Every organization is undergoing a massive transformation about how they get things done. How we work with data and analytics is also changing dramatically. Instead of relying on dashboards and static KPI reports that take time to interpret, teams are beginning to learn from their data by simply talking to it powered by Reliable Data AI Agents. This shift toward conversational access is powerful. When done right, it removes friction, lowers the learning curve, and brings insights closer to the people who need them.
While the concept of conversing with data is not new and various commercial offerings propose solutions, a large-scale implementation is still unviable due to practical challenges and limitations of these solutions.
This blog post will showcase how Trust3 IQ addresses common data challenges using a practical, real-world example and a sample dataset.
A high-level roadmap for the blog:
- Data Setup: Generate and load the demonstration dataset into Snowflake.
- Trust3 IQ Setup: Create a Trust3 trial account and initialize the IQ spaces.
- Insights: Query the data and get insights using Trust3 IQ.
- Summary: Present final thoughts and conclusions.
Case Study: Sales Intelligence
In a previous blog, we explored the Sales AI Assistant use case and the real-world challenges of deploying such assistants inside organizations.
Common challenges faced by teams building these AI Assistants:
- Ambiguity in Business Context: AI struggles to interpret terms like “similar customer” or “healthy pipeline” without a clear definition of what “similar” means or what defines a healthy pipeline.
- Repetitive & Extensive Manual Configuration: Defining the necessary data joins and metrics for each use case requires extensive manual effort that is repeated for each AI agent use case.
- Context Space Utilization: Simply providing schemas and table definitions as context to LLMs is inefficient, resulting in context overload and inaccurate results.
- Lack of Trust and Transparency: Users cannot validate how results are generated, leading to skepticism and low adoption.
Scenarios
A fictitious product company Nimbus Inc. wants to empower their sales, marketing and strategy organizations with data to improve not just productivity but increase their top line growth. Instead of relying on specialized data and analyst teams, enable business teams to self-serve.
A sales rep wants to know:
- Give me a list of all prospects who attended a recent webinar on Tech Workshop
- I am getting into a meeting, tell me everything about Account X’s products, current spend, usage to help me prep for an upcoming meeting.
A sales manager wants to know:
- What’s the health of my pipeline compared to the targets this quarter? Is my team likely to meet the targets?
- Which customers are at risk of churn based on their usage?
To demonstrate this, we will use a synthetic dataset and load it into a Snowflake.
- Sign up for a free Snowflake account here.
- Clone the github repo.
- Follow the instructions in the QUICK_START_SNOWFLAKE to load the data
- Validate that the dataset has been loaded correctly from Snowpark. Run a few statements to test everything was loaded successfully.
The generated dataset represents data typical to various functional departments such as marketing, sales, finance, product and support. This dataset contains 62 tables across 5 schemas, simulating the reality of data for most organizations.
- CRM and Marketing Data: Opportunities, campaigns, customer segments, and attribution metrics showing how marketing activities impact revenue.
- Product Usage Data: Feature adoption, license utilization, and engagement trends revealing produ
ct stickiness and renewal likelihood. - Support Data: Tickets, resolutions, and satisfaction scores representing customer health signals.
- Operations Data: Customer operations data that indicates customer health, satisfaction ratings, customer success engagements.
Let us now proceed with doing the necessary to start using this data set with Trust3 IQ.
Trust3 IQ
Register today to create a Free Trust3 AI account. After registration, proceed to building an IQ space.
Build an IQ Space
IQ Space is a governed context that represents your business entities, concepts, definitions, physical table descriptions, relationships, and policies extracted from the data sources. IQ Spaces are highly flexible, allowing you to tailor what’s included in each space, whether you organize them by domain, department, or environments.
To build an IQ Space:
- Click “Create IQ Space” and provide a meaningful name and description
- Click Add Data Assets to start connecting your datasource and adding the database and required schemas
- Connect your data sources. You can follow this step-by-step tutorial for connecting with Snowflake.
- Browse through the database and choose the schemas, for this tutorial, we will add the SALES_DATA and MARKETING_DATA schemas
- Ask IQ to auto-generate semantics.
- Finish adding data assets and click to create the IQ space
- Give it some time to derive the semantic model automatically from the data and metadata definitions.
- Start asking questions directly or through an AI agent.
A quick recap of what we have done thus far:
- Data is loaded into Snowflake.
- Created a Trial account with Trust3 IQ.
- Created an IQ space.
Now, we are ready to start chatting with our data using IQ.
Demonstration: Gain insights with IQ
Q1: As a sales rep, I want to identify high value prospects to attain my quota.
Show me prospects from the webinar topics beginning with “Tech Workshop” who asked questions or stayed for 75%+ of the session, prioritized by company size and territory fit.
To address this question, AI needs to understand the following:
- Data Discovery and Joins:
- Identify the sources of leads data and join it with webinar event records to accurately filter for attendees of the “Tech Workshop” webinar sessions.
- The semantic model is crucial here, as it provides the necessary information on the join relationships between different tables.
- Filtering on Engagement Criteria:
- Apply the user-defined engagement criteria.
- To create a filter for 75% engagement, the AI must know the session duration and where to retrieve this information within the data.
- Prioritization and Presentation:
- Prioritize the results based on company size and territory alignment.
- This requires knowing the existing territories and how company sizes are categorized within the organization.
Let’s proceed with Asking IQ.
Asking IQ, provides the results with detailed analysis and the SQL.
Analysis by IQ
# Tech Workshop Prospect Analysis Results
I've successfully identified **48 qualified prospects** from the "Tech Workshop 2023" webinar who either asked questions OR stayed for 75%+ of the session. Here's the strategic breakdown:
## Key Highlights
### 📊 Company Size Distribution (Prioritized)
#### Enterprise (10 prospects)
**Highest priority**
- Strong engagement from APAC, EMEA, Latin America, and North America
- Notable titles: CTOs, Directors IT, CEOs, Administrators
#### Large (12 prospects)
**Second priority**
- Balanced global representation
- Key roles: CTOs, VP Engineering, Directors IT
#### Medium (13 prospects)
**Third priority**
- Strong APAC and Latin America presence
- Mix of technical and executive roles
#### Small (13 prospects)
**Fourth priority**
- Heavy North America concentration
- Diverse role mix including CEOs and technical leads
### Geographic Distribution
- **North America**: 19 prospects (40%)
- **APAC**: 13 prospects (27%)
- **Latin America**: 10 prospects (21%)
- **EMEA**: 6 prospects (12%)
### Engagement Quality
- All 48 prospects asked at least 1 question, showing active engagement
- Question voluGenerated SQL
SELECT DISTINCT
l.LEAD_ID,
l.CONTACT_NAME,
l.COMPANY_NAME,
l.COMPANY_SIZE,
l.REGION,
l.TITLE,
wa.QUESTIONS_ASKED,
wa.ATTENDANCE_DURATION_MIN,
e.EVENT_NAME,
e.EVENT_TYPE,
CASE
WHEN wa.QUESTIONS_ASKED > 0 AND wa.ATTENDANCE_DURATION_MIN >= 75 THEN 'Asked questions AND stayed 75%+'
WHEN wa.QUESTIONS_ASKED > 0 THEN 'Asked questions'
WHEN wa.ATTENDANCE_DURATION_MIN >= 75 THEN 'Stayed 75%+'
ELSE 'Other'
END as QUALIFICATION_REASON
FROM SALES_ANALYTICS_DEMO.MARKETING_DATA.WEBINAR_ATTENDANCE wa
JOIN SALES_ANALYTICS_DEMO.MARKETING_DATA.EVENTS e ON wa.EVENT_ID = e.EVENT_ID
JOIN SALES_ANALYTICS_DEMO.MARKETING_DATA.LEADS l ON wa.LEAD_ID = l.LEAD_ID
WHERE e.EVENT_NAME LIKE 'Tech Workshop%'
AND (wa.QUESTIONS_ASKED > 0 OR wa.ATTENDANCE_DURATION_MIN >= 75)
AND l.LEAD_ID IS NOT NULL
AND l.CONTACT_NAME IS NOT NULL
AND l.COMPANY_NAME IS NOT NULL
ORDER BY
CASE l.COMPANY_SIZE
WHEN 'Enterprise' THEN 1
WHEN 'Large' THEN 2
WHEN 'Medium' THEN 3
WHEN 'Small' THEN 4
ELSE 5
END,
l.REGION,
wa.QUESTIONS_ASKED DESC,
wa.ATTENDANCE_DURATION_MIN DESC
Observations:
IQ delivered impressive accuracy almost immediately after the space was created, requiring no preliminary setup with entity relationships, metrics, or example queries.
- A key advantage was that all necessary schemas were part of the same IQ space, eliminating the need to switch between different agents.
- Initially, IQ provided results without the necessary filter for “Tech workshops.” However, by making the questions more specific—for example, Webinar topic name starting from Tech Workshops—IQ successfully adjusted the filter. Subsequent questions were then resolved perfectly with the correct filtering in place.
Q2: As a sales leader, determine the health of my team’s commitments this quarter?
Show me the gap-to-quota by rep with recommended actions: which deals need executive engagement, pricing concessions, or are likely to slip.
- Identify Quota Gaps: Locate sales representative and quota data to accurately assess the current quarter’s sales shortfall.
- Deal Analysis for Q-Close: Evaluate deals to determine the probability of closure within the current quarter.
- Suggest Best Practices: Utilize institutional knowledge or access to relevant resources to recommend optimal strategies.
A snippet of the response from IQ
Analysis by IQ
# Gap-to-Quota Analysis with Recommended Actions
Based on your request to analyze gap-to-quota by sales rep with specific recommendations for deals needing executive engagement, pricing concessions, or likely to slip, here's the comprehensive analysis:
## Key Findings Summary
### Critical Attention Needed (P1-P3)
- **Rep_5**: $3.78M gap (5.46% attainment) - CRITICAL SITUATION
- **Rep_42**: $1.84M gap (53.93% attainment) - Medium priority
- **Rep_4**: $1.68M gap (58.02% attainment) - Medium priority
- **Rep_10**: $1.42M gap (64.48% attainment) - Medium priority
---
## Immediate Actions Required
### This Week
- Emergency pipeline review for Rep_5
- Deal acceleration coaching for Rep_4
- Pipeline forecasting session for Reps 42, 10, 47
### This Month
- Implement Rep_48's and Rep_35's best practices across team
- Review territory assignments and lead distribution
- Establish regular deal review cadence for at-risk reps
---
## Conclusion
The analysis shows a clear need for immediate intervention on pipeline development and deal acceleration rather than pricing or executive engagement issues.To answer the question, IQ generated a comprehensive analytical query that identifies underperforming sales reps, diagnoses the root causes of their gaps, and provides actionable recommendations for improvement.
It created 3 CTEs and a final output.
- Rep_performance- calculates fundamental performance metrics for the sales rep such as revenue metrics, activity i.e deals won or lost, and gap analysis using the SALES_TARGET, CLOSED_DEALS and TERRITORIES tables.
- deal_insights – used to analyze the characteristics and patterns of deals lost or won. Specifically checks for pricing discounts applied, customer mix, and deal velocity. Uses QUOTES, CUSTOMERS AND CLOSED_DEALS tables.
- open_pipeline – calculates forward looking indicators on the open pipe to assess the deal quality, pipeline health, and risk indicators such as age of the deal, competitive threats etc. Uses OPPORTUNITIES, CLOSED_DEALS tables
Snippet of deal_insights CTE
deal_insights AS (
SELECT
cd.SALES_REP,
COUNT(*) as total_deals,
AVG(q.DISCOUNT_PERCENT) as avg_discount,
COUNT(CASE WHEN q.DISCOUNT_PERCENT > 20 THEN 1 END) as high_discount_deals,
COUNT(CASE WHEN c.CUSTOMER_TIER = 'Enterprise' THEN 1 END) as enterprise_deals,
COUNT(CASE WHEN cd.CLOSE_DATE >= CURRENT_DATE - 90 THEN 1 END) as recent_deals
FROM SALES_ANALYTICS_DEMO.SALES_DATA.CLOSED_DEALS cd
LEFT JOIN SALES_ANALYTICS_DEMO.SALES_DATA.QUOTES q ON cd.DEAL_ID = q.OPPORTUNITY_ID
LEFT JOIN SALES_ANALYTICS_DEMO.SALES_DATA.CUSTOMERS c ON cd.CUSTOMER_ID = c.CUSTOMER_ID
GROUP BY cd.SALES_REP
)Finally, IQ puts everything together to generate concrete recommendations and prescriptive actions based on performance patterns.
Observations:
IQ is not just reporting numbers, it is diagnosing WHY reps are behind and prescribing specific remediation actions based on patterns in their deal data, discount behavior, pipeline health, and competitive dynamics.
We needed to refine some prompts to ensure accurate filtering by IQ. Specifically, we had to clarify that “Won” should be used as the status, rather than “Closed Won.” It was a one time prompt and IQ adjusted its subsequent inquiries to automatically add this specific filter.
Now managers can use this to power their weekly review calls to:
- Spot Quickly spot at-risk quota attainment
- Understand root causes (pricing? pipeline? velocity?)
- Take targeted coaching actions
- Track progress over time
Extending Context to AI Agents via MCP
Once IQ Spaces are live, they can be accessed by AI agents through the Model Context Protocol (MCP). MCP allows agents to act within governed boundaries, using context—not raw data.
Examples include:
- A renewal agent that flags at-risk customers automatically.
- A campaign agent that optimizes marketing spend.
- A finance agent that reconciles invoices using trusted definitions.
Through MCP, Trust3 IQ becomes the control plane for enterprise AI, ensuring every agent action is governed, semantic, and auditable.
Watch out for this space for our next blog demonstrating how to build an AI agent with IQ as a tool.
Closing Thoughts
In this blog, we showed how IQ builds, manages, and delivers the context that AI systems need to answer business questions accurately. IQ understands the intent behind a question, plans how to resolve it, identifies the right entities and relationships, and produces a clear, grounded answer based on the retrieved data.
IQ also works immediately. It doesn’t require months of prompt tuning or manual setup. It learns from every interaction, improving how it interprets questions and how it connects meaning across your data.
Trust3 IQ solves the core challenge that holds most chat‑with‑data systems back. By providing a reusable, governed context layer that unifies data, semantics, and policies, IQ turns AI from a guessing system into a dependable operating engine. That’s what makes IQ uniquely equipped to handle real, cross‑domain business questions that other tools simply cannot support.
Did you try the tutorial and have thoughts or feedback to share with the product team – contact us at [email protected].
