• 151

    Projects submitted

  • 109

    Published projects

  • 47

    Comments left

  • 66

    Number of votes

  • 52

    days until voting ends

← Back to list

KautilyAI: an agentic counsel for the C-suite & Business Leaders

  • Project manager on customer side

    Tushar Tiwari

    LTM

    Head of Data & AI

  • Category

  • Nomination

  • Goals

    1. Give LTM's CXOs one conversational, governed source for decisions across 6+ business domains: FP&A, Collections, Finance, Workforce, HR and Sales.

    2. Anchor every answer to a single certified catalog of 291 KPIs with traceable definitions, source systems and calculation logic.

    3. Cut the time to a cross-functional leadership view from 2-3 weeks to about 30 minutes

    4. Reduce monthly and quarterly business review preparation from 4 days to 1 hour per business unit.

    5. Enforce role-based access so each leader sees only their own scope, with responsible-AI controls (disclaimers, refusal rules, red-team testing) built in from day one.

  • Timeline

    march, 2026 — august, 2026
  • Project completion year

    2026

  • Project scale

    58 automated workplaces
  • Results

    Results

    - 58 business users across finance, delivery, growth, HR and business leadership

    - 3,171 production requests in the first 10 weeks, peaking at 683 in a week

    - Cross-functional leadership view: from 2-3 weeks to about 30 minutes

    - Monthly and quarterly business reviews: dashboards and presentations generated in the platform, cutting effort from 4 days to 1 hour per business unit each month

    - Passed CISO, Data Privacy and Responsible AI review

  • Project uniqueness

    KautilyAI is not a chatbot on top of dashboards; it is an AI counsel that tells leaders what a number means and what to do next. Every answer follows the same executive structure: a verdict, key risks, recommended actions and openly declared data gaps.

    What makes it unusual:

    - Governed by design: the AI only answers from a certified 291-KPI catalog and 163 verified queries, and refuses KPIs it cannot compute instead of guessing.

    - Ambiguity handled explicitly: a three-tier routing model settles metrics with competing definitions, such as headcount and revenue.

    - Nine role lenses (CFO, CEO, delivery, sales, HR, board and others) and three modes (Operator, Executive, Governance) shape answers for each leader.

    - Security travels with the question: row-level security on 54 data objects, so each leader sees only their scope.

    - Next step already in build: an agent-to-agent mesh where domain agents consult each other with no master agent.

  • Software used

    Snowflake (Cortex Agents, Cortex Analyst, semantic views, row access policies); Anthropic Claude models via Snowflake Cortex; Microsoft Azure (web application hosting, Azure single sign-on); source systems: SAP ERP, Salesforce CRM, HRMS, resource management, procurement and risk platforms; existing enterprise data lake.

  • Solution from the Global CIO catalog

    The project does not use solutions from the Global CIO catalog

  • Project implementation complexity

    - Data from more than a dozen source systems had to be reconciled into one certified catalog of 291 KPIs; the same KPI was often defined differently across reports.

    - Competing definitions (for example several headcount and revenue sources) required a three-tier routing model so the AI picks the right one, or discloses its default.

    - Preventing invented figures: the agent must refuse KPIs it cannot compute, backed by 163 verified queries and a red-team rubric (sycophancy, pressure to fabricate, mode leakage, governance integrity).

    - Role-scoped access across 54 data objects, so CXOs see only their own scope.

    - Passing CISO, Data Privacy and Responsible AI review for an AI that advises the C-suite.

    - Delivered from prototype to leadership pilot in about five months by reusing the existing data lake, security policies and earlier prototype agents.

  • Description

    Business challenge

    LTM's leadership relied on separate MIS reports drawn from more than a dozen enterprise systems (ERP, HRMS, CRM, resource management, procurement, risk). Cross-domain questions, such as how margin, receivables and workforce utilization interact for a business unit, took 2-3 weeks of manual reconciliation, and the same KPI was often defined differently in different reports. Leaders needed one trusted source that explains why a number moved and what to do next.

    Solution

    KautilyAI ("The Enterprise Agentic Council") is built natively on Snowflake, so enterprise data never leaves the governed platform. A Snowflake Cortex Agent sends each question to six domain agents (FP&A, Collections, Finance, Workforce Management, HR, Sales), each running on a governed semantic view. A three-tier routing model resolves metrics with competing definitions. Nine role lenses and three operating modes shape every response into a verdict, key risks, recommended actions and declared data gaps. Users sign in with Azure single sign-on, and row-level security limits each leader to their own scope. Every change moves through development, QA and production environments with twice-daily automated change tracking.

    Scale

    - Governed catalog of 291 enterprise KPIs

    - 6 semantic views with 136 metrics, 338 facts, 1,764 dimensions and 163 verified queries

    - Row-level security on 54 data objects

    - Built for five C-suite personas: Finance, Delivery, Growth, HR and Business

    Timeline

    Prototype March 2026; controlled leadership pilot from August 2026; production governance sign-off October 2026.

    Importance and next steps

    KautilyAI is now the foundation for LTM's AI-driven Unit Scorecard, where 17 of 28 scorecard KPIs are already available. Next: five more domains (Accounts Payable, Bid vs Delivery, Contracts, Attendance, Enterprise Risk) and an agent-to-agent mesh in which domain agents consult each other directly, with the user's permissions carried through every hop.

  • Project geography

    India

  • Additional presentations

Comment

Only authorized users can comment.
We suggest you or register.

  • Customer

    LTM

We use cookies for analytical purposes and to deliver you the best experience with our website. Continuing to the site, you agree to the Cookie Policy.