The AI-native operating layer that sits on top of SAP, Tally and other systems you already run, hands back the decisions — named, priced, ready to send.
Module 01 is Crestline Core — the org graph + memory + workflows shipping today. The other twelve are verticalised intelligence layers that sit on top of your existing systems of record. Same memory, same graph, same console — different surface per buyer.
The org graph, four-layer memory, decision register, and workflow engine. The base everything else sits on.
AI-native people ops on top of your HRMS — not replacing it.
Enterprise PM with AI risk-detection + cross-functional impact maps.
Forward-looking financial intelligence on top of your accounting stack.
Revenue intelligence with organisational context — on top of your CRM.
People, departments, managers, reporting lines — typed FKs with referential integrity. Tokens like @manager_of(Sarah) and @dept-finance resolve live against the graph.
Membership.managerId is the reporting-line foreign key. Cycle-safe writes, hierarchy-aware row-level access — permissions never configured separately.
@manager_of(x) · @dept-finance · @reports_of(x) · @cxo · @self. Same shape across workflows, messages, and the Constitution.
Every change to the graph audited. asOf=2025-12 returns the org as it was six months ago. Required for compliance and forensic audits.
Every query fans out across all four memory layers in parallel. The Hub synthesises a single answer with inline citations to the exact memory items that grounded each claim.
Upload an event log — or connect SAP, Salesforce, Tally. Crestline reconstructs the actual process, quantifies every bottleneck in real money, and names the people who can fix it. Grounded in your verified org graph — anti-hallucination by construction.
Describe automations in plain English — Hindi works too. Five agents stream a validated spec. @manager_of(x), @dept-finance, @cxo resolve against the verified graph at runtime.
Cmd-K opens the Console from anywhere. Slash commands run deterministic queries against the graph; free-form questions enter an agentic ReAct loop that plans, calls tools, and cites memory items inline.
Describe a tile in English — the AI Director emits a validated CustomTileSpec and the runtime renders it. Cockpits compose, remix, and persist. Not codegen — composable spec runtime.
Geography-aware query refinement → parallel Tavily fan-out → optional URL extract → structured ranked candidates with tradeoffs and a recommendation. Replaces 'one summary + action items' with something you can actually act on.
A multi-agent pipeline profiles every column, hypothesises, computes locally, visualises, and narrates. The LLM never sees rows in bulk — it picks the chart and writes the story. Server-side stats, AI only narrates.
ARR by segment, last 12 months, ₹ Cr — only BFSI still double-digit.
Crestline Meet runs in-app on LiveKit. Crestline Scribe joins as a real participant — listens, transcribes via Deepgram, and files every decision · task · blocker into the right place in real time. Closing the meeting publishes the notebook to org memory.
Policies, SOPs, MOMs, POs, Templates — authored by org or dept admins, read by everyone in scope. Every publish snapshots a hash-chained version; every body auto-embeds into the right memory layer.
Vendor approvals > ₹40L require CFO + CTO sign-off. Concentration ceiling: no single vendor > 18% of OpEx without board note.
1. Scope. Applies to all procurement requests above the policy threshold issued by Crestline Org members from any department.
2. Approval cascade. Triggers route through @manager_of(@self) → @dept-finance → @cfo.
Multi-agent productivity cockpit + trainer + analyst + skill-path designer. Five parallel agents stream chat, learning, preferences, artifacts, and inline cards. @me.signature, @me.po-template resolve across Messages, Console, and Workflows.
Straight answers on what an Organisational System of Record is, how Crestline AI compares to Glean, Celonis and n8n, and what it costs to run on top of SAP, Tally or Salesforce.
30-minute walkthrough with a founder. We bring real data — yours or ours. You walk out with one named bottleneck and a ₹ number against it.