Can Snowflake store unstructured data? How Snowflake handles documents, images, and other data in 2025

Mid-market data and analytics leaders face a unique challenge when implementing Snowflake. You need enterprise-grade security controls but often lack the dedicated security teams that larger organizations maintain. The 2024 Snowflake breaches proved what can happen when access controls are weak. With 165 organizations affected, the lesson is clear: security must be part of the implementation, not an afterthought. Snowstack helps enterprises implement Snowflake security with governance frameworks embedded from day one, ensuring mid-market teams achieve compliance confidence without slowing down delivery.

This guide breaks down seven security and governance essentials that mid-market teams need to address when deploying Snowflake. Each essential covers what to implement, why it matters, and how to get it right the first time.

Key Takeaways: 7 Snowflake Security Essentials for Mid-Market Teams

• Role-based access control structures permissions around business functions rather than individual users for scalable governance.
• Multi-factor authentication blocks credential-based attacks, which caused most recent Snowflake security incidents.
• Data encryption at rest and in transit protects sensitive information from unauthorized access and interception.
• Network policies restrict platform access to approved IP ranges and reduce external attack surface.
• Snowstack embeds governance controls during initial architecture design, helping mid-market teams achieve 100% audit readiness.

Security and Governance Essentials for Mid-Market Snowflake Implementations

1. Role-Based Access Control and Least Privilege

RBAC forms the foundation of Snowflake security. Instead of granting privileges directly to users, you assign privileges to roles and then grant those roles to users. This approach simplifies administration, supports compliance requirements, and makes access audits straightforward.

Mid-market teams should create a structured role hierarchy that separates functional roles from administrative ones. A finance analyst should have read access to reporting tables only. An ETL engineer needs write access to staging schemas. Keep these responsibilities distinct with specific, targeted grants rather than broad database-level permissions.

Critical practices include isolating compute access from data access, using separate roles for warehouse usage and data queries, and reserving ACCOUNTADMIN for emergency situations only.

2. Multi-Factor Authentication Enforcement

MFA blocks the most common attack vector: stolen credentials. The 2024 breaches happened because passwords were compromised and MFA was missing. Enforcing MFA across your entire Snowflake environment is the single most effective step you can take to protect user accounts.

Integrate Snowflake with your existing identity provider through SAML or OAuth for centralized management. Require MFA at the IdP level so all integrated applications, including Snowflake, inherit the same authentication standards. Document break-glass procedures for critical roles in case your SSO provider experiences downtime.

Do not make MFA optional. A universal enforcement policy is the only way to ensure this control cannot be circumvented by individual users.

3. Data Encryption Configuration

Snowflake encrypts data at rest with AES-256 and data in transit with TLS 1.2+ by default. For mid-market organizations handling regulated data, consider customer-managed encryption keys through your cloud provider's Key Management Service. This adds control over key access and the ability to revoke access instantly if needed.

Tri-Secret Secure combines a customer-managed key with Snowflake-managed and cloud provider keys. No single entity can decrypt the data independently. Establish key rotation policies and implement separate keys for development, staging, and production environments.

Monitor your KMS audit logs for unusual key access attempts. Early detection of anomalous activity can prevent security incidents from escalating.

4. Network Policies and IP Allowlisting

Network policies restrict access to your Snowflake environment based on IP addresses. This limits potential attack surface by ensuring only authorized networks can connect to your data platform.

Define allowlists based on your corporate network ranges, VPN endpoints, and trusted partner connections. For organizations with distributed teams, combine network policies with private connectivity options like AWS PrivateLink or Azure Private Link.

Review and update network policies quarterly as your organization's network footprint changes. Remote work and cloud-based tools can introduce new IP ranges that need authorization.

5. Activity Monitoring and Audit Logging

Continuous monitoring of user activities and access patterns identifies potential security threats before they become incidents. Snowflake's ACCOUNT_USAGE schema stores query history, login history, and administrative changes for up to one year.

Forward these logs to your SIEM platform for correlation with other security events across your infrastructure. Configure automated alerts for high-risk activities: ACCOUNTADMIN logins, unusual data export volumes, failed authentication attempts from new locations.

Create visualization dashboards to spot anomalies in query patterns and login trends. A sudden spike in data access outside business hours warrants immediate investigation. Grant access to ACCOUNT_USAGE views only to a dedicated AUDITOR role to preserve log integrity.

6. Data Classification and Dynamic Masking

Data classification identifies and tags sensitive columns, while dynamic masking automatically redacts that data based on the querying user's role. This protects sensitive information without altering source data or limiting legitimate analytics work.

Use Snowflake's EXTRACT_SEMANTIC_CATEGORIES function or partner tools to scan and tag sensitive columns automatically. Create masking policies with conditional logic that returns full values for authorized roles and redacted values for everyone else.

Apply masking policies through classification tags rather than individual columns. Any column tagged as PII automatically inherits the correct masking policy, reducing manual configuration and ensuring consistent protection across your environment.

7. Governance Framework and Compliance Alignment

Mid-market organizations operating under SOC 2, HIPAA, GDPR, or PCI DSS need governance controls built into initial architecture. Retrofitting compliance is expensive and error-prone. Embed lineage tracking, access documentation, and audit trails from day one.

Document your data governance framework including data classification standards, retention policies, access review cadences, and incident response procedures. Regular access reviews, conducted quarterly at minimum, verify that role assignments remain appropriate as team members change responsibilities.

Snowstack delivers Snowflake consulting with compliance expertise for regulated industries. Our implementations achieve 100% audit readiness for SOC 2 and GDPR frameworks with governance controls, access audits, and full traceability embedded during the initial build.

How Mid-Market Teams Can Secure Their Snowflake Platform

Security and governance decisions made during Snowflake implementation determine long-term platform health. Mid-market teams that address these seven essentials from the start avoid costly remediation projects later.

The challenge for many mid-market organizations is internal expertise. Specialized Snowflake knowledge for security architecture, RBAC design, and compliance frameworks requires experience across multiple deployments. Snowstack brings this Snowflake expertise to mid-market teams through Platform Team as a Service, compressing typical implementation timelines while embedding enterprise-grade security controls.

Ready to implement secure Snowflake data governance for your organization? Contact Snowstack to discuss your specific security and compliance requirements.

‍

‍

Choosing a Snowflake consulting and support services partner shapes whether your data platform becomes a competitive asset or an ongoing headache. Generic partner rankings rarely help mid-market data leaders make confident decisions because they overlook the operational realities that matter most during modernization projects.

Mid-market organizations face a distinct set of challenges. Limited internal Snowflake expertise, fluctuating project demands, and strict budget constraints make vendor selection critical. The wrong partner can leave your team managing technical debt for years.

Snowstack helps enterprises implement Snowflake solutions with fast migrations, trusted data platforms, and AI-ready infrastructure designed for speed, compliance, and business growth. This guide covers seven essential evaluation criteria that go beyond certifications and logos.

Quick guide: 7 things to know when choosing Snowflake support partners

1. Snowstack: Top choice for mid-market teams needing dedicated Snowflake expertise with 90-day delivery
2. Service model clarity: Understand the difference between project-based and ongoing support
3. Operational depth: Look for 24/7 monitoring and defined SLAs
4. Cost optimization focus: Confirm FinOps capabilities that reduce Snowflake spend
5. AI readiness: Verify Cortex AI and ML pipeline experience
6. Knowledge transfer: Ensure documentation and training are included
7. Governance expertise: Check for RBAC, compliance, and audit trail capabilities

How we chose the evaluation criteria for Snowflake support partners

Mid-market data leaders need partners who deliver results without the overhead of large consultancy engagements. We focused on factors that directly impact project success, team productivity, and long-term platform health.

• Delivery speed: How quickly can the partner move from discovery to production? Faster timelines mean faster business value.
• Service model flexibility: Does the partner offer project-based work, managed services, or both? Your needs may shift over time.
• Cost management capability: Can the partner demonstrate measurable Snowflake cost reductions through optimization?
• AI and advanced analytics readiness: Does the partner have hands-on experience with Cortex AI, ML pipelines, and vector search integration?
• Governance and compliance track record: Can they implement RBAC, data masking, and audit trails for regulated industries?
• Knowledge transfer approach: Will your internal team gain skills, or will you remain dependent on the partner indefinitely?

The 7 best things to know for Snowflake support partner evaluation

1. Snowstack: Top Snowflake consulting partner for mid-market modernization

Snowstack delivers Platform Team as a Service, compressing typical 12-month projects into 90-day engagements. The firm specializes in mid-market to enterprise organizations that need dedicated Snowflake expertise without building large internal teams.

Every engagement includes senior architects from start to finish. This differs from larger consultancies where senior resources often disappear after the sales cycle. FinOps cost optimization comes standard, not as an expensive add-on.

Client outcomes include 30 to 50 percent cost reduction and 80 percent faster reporting cycles across pharma, financial services, and FMCG implementations. The structured collaboration model features bi-weekly reviews, backlog planning, and clear RACI ownership.

Snowstack benefits

• 90-day delivery framework: Enterprise implementations move from discovery to production in compressed timeframes, letting your team see results while larger projects would still be in planning
• Embedded FinOps: Every engagement includes cost analysis that identifies 30 to 50 percent spending reduction through warehouse right-sizing and query optimization
• AI-ready architecture: Implementations support Cortex AI integration, vector search, and ML pipeline deployment from day one
• 24/7 monitoring: SLA-driven support with automated alerting delivers 60 percent faster incident resolution
• Knowledge transfer built in: Documentation, runbooks, and training reduce long-term consulting dependency
• Compliance expertise: Governance controls for SOC 2, GDPR, and HIPAA get embedded during initial architecture

Snowstack pros and cons

Pros:

• Senior architects remain involved throughout the entire engagement
• Proven delivery across regulated industries including pharma and financial services
• Transparent methodology with bi-weekly reviews and clear accountability

Cons:

• Focused specifically on Snowflake rather than multi-platform data strategies
• Mid-sized firm may have capacity constraints during peak demand periods
• Engagements require active client participation in reviews and planning sessions

2. Service model clarity: Project-based vs. managed support

Many organizations begin with a migration project and later realize they need ongoing operational support. Understanding the difference between engagement models prevents costly transitions later.

Project-based engagements work for defined initiatives like migrations or platform builds. Managed services make sense when your team lacks the capacity for day-to-day platform operations. Some partners offer hybrid models that combine implementation with ongoing support.

Service model benefits

• Project-based clarity: Fixed scope and timeline give predictable costs for budget planning
• Managed service continuity: Ongoing support ensures platform stability without internal hiring
• Hybrid flexibility: Combine implementation expertise with operational coverage as needs evolve

Service model pros and cons

Pros:

• Clear engagement boundaries help manage stakeholder expectations
• Managed services reduce internal team burden during high-demand periods
• Hybrid models adapt to changing organizational priorities

Cons:

• Project-based work may leave gaps in ongoing optimization
• Managed services require trust in external teams for critical operations
• Hybrid arrangements can create confusion about responsibilities

3. Operational depth: 24/7 monitoring and SLA commitments

Platform stability directly impacts business operations. Partners with operational depth offer monitoring, incident response, and defined service level agreements that protect your investment.

Look for partners who implement automated alerting and can demonstrate response time commitments. The difference between reactive support and proactive monitoring often determines whether issues become outages or get resolved before anyone notices.

Operational depth benefits

• Proactive monitoring: Automated systems catch issues before they impact downstream reporting
• Defined SLAs: Response and resolution time commitments create accountability
• Incident documentation: Clear records help identify patterns and prevent recurring problems

Operational depth pros and cons

Pros:

• 24/7 coverage protects against overnight and weekend incidents
• SLA commitments give measurable partner accountability
• Proactive monitoring reduces firefighting for internal teams

Cons:

• Round-the-clock support adds to engagement costs
• SLAs only matter if partners actually meet them consistently
• Monitoring requires proper configuration to avoid alert fatigue

4. Cost optimization focus: FinOps that reduces Snowflake spend

Snowflake's consumption-based pricing can spiral quickly without proper governance. Partners with genuine FinOps expertise demonstrate measurable cost reductions through warehouse right-sizing, query optimization, and automated scaling policies.

Ask for specific examples of cost savings from previous engagements. Credible partners show before-and-after metrics from client environments rather than theoretical projections.

Cost optimization benefits

• Warehouse right-sizing: Matching compute resources to actual workload demands eliminates waste
• Query optimization: Efficient queries reduce credit consumption without sacrificing performance
• Automated scaling: Auto-suspend and resource monitors prevent runaway costs

Cost optimization pros and cons

Pros:

• Cost savings often pay for the consulting engagement within months
• Optimization creates ongoing value beyond the initial project
• Visibility into spending patterns supports better budget planning

Cons:

• Aggressive optimization may impact query performance if poorly executed
• Cost management requires ongoing attention as workloads change
• Some partners treat FinOps as an upsell rather than a core capability

5. AI readiness: Cortex AI and ML pipeline experience

Modern Snowflake implementations must support machine learning pipelines, large language model integrations, and retrieval-augmented generation patterns. Partners without hands-on AI experience deliver platforms that require expensive redesign when your organization advances AI initiatives.

Verify that partners have implemented Cortex AI, vector search, and Snowpark ML in production environments. Theoretical knowledge differs significantly from practical deployment experience.

AI readiness benefits

• Cortex AI integration: Built-in AI capabilities run where your data already lives
• Vector search support: Embeddings enable semantic search across documents and products
• ML pipeline deployment: Snowpark enables model training and serving inside Snowflake

AI readiness pros and cons

Pros:

• AI-ready architecture eliminates costly platform redesign later
• Keeping ML workloads in Snowflake simplifies governance
• Native AI features reduce dependency on external ML infrastructure

Cons:

• AI workloads require careful compute cost management
• Cortex AI capabilities continue evolving rapidly
• Teams need training to take advantage of AI features

6. Knowledge transfer: Documentation and team enablement

The goal of any consulting engagement should be reducing long-term dependency, not creating it. Partners who invest in knowledge transfer leave your team with runbooks, documentation, and practical skills.

Ask about training components, documentation standards, and how the partner handles handoff at project completion. Organizations that skip this step often find themselves calling consultants for basic operational questions.

Knowledge transfer benefits

• Runbooks and documentation: Clear guides help internal teams handle routine operations
• Hands-on training: Practical skill building creates lasting internal capability
• Architecture decision records: Documentation explains why decisions were made, not just what was built

Knowledge transfer pros and cons

Pros:

• Internal capability reduces ongoing consulting costs
• Documentation supports team member transitions
• Trained teams can extend the platform without external help

Cons:

• Knowledge transfer requires time investment from internal teams
• Documentation quality varies significantly between partners
• Training effectiveness depends on participant engagement

7. Governance expertise: RBAC, compliance, and audit trails

Data governance and security requirements have become business-critical for organizations operating under SOC 2, HIPAA, GDPR, and PCI DSS. Partners with governance expertise implement controls during initial architecture rather than retrofitting them later.

The cost differential between proactive governance and reactive compliance can reach significant amounts in enterprise environments. Look for partners who demonstrate role-based access control design, data masking policies, and lineage tracking capabilities.

Governance benefits

• Role-based access control: Structured role hierarchies provide clarity and security
• Data masking: Column-level masking protects sensitive information automatically
• Audit trails: Complete lineage tracking supports compliance reporting

Governance pros and cons

Pros:

• Built-in governance prevents expensive compliance remediation
• Proper access controls reduce data breach risk
• Audit readiness simplifies regulatory examinations

Cons:

• Governance implementation adds complexity to initial projects
• Overly restrictive controls can slow down legitimate data access
• Compliance requirements vary by industry and geography

Comparison table: Snowflake support partner evaluation criteria

‍

What questions should you ask a Snowflake consulting partner before signing?

The right questions reveal whether a partner can deliver on their promises. Focus on specifics rather than accepting vague assurances.

Start with their methodology. Ask them to walk through their implementation approach and show sanitized architecture diagrams from similar projects. Credible partners have documented processes they can explain clearly.

Dig into their team structure. Find out who will actually work on your project day-to-day, not just who attends the sales meetings. Junior resources may struggle with complex Snowflake architecture decisions.

• Request case studies with measurable outcomes and cost savings data
• Ask about their approach to knowledge transfer and documentation
• Clarify what happens if the project runs over budget or timeline
• Verify their hands-on experience with Cortex AI and Snowflake migrations

How do you evaluate a Snowflake partner's AI and analytics capabilities?

AI capability has become critical as organizations deploy machine learning pipelines and LLM integrations. Consultants without practical Cortex AI experience deliver platforms that need expensive rework when AI initiatives mature.

Ask for specific examples of AI implementations. Request details about vector search deployments, ML model serving, and governance patterns for AI workloads. Theoretical knowledge differs significantly from production deployment experience.

Verify they can demonstrate cost management for AI workloads. Cortex AI and ML pipelines can consume significant compute resources without proper guardrails. Partners should explain their approach to balancing AI capability with cost efficiency.

Why Snowstack is the top Snowflake consulting partner for mid-market data platform modernization

Snowstack combines deep Snowflake consulting expertise with proven delivery methods and transparent team structures. The firm delivers production-ready environments in 90 days while larger consultancies require 12 to 18 months for equivalent capability.

Cost optimization gets delivered as core methodology rather than optional add-on. Every Snowstack engagement includes FinOps analysis that identifies spending reduction opportunities through warehouse right-sizing, query optimization, and automated scaling policies.

AI readiness comes embedded in architecture from day one. Snowstack implementations support Cortex AI integration, vector search capabilities, and machine learning pipeline deployment without requiring platform redesign. The Platform Team as a Service model provides ongoing senior architect access rather than transitioning to junior support resources post-implementation.

Ready to evaluate Snowflake support partners for your data platform modernization? Contact Snowstack to discuss your specific requirements.

‍

‍

‍

Under the Hood: grounding CoWork with Cortex Sense — not just prompting it

Your agent isn't wrong because your data is dirty. It's wrong because it doesn't know what your data means. An LLM can read your schema perfectly — table names, column types, row counts — and still have no idea that net revenue means gross revenue after discounts, that the fiscal year starts in February, or that "active customer" excludes anyone who churned last quarter. That gap between the schema an agent sees and the business meaning it doesn't is where confident, wrong answers come from. Closing it is now the single highest-leverage thing a data team can do for AI.

That's also the thesis Snowflake built its entire Summit 2026 agentic story around.

What actually changed at Summit 2026

Two things matter for anyone running agents on Snowflake:

Snowflake Intelligence is now CoWork. Same product lineage — the personal work agent that decomposes a question, researches across structured and unstructured data, and returns a cited answer — new name. If you saw Episode 3, this is the thing you already built against. Existing deployments migrate automatically.

Cortex Sense is the headline, and it's about accuracy, not features. Cortex Sense is a runtime context-enrichment layer: it automatically assembles business context — query history, object metadata, BI dashboards, and Horizon Context semantic views — and feeds it to CoWork and CoCo at query time, with no manual configuration. Snowflake's own internal benchmark puts the difference starkly: 47% accuracy on complex enterprise queries without it, 83% with it— and just 23% for frontier coding agents wired up through Snowflake's MCP connector alone. The message Snowflake is sending could not be clearer: context, not the model, determines agent quality.

We agree with that framing. But there are two catches, and they're exactly where a data team's real work lives.

The two catches nobody puts on the keynote slide

Catch #1 — Cortex Sense is private preview (as of June 2026). CoWork is shipping to enterprises now; Cortex Sense is not generally available yet. So the default CoWork deployment today operates closer to that 47% baseline, withoutSnowflake's own context infrastructure at production readiness. You can't wait for the feature to flip on and rescue answer quality before your stakeholders start trusting (or distrusting) the agent.

Catch #2 — even at GA, Cortex Sense is only as good as what's underneath it. Read the description again: it assembles context from your semantic views, metadata, and dashboards. If those definitions are missing, ambiguous, or contradictory, Cortex Sense faithfully assembles ambiguous context. And there's a deeper trap that governance alone never solves: access control is not correctness. RBAC enforces who can query the revenue table; it says nothing about whether that table is accurate, consistently defined, or current. An agent querying a revenue figure with an upstream ingestion error will return a confident, beautifully-cited, wrong number — and every guardrail will have done its job.

So the work is the same whether Cortex Sense is in preview or GA: you build the governed context layer and you make sure the data beneath it is actually right. The good news is that this work is not throwaway — the semantic layer you build now is precisely the substrate Cortex Sense consumes later. You're not waiting for the feature; you're getting ahead of it.

Here's how we build it.

Under the Hood: the context layer, step by step

Step 1 — Put the business definitions in a governed semantic view

A semantic view is a schema-level Snowflake object that maps physical columns to business concepts — facts, dimensions, and metrics — and stores the definitions natively, under RBAC, where both Cortex Analyst and (eventually) Cortex Sense read them. This is where you kill ambiguity once, centrally, instead of in fifty different dashboards.

The canonical example is the one Snowflake itself uses: revenue is physically stored in a column called amt_ttl_pre_dsc, but the business always means gross revenue after discounts. You encode that once:

CREATE OR REPLACE SEMANTIC VIEW analytics.sales.revenue_model
  TABLES (
    orders AS prod.sales.orders
      PRIMARY KEY (order_id)
      WITH SYNONYMS ('sales', 'bookings')
      COMMENT = 'One row per order line. Source of truth for revenue.',
    unit AS prod.sales.business_unit_dim
      PRIMARY KEY (unit_id)
  )
  RELATIONSHIPS (
    orders_to_unit AS orders (unit_id) REFERENCES unit (unit_id)
  )
  FACTS (
    orders.gross_amount   AS amt_ttl_pre_dsc,
    orders.discount_rate  AS disc_rate
  )
  DIMENSIONS (
    unit.unit_name    AS unit_name WITH SYNONYMS ('business unit', 'segment'),
    orders.order_date AS order_dt
  )
  METRICS (
    orders.net_revenue AS SUM(orders.gross_amount * (1 - orders.discount_rate))
      COMMENT = 'Net revenue = gross revenue after discounts. Use this for ALL
                 revenue reporting. Never sum amt_ttl_pre_dsc directly.'
  )
  COMMENT = 'Governed revenue model. These definitions are the single source
             of truth for agents and BI alike.';

‍

Now anyone — human or agent — asks the question the same way and gets the same number:

SELECT * FROM SEMANTIC_VIEW (
  analytics.sales.revenue_model
  METRICS    net_revenue
  DIMENSIONS unit_name
);

‍

Step 2 — Write your comments like prompts, because they are

This is the part that separates "it compiles" from "the agent is actually right." In a semantic view, Cortex Analyst reads your COMMENT text as instructions, not documentation. The comment on net_revenue above isn't a note for a future engineer — it's telling the model which column is not revenue. Be that explicit everywhere: define what a metric means, when to use it, and what to avoid. If you don't write it down, the model guesses, and a guess is how you get a wrong answer on clean data.

Two more high-leverage moves on the same object:

• Synonyms so "business unit," "segment," and "BU" all resolve to one dimension. Agents fail constantly on vocabulary mismatch; this fixes it cheaply.
• Verified queries — known-good question/SQL pairs that anchor the model on your hardest or most political metrics:

-- inside CREATE SEMANTIC VIEW, after the COMMENT clause:
AI_VERIFIED_QUERIES (
  net_rev_by_unit AS (
    QUESTION  'What was net revenue by business unit last quarter?'
    VERIFIED_AT 1717200000
    VERIFIED_BY '(owner = data-platform@yourco.com)'
    SQL 'SELECT * FROM SEMANTIC_VIEW (analytics.sales.revenue_model
           METRICS net_revenue DIMENSIONS unit_name)'
  )
);

‍

One discipline worth stating plainly: only add verified queries you have actually validated. One wrong example teaches the model a bad habit at scale.

Step 3 — Measure the lift on your KPIs, don't take 47→83 on faith

Snowflake's benchmark is theirs, on their data. Before you tell your CFO the agent is trustworthy, prove it on your questions. Snowflake ships a Cortex Agent evaluation framework for exactly this — define a dataset of real questions with expected answers, then score the agent against it:

evaluation:
  agent_params:
    agent_name: "revenue_agent"
    agent_type: "CORTEX AGENT"
  run_params:
    label: "Baseline — before semantic layer"
    source_metadata:
      type: "dataset"
      dataset_name: "kpi_eval_set"
  metrics:
    - answer_correctness        # how close the answer is to ground truth
    - tool_selection_accuracy   # did it call the right tools? (public preview)
    - logical_consistency       # reference-free; consistency across the run

‍

Run it once before the semantic layer exists, run it again after. The delta is your evidence — and your regression test. Wire it into CI so a careless change to a metric definition can't silently re-break answer quality next month.

Step 4 — Fix the data the context layer points at

A perfect semantic layer over a stale or half-loaded table still produces a wrong answer, just a well-defined one. So the context work has a twin: source-to-report reconciliation, freshness checks, and catching the broken or partial feeds that quietly poison a metric. That's a whole topic — it's Episode 5 — but flag it now, because "the agent gave the wrong number" is at least as often an ingestion problem as a semantics problem.

What this means, by role

If you lead data or analytics: the semantic layer is no longer a BI nicety — it's the accuracy substrate for every agent you're about to be asked to deploy. Building it now pays twice: better Cortex Analyst answers today, and a ready-made context source for Cortex Sense when it GAs.

If you're the architect or lead engineer: treat semantic views as strict contracts, not flexible SQL. Model relationships explicitly, comment like you're prompting, anchor hard metrics with verified queries, and put an evaluation set in CI. This is the build work that makes the demo survive contact with production.

If you own the platform strategy (VP / CDO): the question your stakeholders are really asking is "can we trust this for a real decision?" The honest answer is "only as far as our governed definitions and our data quality go." That's a roadmap, not a blocker — and it's a far better place to invest than another model evaluation.

How we'd approach it

Most teams we talk to don't have a context problem they can see — they have a trust problem they can feel: two dashboards disagree, an agent answer doesn't match the board deck, nobody's quite sure which number is right. The fix starts with finding where the definitions diverge and where the data underneath is wrong, before pointing any agent at it.

That's the shape of our AI-readiness assessment — a fixed-scope first step that maps your sources, definitions, and the gaps between what your reports say and what your data actually contains, so the agents you ship are accurate by construction. If your CoWork answers are landing in the "confident but wrong" zone, that's the place to start.

‍

‍

‍

Notes & sources: Cortex Sense status and the 47%→83% figure are from Snowflake's own materials and product announcements (Snowflake Summit 2026, June 2); Cortex Sense is in private preview as of June 2026, so treat the figure as a vendor benchmark and validate on your own data. Semantic-view DDL and the comment-as-instruction behavior follow Snowflake's CREATE SEMANTIC VIEW and semantic-view documentation; semantic-view SQL is stricter than ordinary SQL, so validate any DDL against current docs for your account version. Cortex Agent evaluation metrics (answer correctness, tool-selection accuracy, logical consistency) are from Snowflake's Cortex Agent evaluations documentation.

‍