- Is Snowflake available in South Africa, and can our data stay in the country?
- Yes. Snowflake announced general availability of its AI Data Cloud on the AWS Africa (Cape Town) Region on 1 September 2025, which lets you keep data in-country for POPIA and the National Policy on Data and Cloud. Some Cortex AI models are not yet hosted in every region; Snowflake's cross-region inference can route those requests to another AWS region over its private network, and we make that an explicit, documented design decision with your information officer rather than a surprise.
- What is the difference between Snowflake CoWork and CoCo?
- CoWork (formerly Snowflake Intelligence) is the agent for business users: it answers questions over governed data, runs Deep Research reports, publishes Artifacts dashboards and can act in Slack, Gmail and Salesforce. CoCo (formerly Cortex Code) is the coding agent for data engineers and developers: it reads your schemas, permissions and lineage, writes and runs SQL, Python and dbt, and can run on a schedule through Automations. Both were renamed at Snowflake Summit in June 2026 and both sit on Cortex AI.
- Do we need a separate vector database to build RAG on Snowflake?
- No. Cortex Search is a managed hybrid vector and keyword search service that handles embedding, indexing and retrieval inside Snowflake. For most enterprise RAG use cases it removes the need for a separate vector store and the pipeline that would otherwise keep it in sync.
- Which AI models run inside Snowflake Cortex?
- Cortex AI serves models from Anthropic (Claude), OpenAI, Meta, Mistral AI and DeepSeek, selectable per function or per agent. Anthropic and Snowflake expanded their partnership in December 2025, and Claude models power CoCo and much of CoWork. As an independent member of Anthropic's Claude Partner Network we tend to default to Claude for reasoning-heavy agents and pick smaller models where cost per token matters more than nuance.
- Should we choose Snowflake, BigQuery or Databricks?
- It depends on your existing stack, your team and your workloads, and we will tell you honestly if Snowflake is not the right fit. Snowflake suits organisations that want strong governance, easy data sharing and business-user AI with minimal platform engineering. BigQuery suits Google Cloud-centric teams; Databricks suits heavy Spark and ML engineering. We build on all three, which is exactly why the free AI assessment starts with your constraints, not our preferences.
- How is Snowflake Cortex AI priced, and how do we keep it under control?
- Cortex AI functions and agents consume Snowflake credits, mostly on a per-token basis that varies by model, on top of normal warehouse compute. We put resource monitors, model selection rules and Adaptive Compute in place from day one, and we review query and token cost as part of every engagement so that an AI feature that works in a demo also works in the budget.
- How long does a Snowflake migration or first AI workload take?
- A free AI assessment gives you a scoped plan within two weeks. A first production workload — a governed warehouse with one or two source systems, dbt models and reporting — typically lands inside a quarter. Adding Cortex AI features such as document extraction or a CoWork rollout is usually a matter of weeks once the semantic layer exists, which is why we build that layer first.
- Are you an official Snowflake partner?
- We describe ourselves as a Snowflake partner practice: that phrase describes our delivery capability on the platform, not a signed programme membership, and Automation Architects is independent of Snowflake Inc. Our only formal programme membership is Anthropic's Claude Partner Network. We would rather be precise about that than let a badge do the talking.