Tellius × Snowflake Technology Partner

Tellius × Snowflake

Turn Snowflake data into finished, deterministic work.

Q: Does my data leave Snowflake?

No. Tellius runs in-perimeter — it reads governed data, reasons on Snowflake's own AI, and writes results back, all inside your account. Because the data never moves, the AI-governance and security review is typically shorter.

Q: How does Tellius use Snowflake Cortex and Semantic Views?

Tellius consumes Semantic Views as the shared meaning layer rather than rebuilding them, and reasons on Cortex models inside your perimeter on committed Snowflake spend. It builds semantics where none exist and respects the ones you already govern in Horizon.

Q: What does "deterministic" actually mean here?

It means the same defensible answer every time the workflow runs on the same data. Tellius gets there by validating the data before it computes (null, join-key, cardinality, double-counting, fan/chasm-trap checks and more) and by locking the workflow specification through Missions before execution — so there's nothing left for a model to reinterpret at runtime.

Q: Can we buy Tellius through Snowflake?

Yes — Tellius is available through the Snowflake Marketplace, so a purchase can draw down existing committed spend and collapse procurement from months to weeks.

Tellius brings agentic domain depth to Snowflake's data, AI models, governance so you get finished, deterministic, analytically sound work + actions — across pharma commercial, CPG, FP&A, and RevOps.

Book a demo See how it works →

Kaiya

How can I help, Sanjeev?

"Where is our fleet margin-at-risk concentrated, and what's driving it?"

Iteration 1

Step 1Data agent · align IQVIA, P&L and booking feeds

Step 2Semantic layer · resolve margin tier, department & voyage defs

Step 3SQL · decompose revenue: tier → department → category

Step 4Python · rank drivers; quantify low-margin concentration

Step 5Validation · re-run on locked spec; trace every number to source

Step 6Synthesis · draft the brief & build the export

Fleet margin-at-risk — where the leak is

5.0% of $15.33B gross revenue sits in the low-margin tier — but 100% of that exposure originates from just three departments: Entertainment, Retail, and Dining.3Live from SnowflakeDEMO_DATASETS.MSC_FLEET.REVENUE_BY_DEPARTMENT All other departments contribute zero low-margin revenue.

Low-Margin Revenue · Fleet Total

$771.7M

5.0% share · +0.05 pp YoY (2023→2024)

Duty Free Fragrances$298M

Casual Dining Extras$254M

Kids Club Premium$219M

Key observations

The leak is narrow but deep — 3 departments, 3 categories, 100% of fleet low-margin revenue.7Live from SnowflakeDEMO_DATASETS.MSC_FLEET.MARGIN_TIER_DETAIL
Fleet low-margin share is structurally stable, oscillating 4.95–5.04% for 22 of 24 months — the lone spike is Sept 2024 (+0.83 pp).8Live from SnowflakeDEMO_DATASETS.MSC_FLEET.LOW_MARGIN_SHARE_MONTHLY
Itinerary mix is the lever: shifting toward Extended & Grand Voyage formats lowers fleet-wide at-risk share.

Fleet_Margin_at_Risk.pptx

Synthesis · 9 slides · generated by Kaiya

Opening…

Ask Kaiya anything about your data…

Deep Insight

bv_MSC_Cruises_BV ⌄

Fleet_Margin_at_Risk.pptx — PowerPoint SLIDE 1 / 9

FLEET MARGIN-AT-RISK

$771.7M

Low-Margin Revenue — Fleet Total

5.0% of $15.33B gross revenue sits in the low-margin tier — but 100% of that exposure originates from just three departments: Entertainment, Retail, and Dining. All other departments contribute zero low-margin revenue.

$15.33B

Fleet Gross Revenue

5.0%

Low-Margin Share

+0.05 pp

YoY Change 2023→2024

The leak is narrow but deep — 3 departments, 3 categories, 100% of fleet low-margin revenue.

Source: Fleet Margin Risk Analysis 2023–2024 | Margin tier = categorical variable (Low / Medium / High)

Fleet Revenue by Margin Tier ($B)

20151050

High-MarginMedium-MarginLow-Margin / MAR

Low-Margin Revenue by Product Category ($M)

Kids Club Premium

219

Casual Dining Extras

254

Duty Free Fragrances

298

0100200300400

How can I help, Sanjeev?

Ask Kaiya anything about your data…

Deep Insight

bv_MSC_Cruises_BV ⌄

DEMO_DATASETS / SUN_PHARMA / HCP_SALES_PERFORMANCE ··· Load Data

Table DEMO_INGEST_ROLE 3 months ago 300K 5.9MB

Table Details

Columns

Data Preview

Copy History

Data QualityPREVIEW

Lineage

DEMO_WH 100 of 300K Rows · Updated just now

	HCP_ID	HCP_NAME	HCP_SPECIALTY	BRAND	TERRITORY	NBRX_QTD

CoCo

How can I help?

Show me what CoCo can do

Recent chats

What Can This Agent Do

2 days ago

Help Finding Tables and Building SQL Query

2 days ago

Tellius MCP · deep_insight

authenticate → app.tellius.com

map HCP_SALES_PERFORMANCE → bv_HCP_Sales semantic layer

run deep insight · NBRx decline · Northeast · QTD

6 steps complete · 3 drivers ranked · traced to source

NBRx −12% · Northeast · QTD. The Levulan decline traces to three quantified drivers:

−3.1 pts

PA rejections

a top commercial plan moved the brand to non-preferred mid-quarter

−2.2 pts

Copay expiry

the support card lapsed for ~40% of commercial scripts

−1.4 pts

Competitor launch

a new entrant took early share across three key territories

Data sourcesIQVIAMMITVeeva CRM

Ask, build, @ for context, / for skills

Plan

Auto ⌄

CoCo can make mistakes, double-check responses

Trusted by the world’s most innovative teams

Finished work in action

Ask "why". Get back finished work.

Kaiya plans a Mission, runs it on your Snowflake data, and returns the brief — ranked drivers, recommended actions, lineage, written back to the warehouse. Same method, every vertical.

Why did NBRx decline 15% in the Northeast?

Mission · auto-planned · 6 steps

Iteration 1

Data agent · dedupe & align lagged IQVIA, MMIT & Veeva feeds✓
Semantic layer · resolve NBRx, decile & region definitions✓
SQL · decompose the drop: region → payer → prescriber decile✓
Python · rank drivers; quantify each in NBRx points✓
Validation · re-run on the locked spec; trace every number to source✓
Synthesis · draft the brief; push next-best actions to Veeva✓

NBRx −12% · Northeast · QTD

PA rejections−3.1 pts

Copay expiry−2.2 pts

Competitor launch−1.4 pts

PA rejections — a top commercial plan moved the brand to non-preferred mid-quarter
Copay expiry — the support card lapsed for ~40% of commercial scripts
Competitor launch — a new entrant took early share across three key territories

Monday_Regional_Brief.pptx

Data sources IQVIA MMIT Veeva CRM

Better together

Snowflake is the foundation. Tellius agents build work on top.

Snowflake unifies/governs the data + AI. Tellius is the intelligence layer on top — knows your business, investigates the why, and ships finished work.

Snowflake

Governed, AI-ready data cloud

Unified, governed data — IQVIA, Veeva, claims, payer, CRM, syndicated, all in one place
Native AI — Cortex models in your perimeter
Governance — Horizon, RBAC/RLS, lineage, audit
Open semantics — Semantic Views as the shared meaning layer

Tellius

Domain reasoning, finished agentic work

Domain — AI agents act like seasoned pharma/CPG/finance analysts
Deterministic — repeatable, defensible answers every run
Delivery — finished briefs, decks, and actions — not just chats
Open — portable reasoning, callable from any agent via MCP

Snowflake brings the data, AI models, governance, and the agentic control plane together for the enterprise. Tellius adds the domain depth that turns all of it into finished, deterministic work — diagnostic when you need the why, deterministic when the work repeats, delivered as finished work and action.

Why Tellius

Vertical depth to Snowflake's horizontal scale

General-purpose agents give generic answers. Dashboards get stale. Tellius speaks your business language out of the box.

Domain

It already knows your business

Purpose-built for pharma commercial, CPG, and finance — it knows TRx and NBRx, formulary access, category share, and GL structure, and it works the way your best analysts do. No prompt-engineering the business. The context and memory compound with every Mission, so it gets sharper the more it runs.

Speaks your business — out of the box

TRx / NBRxformulary accessprescriber decilecategory shareTDP / ACVGL structureplan vs. actuals

Context + memory compound every Mission

Delivery

Finished work, not prompts

The unit of output is a completed brief, deck, or recommended action — generated and written back, not a chatbot turn that suggests and waits. Triggered from CoWork, Slack, the field app, email, or a schedule; the work arrives done.

Monday_Regional_Brief.pptx

Delivered where the team already works

CoWorkSlackField appEmailSchedule

Deterministic

The same defensible answer, every single run

Ask a general-purpose agent why a number moved and you'll get a different story each time — it reinterprets the gaps in your request on every pass. Tellius computes the answer instead of narrating it, on two layers of rigor.

Why AI workflows drift — and how to make them deterministic →

Data & analytics validation before it reasons

Tellius checks the data before it computes anything — so the inputs aren't quietly wrong.

null analysistype compatibilityjoin-key compatibilitycardinalityprimary-key uniquenessrow-count integrityjoin-explosiondouble-countingfan & chasm traps

Consistency via Missions + Skills when it runs

Describe the workflow in plain language; Tellius confirms its interpretation, locks the specification, then runs it on a deterministic engine — identical every pass, with lineage on every number.

describe→confirm→lock→run

Open

Open across your ecosystem

Tellius consumes open semantics (Semantic Views, Horizon, dbt), is callable from any agent via MCP or API, and brings hundreds of connectors and syndicated feeds to put more data behind your Snowflake decision. Your reasoning is portable — encoded once.

Semantic Views · Horizon · dbt

↓ consumes open semantics

Tellius reasoning encoded once

↑ callable via MCP / API

Cortex · CoWork · any agent

No data movement Governance inherited (RBAC/RLS) Warehouse-scale performance Audit-ready lineage Snowflake Marketplace billing

Computed, not narrated

Trusted Domain-specific Insights

Numbers must survive re-runs and audits to be trustworthy. Ask the same thing 5 times differently, get one trusted answer.

Before Tellius

Generalist AI agents narrate a plausible cause

Estimates the number — "payer access, roughly 12 of 15 points" — without ever computing it.
Skips the messy-data work — duplicates, restated history, lagged feeds — so the inputs are quietly wrong.
Gives a different answer each run — nothing to take to the retailer, payer, or board.
No lineage when compliance or audit asks how the number was produced.

With Tellius

Computes the defensible one

Reconciles and validates sources first — dedupes, aligns lags, checks joins across IQVIA, MMIT, and more.
Runs the full driver search with statistics and machine learning — not just SQL — across every factor at once.
Computes the bridge: the formulary shift accounts for 11 of 15 points. Calculated, not guessed.
Same answer every run, with lineage attached and the recommended action included.

A semantic layer gives AI agents your vocabulary.
Tellius gives them your judgment.

How it works

Tellius runs inside Snowflake

Tellius reads governed data, leverages Snowflake's own AI inside your perimeter, and writes every decision back. You get finished work.

01Read

Governed data & semantics

Reads the warehouse and consumes Horizon and Semantic Views as the shared meaning layer — building semantics where none exist.

02Run

On Snowflake's own AI

Reasons on native AI like Cortex inside the customer's perimeter, on committed Snowflake spend.

03Serve

Callable from your agents

Tellius is an MCP server and optionally a full UI. Cortex and CoWork agents call it for the hard part and get back the why, the recommended action, and the lineage.

04Record

Decisions written back

Every decision lands in the warehouse as a governed, auditable record — then fans out to Veeva, planning tools, Slack, and email.

05Govern

Security posture inherited

Row-level security, audit trails, in-perimeter processing. The AI-governance review gets shorter because the data never moves.

06Transact

Bought where you buy

Available through the Snowflake Marketplace, drawing down committed spend — collapsing procurement from months to weeks.

Proven in production

Finished work, measured by the customer

Results from Tellius deployments on the data cloud — reported by the teams who run them.

3×

faster category insights

95%

less time on analysis (20 hrs → 30 min)

10→2

days to insight for field teams

275%

adoption growth across field + home office

Now the answer is on the table before Monday's call — so the team spends its time deciding what to do, not chasing the number.

Sanjeev Mankar · Senior Director, Business Applications / Data & Insights · Novo Nordisk

Common questions

Tellius on Snowflake, answered

Does my data leave Snowflake?

No. Tellius runs in-perimeter — it reads governed data, reasons on Snowflake's own AI, and writes results back, all inside your account. Because the data never moves, the AI-governance and security review is typically shorter.

How does Tellius use Snowflake Cortex and Semantic Views?

Tellius consumes Semantic Views as the shared meaning layer rather than rebuilding them, and reasons on Cortex models inside your perimeter on committed Snowflake spend. It builds semantics where none exist and respects the ones you already govern in Horizon.

How is this different from Cortex Analyst or a general-purpose AI agent?

A general-purpose agent is horizontal and probabilistic — great for broad questions, but it narrates a plausible cause and can give a different answer each run. Tellius is vertical, deterministic, and finished: it knows your industry, validates the data first, computes the driver decomposition the same way every time, and returns a complete brief with lineage. Tellius complements Cortex and CoWork — it's the callable layer they reach for when the answer has to be defensible.

What does "deterministic" actually mean here?

It means the same defensible answer every time the workflow runs on the same data. Tellius gets there by validating the data before it computes (null, join-key, cardinality, double-counting, fan/chasm-trap checks and more) and by locking the workflow specification through Missions before execution — so there's nothing left for a model to reinterpret at runtime.

How fast can we deploy on our Snowflake data?

Weeks, not quarters. Tellius ships with the deterministic engines and the domain models already built, so it runs on your governed Snowflake data out of the box — no custom pipeline to assemble and re-check.

Can we buy Tellius through Snowflake?

Yes — Tellius is available through the Snowflake Marketplace, so a purchase can draw down existing committed spend and collapse procurement from months to weeks.

See Tellius reason on your Snowflake data

Bring a question your team asks every week. We'll show you the finished work come back — computed, reproducible, and written to your warehouse.

Book a demo Start for free

❄ Snowflake Technology Partner Gartner Magic Quadrant Visionary Nucleus Research Accelerator

DISCOVER MORE

Breakthrough Ideas, Right at Your Fingertips

Dig into our latest guides, webinars, whitepapers, and best practices that help you leverage data for tangible, scalable results.

ThoughtSpot alternative: migrating from ThoughtSpot to Tellius (2026 guide)

Many organizations that adopted search-driven analytics are now looking for platforms that go beyond answering questions to proactively investigating business performance and delivering finished analytical work. This migration guide explains how to move from ThoughtSpot to Tellius while preserving trusted metrics, semantic definitions, dashboards, and user adoption. It compares the two platforms across conversational analytics, automated insight discovery, root cause analysis, AI agents, governance, semantic context, and enterprise scalability.

Agentic AI in Supply Chain: Use Cases, Platforms, and What's Shipping (2026)

Supply chain leaders are moving beyond predictive dashboards and generative AI toward agentic systems that can monitor operations, investigate disruptions, recommend actions, and orchestrate workflows across the enterprise. This guide explores the state of agentic AI in supply chain management in 2026, comparing the leading platforms and highlighting the use cases that are delivering measurable business value today. Learn how organizations are deploying AI agents for demand sensing, inventory optimization, supplier risk monitoring, procurement intelligence, logistics planning, production scheduling, transportation optimization, and exception management.

Best FP&A Software in 2026: The Tools That Plan, and the Layer That Explains Why

Modern FP&A teams need more than planning software—they need the ability to explain financial performance, identify the drivers behind change, and act faster on emerging risks and opportunities. This buyer's guide compares the leading FP&A platforms in 2026 across budgeting, forecasting, scenario modeling, financial consolidation, reporting, AI-assisted planning, and enterprise analytics. It also introduces a new evaluation framework: distinguishing the systems that plan from the intelligence layer that explains why.

Tellius 6.3: Kaiya Everywhere, Grounded in Your Business

AI answers are only useful when they're grounded in your business and available where decisions actually happen. See how Tellius 6.3 delivers fast, traceable insights across pharma, CPG, RevOps, and FP&A — right in the tools your team already uses.

Tellius 6.1: AI Agents Across Every Metric, Document, and Conversation

Tellius 6.0: Agent Mode for Deep Analytics + Insights

PMSA Fall Symposium 2025 in Boston

Join Tellius at PMSA Oct 2–3 for two can’t-miss sessions: Regeneron on how they’re scaling GenAI across the pharma brand lifecycle, and a hands-on workshop on AI Agents for sales, HCP targeting, and access wins. Discover how AI-powered analytics drives commercial success.

Tellius AI Agents: Driving Real Analysis, Action, + Enterprise Intelligence

Tellius AI Agents transform business intelligence with dedicated AI squads that automate complex analysis workflows without coding. Join our April 17th webinar to discover how these agents can 100x enterprise productivity by turning questions into actionable insights, adapting to your unique business processes, and driving decisions with trustworthy, explainable intelligence.

View All Resources