Salesforce CRM Analytics + AI Co-Pilot. Pickup pace alerts, BoB anomaly detection, late-order surge prediction.Confirms the prompter's belief. RLHF-trained sycophancy. The rep who already thinks the deal will close asks the AI; the AI confirms.
Runs against firm-wide objective scoring rules. Lead scores reflect documented qualification criteria — not the optimism of the user holding the keyboard. The disciplined-denier role.
Underlying: Salesforce lead-scoring framework, DNQ rules, ceiling rules, displacement thresholds — institutional, not individual.Generic foundation model. Broad. Not deep on hospitality.
Grounded in the deepest publicly-documented venue ontology. Knows what an ESC late order is, what a TRF report is, what a multi-event booking with sub-events looks like, what a German Bietererklärung differs from a UK proposal in.
Underlying: Thynk's open hospitality data model — 212 Salesforce objects, public OpenAPI specification."I drafted the proposal in two minutes." Productivity. A vanity metric.
"The proposal closed two days faster. RFP conversion rose 32% on first-response-under-four-hours leads. Exhibitor late-orders dropped 60%." Outcomes. Revenue moved.
Underlying: 79% of MICE RFPs are won by one of the first three responders. AI Co-Pilot + Thynk Direct sold against revenue outcomes — not productivity.Empowers whoever has the credit card and the licence.
Encodes firm processes. Empowers everyone the venue depends on — ops staff with no software licence, exhibitors with no CRM training, suppliers with no portal account, finance teams with no AI literacy.
Underlying: five portals — CRM, Direct, ESC, Supplier, Ops. Magic-link or SSO. No Salesforce licence required for floor staff, exhibitors, or suppliers.Waits to be asked.
Acts on what nobody asked. "This exhibitor is two days from a late-order surcharge — here's the draft reminder." "This BEO has changed since the kitchen acknowledged it — here's the diff."
Underlying: AI Co-Pilot's automated triage; Ops Portal push notifications; ESC late-order detection; CRM Analytics anomaly detection.Most venue AI runs on a proprietary database the customer cannot inspect. Thynk runs on Salesforce — the world's most documented enterprise SaaS data layer. Your AI runs on your data, inside your perimeter, governed by your permission model.
Coordination breaks the moment a workflow leaves the AI's data layer. Thynk's five-portal architecture closes every loop on the same data model — sales, booker, exhibitor, supplier, floor-staff member all act inside the same platform.
212-object hospitality data model is the deepest venue ontology publicly available. Any AI agent retrieving against it has institutional venue knowledge a generic LLM cannot match.
The delivery team brings years of Momentus expertise across delivery, project leadership and sales. When you migrate twenty years of venue history, you migrate it with people who built the system you are leaving.
What your procurement team will ask, answered.
Your data does not train Anthropic, Salesforce, or any model you have not chosen.
Every AI action is recorded on the record. Show your auditor.
Every AI draft requires human review before sending. You set the threshold.
AI processing inside the EU Salesforce perimeter. No exfiltration.
AI inherits the same access controls as the human user it acts on behalf of.
Salesforce Einstein and Anthropic Claude today. Other models tomorrow. The institutional layer does not change.
Indicative outcomes from Thynk customers. Yours will depend on baseline maturity, configuration, and adoption.
Bring your hardest exhibitor escalation, your busiest sales week, your most complex multi-venue rate question. We will show you what coordinated AI looks like on real data, with the exact governance you would run in production.
Operate hotels too? See the Thynk hotels platform →
