Bridging the Data Divide: How Motive Uses Glean and Snowflake Agents to Give Executives a 360° Account View

Imagine you’re a VP of Sales or Customer Success, and you have a massive renewal coming up. You need a 360-degree view of the account, but right now, that “view” is a fragmented mess.

The information you need lives all over the place. It’s in the Slack threads where engineers are discussing a bug. It’s in the Quarterly Business Reviews (QBRs) tucked away in Google Docs. It’s in the transcripts of last week’s support calls. Then there’s the structured side: Salesforce deal history, product telematics, and billing data sitting in your data warehouse.

As a busy executive, you don’t have two hours to play detective. You need the “CliffsNotes” version of the account’s reality right now!

At Motive, we’ve solved this internal information collection challenge by building a bridge between Glean and Snowflake Intelligence. Here is how we’re turning fragmented data into actionable insights for our executives.

The “Data Scavenger Hunt” Problem

In most organizations, a “simple” account summary requires an executive to open at least five tabs:

1. CRM for the deal history and opportunity stages.
2. Slack to see if the sentiment is actually positive or just “polite.”
3. Google Drive for the last QBR deck and account plans.
4. BI tools or the data warehouse: for the product usage stats and telematics.
5. Support tools to see if they have 40 open tickets.

On the business intelligence team at Motive, we realized that the friction of moving between these systems often means the most important data points, the “soft” signals in a transcript or a weird dip in telematics, get missed. We needed a way to synthesize the Structured Truth (what the data says) with the Unstructured Narrative (what the people are saying).

The Architecture: Three Engines of Intelligence

To solve this challenge we’ve engineered a powerhouse: three distinct engines, one seamless interface. Here’s the architecture driving the future of our system.

Engine 1: The Structured Foundation (Snowflake Semantic Models)

First, we have the hard numbers. We’ve built out robust semantic views in Snowflake that act as the single source of truth for account stats. This includes:

• Product Telematics & Analytics: Logging from our hardware and software with a device health history.
• Account Financials: Annual Recurring Revenue (ARR) History.
• CRM Deals: History of closed won deals and the products purchased.

By using semantic models, we ensure the AI doesn’t have to “guess” how to calculate churn or usage. It queries a predefined model, ensuring the numbers an executive sees in a chat are the same numbers they’d see on an official finance dashboard.

Engine 2: The Unstructured Customer Hub (Snowflake Cortex Search)

The second piece is where the “external vibe” of the account lives. We’ve moved the history of our conversations with customers into a Snowflake Search Service (Cortex). This repository turns “dark data” into searchable intelligence:

• Transcripts: Sales calls and customer support phone calls.
• Written Records: Customer support emails, chat logs, and case notes.
• Field Notes: Sales rep notes on opportunities and accounts.

By putting these into Cortex Search, we allow our Cortex Agent to “read” through thousands of interactions in seconds to identify recurring pain points or shifts in customer sentiment that never make it into a structured database.

Engine 3: Internal Discussions (Glean Federated Search)

The third engine captures the “internal narrative.” A lot of the most important context about an account engineering blockers, internal dynamics, or feature requests, lives in Slack and Google Docs.

Because Glean uses federated search, it can look through these external systems while respecting the specific security and permissions of the user. If an executive doesn’t have access to a specific private channel, the AI won’t include it in the summary. This ensures we get the “tribal knowledge” without compromising data privacy.

Engine Orchestration: Glean Agents and Cortex Agents

We use a nested “Agent-to-Agent” architecture to tie these engines together. A Snowflake Cortex Agent serves as the specialist for our data warehouse; it orchestrates between Semantic Views (to get structure data from engine 1) and Cortex Search (to retrieve unstructured insights from Engine 2).

Finally, a Glean Agent acts as the primary orchestrator. It stays in the executive’s daily workflow, performing a federated search across internal docs (Engine 3) and simultaneously calling the Snowflake Cortex Agent to pull in the verified customer metrics and transcripts. This multi-agent approach ensures that the “internal narrative” and “structured truth” are synthesized into a single, cohesive response.

Data Security: Governing What the AI Can See

Building a powerful intelligence layer is only half the job. The other half is making sure it doesn’t surface data that someone shouldn’t see. Glean handles this elegantly out of the box. When a user searches, Glean respects the permissions of the underlying source system. If a Google Doc hasn’t been shared with them, it simply won’t appear in the results.

Snowflake is a different story. A data warehouse typically ingests from systems like Salesforce using a service account that can see everything. The moment your CRM data lands in the warehouse, it effectively loses its source-level security guardrails.

We solved this by building a row-level access mapping table that acts as our governance layer. Using a combination of our org hierarchy and account ownership, it controls exactly which accounts each user can query. On semantic views, this is enforced via Snowflake’s native row-level security policies. In the Cortex Agent, we apply a permissions check directly before results are returned from the search service.

The result is three clean tiers of access: users with no data access, super users who can see everything (typically BI, data, and ops teams), and everyone else, whose view is scoped to the accounts they own or manage through the hierarchy.

The Executive Interface: How It Works

We’ve connected the Snowflake Intelligence Agent to Glean to create a seamless workflow. The Glean agent is the entry point; the executive stays in the tool they already use for daily search.

They simply pick a CRM account and ask:

“I have a meeting with this account. Summarize the account history and the open issues we have.”

Glean then orchestrates the search across all three engines. It combines the Internal Context (Engine 3), the External Communication (Engine 2), and the Hard Metrics (Engine 1) into a single, cohesive brief.

The Engineering Frontier: The “Context” Dilemma

While the end-user experience is seamless, configuring this presents a unique architectural challenge: Where does the context live?

When you have three different engines, you have to decide which “brain” handles which thought. Glean has its own prompting logic, Snowflake Agents have their own system prompts, and the semantic models require specific context to pull the right metrics.

Just as when two humans speak, passing a prompt between Glean and Snowflake Agents can lead to lost context like a game of telephone. We’ve found that it’s best to keep the specific instructions into Glean for the final analysis and leave Snowflake to be purely data retrieval and not analysis. We still like using a Snowflake Agent as that allows us to have a single entry point for Glean to talk to multiple semantic views and the search service. We can also use this same Snowflake Agent in Snowflake Intelligence or other interfaces like Slack or Snowflake Intelligence

So Snowflake is directed to return raw data verbatim, and Glean handles all synthesis and final analysis. Then Glean is the brain to synthesize all the information together to surface the final output to our users.

We’ve had to be very intentional about where to put different bits of context. If you put too much instructional context in Glean, the Snowflake query might get muddy. If you put too little, the agent won’t know which account ID to look for.

The Bottom Line

By combining Glean’s federated search with Snowflake’s dual-engine approach, we’re moving away from “searching for data” and toward “receiving intelligence.”

For our leadership, it means the difference between walking into a renewal meeting with a hunch and walking in with a complete, AI-synthesized roadmap of every call, every metric, and every Slack thread that defines the customer’s journey.

Written by Jay Rosenthal, Director of Business Intelligence at Motive, and Shashi Gupta, BI Platform Manager at Motive.