EXPLORINEER
en

KR — ZDNet Korea

AI Models Alone Are Not Enough: Snowflake Bets on Enterprise Data

AI2 min read

Snowflake is betting that connecting company data to AI agents will set it apart from rivals. It unveiled new "Cocoa" and "Cowork" bots and tools to manage what AI agents do, ahead of its online event Expedition 2026.

What happened

Snowflake, a company that stores and analyzes business data, announced new AI tools before its online event Expedition 2026, which starts on the 3rd (US time). The company uses models from OpenAI and Anthropic. Its plan is to help AI understand a company's internal information and then do real work with it. Snowflake says it will support everything from learning the company's work to carrying out tasks and keeping control over them.

Key facts

  • Expedition 2026 runs for three days, starting on the 3rd (US time), and is held online.
  • The Cocoa and Cowork bots are in private preview, a stage where only some customers can try them.
  • Cocoa is used in more than 9,100 accounts. Cowork is used in more than 5,800 accounts.
  • The new Business Ontology will be offered as a public preview, meaning anyone can try it.

Why data matters for AI

A strong AI model is not enough on its own. If the AI does not know a company's data, rules, and way of working, its answers can be wrong. Snowflake says it can link AI to company data while keeping existing access rights and security rules. For example, one department may define "employee turnover" differently from another. Snowflake's Business Ontology aims to create one shared set of terms, approved by owners, so AI answers use the same meaning.

Why it matters

AI agents are software that can plan and do tasks on their own. Companies want them to work safely and within budget. Snowflake's tools record which tools an agent uses, block actions outside its permissions, track costs, and pick smaller AI models for simple jobs. The company also says people should still approve big changes to core data systems. Snowflake advises companies to start with a small task, test it in real use, and then expand.

What to watch

  • Whether the Cocoa and Cowork bots move from private preview to wider public use.
  • How well the Business Ontology works in public preview across different departments.
  • Whether companies accept the rule that people must approve changes to core data systems.
  • How Snowflake's AI Gateway features compare with similar tools from other companies.

Glossary

Snowflake
A company that stores, manages, and analyzes large amounts of business data.
Business Ontology
A shared list of company terms and how they relate, so AI uses the same meaning for each term.

Ad

Add 150 words. Learn 10 a day.

Dicter word books send only the words you learn to your review notes.

More in this topic

AITechCrunch

Anthropic is still grappling with the economics of compute

An updated S-1 filing shows Anthropic's revenue rose to $17 billion in the latest quarter, up from $11.5 billion, but its compute costs grew to $10 billion and consumed 58% of revenue, up from 48%. The company said $5.0 billion of that compute came from short-term commitments that cost significantly more than multi-year deals. Anthropic expects compute to remain its largest operating expense for the foreseeable future.

AIThe Register

Forget retraining - AI can now tweak another AI's weights

MIT CSAIL researchers found that frontier AI models such as OpenAI's GPT-6 Astra can rewrite another model's weights directly rather than through training, altering its behavior. In one demonstration, Astra changed only 0.016 percent of Qwen3.5-0.8B's weights, improving MATH-500 performance by as much as sixfold and IFEval instruction-following roughly threefold, without gradient descent. The researchers call the approach "Weight-Space Programming" and plan to build a benchmark and harness for it.

AIThe Register

Open models can do 90 percent of agentic work at a twentieth of the cost of a frontier LLM: Gartner

Gartner analysts Mark Petty and Raymond Paquet argue that open-weight models can handle about 90 percent of routine enterprise agentic workflow steps at parity with closed models for a quarter of the price of the cheapest closed model. They say the best-case saving against a frontier model run on every step is up to a twentieth, though current open models are more verbose and retry more often, so the realistic saving is closer to five or six times. The analysts also predict that by 2030, 90 percent of enterprise operational agent workflow steps will run on open-weight models, and they argue that harnesses, which route work between models, are where the value in AI products lies.

AIZDNet Korea

Who Works Better? SK Telecom Employees' AI Agents Go Head to Head

SK Telecom held the finals of "AX Spark," an internal competition in which employees' self-built AI agents were judged on work performance and productivity gains. Five finalists, selected from 173 entries, are already used in real operations, including PYLON, which analyzes power usage and costs for about 180,000 telecom facilities nationwide, and M-TripleA, which tracks attack servers behind malicious smartphone apps. The company plans to extend the best cases across the organization, and it has over 5,000 internally developed work AI agents this year.